Development of an artificial neural network to maximize the reproducibility of dyeing polyamide fabrics

Development of an artificial neural network to maximize the reproducibility of dyeing polyamide fabrics | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Development of an artificial neural network to maximize the reproducibility of dyeing polyamide fabrics Francis Dalponte Voigt, Ricardo Antonio Francisco Machado This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3996611/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The Brazilian textile industry is an essential pillar of the country's economy, standing out globally as the fifth-largest textile hub and the fourth-largest in the clothing segment. However, one of the critical challenges faced by this sector is the reprocessing of fabrics, which leads to delivery delays, quality impacts, increased costs, and environmental impacts. Therefore, the aim of this study is to identify reprocessing in the dyeing process of a textile industry through preestablished patterns using a neural network. To achieve this goal, this research is being conducted in partnership with a company in the sector, focusing on data collection, preparation, processing, training and validating the neural network. Specifically, the focus is on the data collected from the production of polyamide, where approximately 95% of the reprocessing is classified as undefined, making the identification and precise resolution of these issues challenging. Thus, this research aims not only to enhance the efficiency of polyamide production but also to contribute to resource savings and compliance with environmental commitments, consolidating the concept of sustainability in the textile industry. The incorporation of artificial intelligence, such as neural networks, has emerged as an essential strategy to drive the textile industry toward more efficient and less impactful practices. Industrial Engineering Artificial intelligence Textile processing Polyamide. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 INTRODUCTION The Brazilian textile industry plays a key role in the country's economy, as Brazil stands out as the fifth largest textile hub in the world, being especially strong in the clothing segment, occupying the fourth position globally. In addition, the country is self-sufficient in cotton production, which contributes to the supply of raw material for this industry, which is one of the most complete chains in the West, ranging from the production of fiber to retail marketing. This vertical integration allows for greater efficiency and quality control throughout the production process. In addition, with the discovery of presalt, Brazil has the opportunity to become a major global exporter of synthetic fibers (ABIT, 2021). The textile and apparel sector played an important role in the world's manufacturing history, even before the industrial period, when products were manufactured by hand. However, with the course of industrial revolutions, the sector underwent transformations in the configuration of the business model, replacing manual work operations with factory operations (DUARTE, 2017). As a result, the company developed a productive infrastructure that was transformed into industrial parks to meet large-scale demands in domestic and foreign markets. This infrastructure currently constitutes a network of independent productive infrastructures, such as spinning, weaving, knitting, processing and clothing, constituting the textile chain (FUJITA; JORENTE, 2015). Within the textile infrastructure, the processing sector plays a key role in adding value and meeting a wide range of needs in this segment. However, it is important to note that this process faces significant challenges, which makes it a critical point throughout the production chain. This behavior is linked to several factors, including the long production cycle, the substantial need for water for the process, the release of toxic substances into the environment, and the use of chemical products such as dyes, heavy metals, acids, caustic soda, and sodium carbonate (CHEN., 2017; KU et al., 2020). These factors pose not only environmental challenges but also concerns about the safety and sustainability of the beneficiation process. In this context, on the one hand, the global demand for textile products has been steadily increasing, driven by the increase in population and economic development; on the other hand, it is evident that companies are increasingly aware of and concerned about the environmental impact resulting from their manufacturing operations (KUMAR et al., et al. 2020, ZHANG et al., 2021; SANDIN; PETERS, 2018). In response to this concern, several areas have adopted Cleaner Production (P + L) approaches, as well as Industry 4.0 concepts, and applied their technologies in production processes to provide innovation for their companies and make them more competitive (KUMAR et al., 2020). In addition to issues related to environmental impacts, textile companies face significant challenges in their production processes due to the need to reprocess the meshes. Reprocesses occur when the product varies in relation to the standards of color, touch, equalization, elongation and elasticity, touch and volume, compromising the efficiency of production and affecting customer service due to delays in the shipment of products and for presenting a different standard than desired at the time of purchase. Such a condition damages the image of a company and reduces its performance in the market, in addition to representing an increase in costs and environmental impact, by consuming more inputs and resources to obtain products. In this sense, the textile industry is generating increasing demands in the global market, and modern technologies have benefited this segment through the automation of long and complex processes, resulting in improvements in the speed, quality and cost of textile manufacturing (SCHWARZ; KOVAČEVIĆ, 2017 ). According to Falani et al. (2020), technological advances are one of the strategic factors for generating changes that contribute to the improvement of techniques, tools, and the use of new inputs. Fujita and Jorente (2015) point out that it is necessary to invest in technological innovation and in the generation of new knowledge through scientific development carried out to encourage research shared with the textile industry. In this scenario, artificial intelligence (AI), in essence, makes it possible to contribute to this need of the textile industry. AI ensures that systems make decisions independently, accurately, and supported by digital data. In an optimistic view, this multiplies the rational capacity of the human being to solve practical problems, simulate situations, think of answers or, more broadly, enhance the capacity to be intelligent. In view of the above, the present research aims to autonomously implement a neural network for prediction and correction, identifying reprocessing in the dyeing of a textile industry through preestablished patterns. In addition to the economic and social aspects, the success of this research increases its relevance by contributing to the saving of resources (water, inputs, energy), with a high environmental impact, reinforcing the company's existing commitments and completing the tripod of sustainability: social, economic and environmental. THEORETICAL BACKGROUND In this chapter, theoretical assumptions are presented based on the national and international literature on the processing process of textile industries with an emphasis on polyamide fibers, reprocessing, and artificial intelligence tools. 2.1 TEXTILE PRODUCTION CHAIN IN BRAZIL 2.1.1 Importance of the textile production chain in the Brazilian economy In general, the textile production chain is a vital cog in the Brazilian economy, contributing significantly to employment, regional development, exports, synergy between different sectors and technological advancement. Its key role in building a robust and diversified economy makes it an inalienable pillar in Brazil's economic landscape. In this context, Brazil is the 5th largest textile industry in the world and the 4th largest in the clothing segment, whose average production, in tons, was approximately 190 million in 2021 (IEMI, 2022). The textile industry promotes 1.36 million direct jobs, and in 2020, it represented 19.8% of the total number of workers allocated to industrial production and 5% of the total value of production in the Brazilian manufacturing industry (IEMI, 2022). According to the Institute of Industrial Studies and Marketing (2022), the Brazilian textile chain covers more than 24.6 thousand factories installed in the country, with approximately 3,030 textile companies and the remaining clothing companies. Along with these data, Brazil is the only country in the West that has a fully verticalized textile chain (ABIT, 2022). In this sense, due to the high relevance of the textile industry in the national economy, as it demands the intensive use of labor, this sector is a great generator of income and jobs for the country. In 2022, the textile industry produced 1.34 million formal jobs and more than 8 million indirect jobs, 60% of which involved women (IEMI, 2022). This fact is linked to the large per capita production of textiles in Brazil, which reached 9.9 kg per inhabitant, with a consumption of approximately 13.3 kg/inhabitant, and this difference is met by the international market (ABIT, 2022). 2.1.2 Stages of the production processes of the textile chain The beginning of the production process of the textile chain is first constituted by the production process of the fibers. Subsequently, several large-scale spinning companies produce wires. These companies have a large volume of capital and technology, are automated, and have a low labor ratio. The next stage of the chain includes the weaving, knitting and processing industries, which have characteristics similar to those of spinning (FALANI et al., 2020). In the last link of the chain are the clothing industries, which have a characteristic low use of capital and technologies but are very intense in the use of labor. For these reasons, they are mostly formed by micro- and small enterprises, most of which are informal (FALANI et al., 2020). Figure 1 shows the representation in the form of a flowchart, with the main stages of the textile production process. For ease of understanding, the flowchart contains only the macro steps, without any detail. Within the flowchart of the textile chain, the beneficiation process usually occurs after the production stage of the fibers and before the stage of making the garments or final textile products. Textile beneficiation involves a series of processes that prepare textile materials for the manufacture of finished products. These processes can include several steps, such as (i) dyeing, where dye is applied to color the fabric or yarn according to the design specifications; (ii) finishing, which involves processes such as bleaching, stamping, calendering and others that can affect the texture, gloss and feel of the material; (iii) washing, where materials are washed to remove impurities and residual chemicals from previous processes; (iv) mechanical finishing, which may include processes such as cutting, sewing and finishing of garments or textile products; and (v) quality inspection to finalize the process. 2.2 PROCESSING OF TEXTILE INDUSTRIES Textile processing, in general, can be defined as the process of aggregating several steps to improve the characteristics of fabrics, fibers and yarns, giving fabrics unique particularities. It is composed of different steps and processes (SAMSAMI et al., 2020). Textile processing consists of the stages of primary processing (preparation), secondary processing (dyeing) and tertiary processing (finishing). Figure 2 shows a flowchart of the main steps and processes involved in textile processing. 2.2.1 Pretreatment Primary processing, or pretreatment, is the initial stage of the processing process. All textile substrates need this treatment before moving on to downstream processes, such as dyeing, printing or finishing. Madhu and Chakraborty (2017) noted that this processing is the first treatment applied to a textile material (yarn or fabric) and aims to obtain a final product with better acceptance in the market. The operations to be carried out follow a sequence and must be preserved to ensure good results. This process is also known as purging or boiling and is responsible for the removal of oils from textile fibers; for this purpose, oil emulsifier products, also called surfactants or detergents, which can be products with nonionic or anionic characteristics, are used. This process is usually carried out at temperatures of 50 to 60°C (SALEM et al., 2005). In general, the effective removal of impurities occurs with the use of 3-6% sodium hydroxide, calcium hydroxide or sodium carbonate. In addition, the proper choice of textile auxiliaries in an alkaline bath is essential for good purging processing. These include sequestering or chelating agents, such as ethylenediaminetetraacetic acid (EDTA), to solubilize insoluble inorganic substances present in hard water, and surfactants, such as sodium lauryl sulfate, which serves as a detergent, dispersing agent, and emulsifying agent to remove unsaponifiable waxes (BARANI, MONTAZER, 2008). Purging is also applied to synthetic fibers that take on a yellow tint due to overheating or the buildup of impurities during manufacturing. This process is gentler than that of natural fibers (BROADBENT, 2011). In synthetic fibers, washing is carried out with soap or detergents containing lower amounts of alkaline solution (0.1-0.2% sodium carbonate) (VIGO, 2002). Moreover, cotton fibers usually require prebleaching unless they are dyed in the dark. The purpose of bleaching is to remove impurities that mask the natural whiteness of the fibers. Oxidizing agents, such as hydrogen peroxide, sodium hypochlorite, and sodium chlorite, are used for this purpose. On an industrial scale, hydrogen peroxide is the most common oxidizing agent. In addition, the process can be conducted by the depletion method or on a continuous basis (ADANUR, 2017). After the preparation stage, most of the textile substrates are subjected to dyeing, where they receive the desired color. 2.2.2 Mercerization In this process, a physicochemical treatment involving the impregnation of the textile material under tension with alkaline solutions under strictly controlled temperature and concentration conditions occurs. The purpose of this processing is to increase the brightness of the product, as well as the absorption of water and dyes, as well as the tensile strength and dimensional stability (ADANUR, 2017). Mercerization is usually applied to yarns and flat fabrics, as well as cellulosic fibers, especially cotton, and can be performed with sodium hydroxide at temperatures ranging from 10 to 18 °C, followed by rinsing and neutralization (BROADBENT, 2011). 2.2.3 Bleaching Bleaching is the chemical process used in the bleaching of textile materials, the ultimate goal of which is bleaching. In addition, this process can also be used in articles that require optical brightening to enhance the degree of whiteness (BROADBENT, 2011). This process is employed in different textile products, such as yarns, flat fabrics and knitwear; oxidative processes, such as sodium hypochlorite, ozone and hydrogen peroxide; and reducing processes, such as sodium hydrosulfite, sodium sulfoxylated formaldehyde, sodium bisulfite and thiurea dioxide. 1.1.1 Thermosetting Thermosetting can be considered a pretreatment of the fabric, as it can be performed before dyeing to provide dimensional stability. This process is carried out only on synthetic fibers such as polyester and its blends and on articles with elastane (RUSCHIONI; ALFIERI, 2010). 1.1.2 Dyeing Dyeing is a process known for the homogeneous coloring of textile substrates with the use of dyes. This process is also known as secondary beneficiation. In general, this process is divided into three stages, mediated by the processes of migration, absorption and fixation of the dye. In the first step, the dye migrates from the medium in which it is diluted to the surface of the fiber. Subsequently, the process of adsorption occurs in the surface layers of the textile material, and then the dye diffuses inside the fiber and is fixed by different types of bonds (ionic, van der Waals, covalent) depending on the type of material used (ADANUR, 2017). Temperature is a primary factor in these steps due to the influence of chemicals and the mechanical action caused by the agitation of the dyeing bath of the textile substrate during processing (SALEM et al., 2005). In addition to the temperature, the dyeing speed must be carefully controlled in the form of a curve, considering the substrate, products and machine. This curve should express the dyeing time as a function of the total percentage of dye that will assemble when it reaches equilibrium as well as the time required to reach half depletion. Many critical factors are important for good equalization and reproducibility (SALEM et al., 2005). Compared to wool, the diffusion of acid dyes in polyamide is slower, and the limited number of sites loaded in polyamide can also cause problems in dyeing deep shade blends, where individual dyes compete for available sites. Under these conditions, the fastest diffusing dyes can block the entry of a second component, and a hue is not achieved (ADANUR, 2017). The greater or lesser saturation of the fiber depends on the percentage of terminal amine groups (AEGs), and these in polyamide fibers are limited in number. Saturation also depends on the number of sulfonic groups in the dye that will react with the terminal amino groups. The greater the number of sulfonic groups in a dye molecule is, the lower the saturation: a dye molecule will occupy more than one terminal amine group. Thus, a trisulfonic dye occupies three terminal amine groups. Therefore, when dyes are combined with mono- and trisulfonic agents in the same recipe, the monosulfonic dye tends to mount first on the fiber, occupy the amine groups (which are limited) and block the assembly of the trisulfonic dye, which remains in the bath (GONDIM, 2016; SALEM, et al., 2005). The permanence of the dye in the fiber is affected by several factors, among which we can highlight the (i) vibration of the molecular structure of the fiber, at each moment, taking new configurations; (ii) constant bombardment of the dye by water molecules during dyeing, making it difficult to fix it in the fiber; and (iii) as the temperature of the system increases, the vibration of the fiber molecules and the bombardment of the water molecules increase (SALEM et al., 2005). Polyamide reprocessing in the textile industry mainly occurs through dyeing processes, where the goal is to change the color or physical properties of polyamide fabrics. This reprocessing may be necessary for a variety of reasons, such as color correction, improved dyeing quality, shade adjustments, or even to correct imperfections that occurred during the initial process (SU et al., 2007). 1.1.3 Finishing Finishing is intended to provide desirable properties to the fabric, including soft touch, wrinkle resistance, and impermeability, among other functional characteristics. This stage involves procedures such as calendering, which uses heated cylinders to smooth and shine the fabric, as well as wire to remove lint and surface impurities. Chemical agents, such as water repellency, stain resistance, or antimicrobial agents, are commonly applied at the finish of a material to impart specific properties to the fabric. These chemical treatments are applied in a controlled manner to ensure the desired functionality without compromising the integrity of the tissue (SALEM et al., 2005). The efficient and accurate integration of these steps is crucial to achieving final textile products that meet the requirements for quality, aesthetics, and functionality. The combination of printing and finishing techniques in textile processing plays a decisive role in the variety and quality of textile products available on the market, reflecting the importance of these processes in the contemporary textile industry. 1.2 ARTIFICIAL INTELLIGENCE AI represents the pinnacle of technological innovation, being a multidisciplinary field that seeks to replicate the human capacity for learning, reasoning, and decision-making through algorithms and computational systems. Based on concepts such as machine learning, neural networks, and natural language processing, AI has gained a central role in several spheres of modern life (SU et al., 1993). Its ability to analyze large volumes of data, identify complex patterns, and make autonomous decisions is redefining industries, driving automation, and transforming the way we interact with technology, from virtual assistants to advanced medical diagnostic systems. The rapid evolution and application of AI are shaping not only the age of technology but also redefining the boundaries of human knowledge and its practical applications (MUKHERJEE; BHA, 2023). Theoretical assumptions about AI are critical to understanding the conceptual basis behind the development and application of this scenario. These assumptions form the theoretical foundation that guides the creation of intelligent systems and algorithms (Schädler and Wysotzki, 1999). They encompass areas such as machine learning, which focuses on the ability of systems to learn and improve from experience; computational logic, which aims to understand how systems can represent knowledge and reason logically; and neural networks, which are inspired by the functioning of the human brain to create highly efficient computational models (HIMMELBLAU, 2008). Understanding these assumptions is crucial for exploring not only the potential but also the limits and challenges of AI in diverse application fields, including textile beneficiation (SIKKA et al., 2022). 1.2.1 Artificial neural networks (ANN) ANNs are computational elements inspired by biological neurons, which, when networked, can reproduce some characteristics of intelligent processes (RAMESH et al., 2004). ANN learning has emerged as a dominant framework in the present day, generating advances in a wide range of applications, including computer vision, natural language processing, and strategic games (SMIRNOV et al., 2014). Some key ideas in this field can be traced back to the human brain, and there is currently a continuous exchange of ideas from neuroscience to the field of artificial intelligence (HASSABIS et al., 2017). At the same time, learning about AI offers powerful new tools for systems neuroscience. In fact, advances in computer vision, especially ANNs, have revolutionized image and video data processing. Uncontrolled behaviors over time, such as the micromovements of animals in laboratory experiments, can now be efficiently tracked and quantified with the help of this technology (MATHIS et al., 2018). Inspired by the ability of humans and other animals to perform functions such as processing sensory information and interacting with poorly defined environments, engineers are concerned with developing artificial systems capable of performing similar tasks. Skills such as incomplete or inaccurate information processing ability and generalization are desirable properties in such systems (LIU et al., 2020). In this sense, ANNs are computational techniques that have the ability to solve problems through simple circuits that simulate the functioning and behavior of the human brain. They present a model inspired by the neural structure of intelligent organisms, which acquire knowledge through experience, that is, by learning, making mistakes and making discoveries. An artificial neural network can have hundreds or even thousands of processing units, while a mammal's brain can contain many billions of neurons (LIU et al., 2020) However, it is important to point out that there are differences between human nerve cells and artificial neuron models. However, the principle of information transfer is the same. Analogous to the human brain, ANNs can interact with and adapt to the external environment. These characteristics give ANNs multidisciplinary importance, which is why this tool has been gaining prominence in different areas of knowledge, such as engineering, mathematics, physics, and computer science (LIU et al., 2020). The mathematical model of the neuron, which encompasses the main characteristics of a biological neural network, parallelism and high connectivity, was proposed in 1942 by McCulloch and Pitts, where the researchers designed the structure known as the first neural network known worldwide as the MCP model (McCulloch-Pitts). The MCP model is a simplification of the biological neuron model, which considers the neuron as a binary information processing unit with multiple binary inputs and a single binary output, showing that these units are capable of performing different logical operations. Figure 3 shows the general model of the artificial neuron, where x1, x2, and xn represent the input signals; W1, W2, and wn are the weights or synaptic connections; BIAS represents the neuron's activation threshold; u is the output of the linear combiner; (f) is the activation function (limits neuron output); and y is the neuron's output signal. The operation of a cell in a neural network can generally be described as follows: (i) signals are presented to the input; (ii) each signal is multiplied by a weight, which indicates its influence on the cell's output; (iii) the weighted sum of the signals is executed, which produces a level of activity; and (iv) when this level exceeds a threshold, the unit produces an output (FINOCHIO, 2014). Over time, ANNs have undergone a process of evolution. First, in 1948, N. Wiener coined the word cybernetics to describe, in a unified way, control and communication in living organisms and machines. Subsequently, in 1949, D. O. HEBB proposed a hypothesis about how the strength of synapses in the brain changes in response to experience. In particular, he suggested that connections between cells that are activated at the same time tend to strengthen, while other connections tend to weaken. This hypothesis had a decisive influence on the evolution of the theory of learning in artificial neural networks (HOPFIELD, 1984). Later, in 1957, Rosenblatt introduced a new approach to the pattern recognition problem with the development of the perceptron (Figure 4). This same researcher also proposed an algorithm for adjusting the perceptron weights and proved their convergence when the patterns are linearly separable. At approximately the same time, B. WIDROW (1965) and his collaborators developed the adaptive linear element (Adaline) (STEINBUCH and WIDROW, 1965). The interesting conclusion adopted by Hopfield was that such equilibrium states can be used as memory devices. Unlike that used by conventional computers, in which access to stored information occurs through an address, access to the contents of the memory of a Hopfield network is given by allowing the network to evolve over time to one of its equilibrium states. Such memory models are called content-addressable memories (HOPFIELD, 1984). However, effectively placing the RNA area as one of the priorities in obtaining resources was the priority of developing a method for adjusting the parameters of nonrecurrent multilayer networks. Therefore, one of the most commonly used architectures is the multilayer perceptron network, which uses the back propagation algorithm, proposed in 1986 by Rumelhart, McClelland and Williams (BEALE; JACKSON, 1990), as a learning rule. 1.2.1.1 Characteristics of artificial neural networks The most important characteristics of artificial neural networks are numerous: (i) the ability to learn the relationships between a set of input data, also called training examples, and thus improve their performance. This ability is due to the training or learning algorithm, which will be discussed below; (ii) the ability to generalize learning to new examples. Networks can provide responses similar to those for which they were trained, for examples not presented in the training. The ability to extract the essence of a dataset and learn from incomplete information; (iii) robustness and fault tolerance: the elimination of some neurons does not substantially affect their overall performance; (iv) flexibility: can be adjusted to new environments through a learning process, being able to learn new actions based on the information contained in the training data; (v) processing of uncertain information: even if the information provided is incomplete and affected by noise, it is still possible to obtain correct reasoning; and (vi) parallelism: an immense number of neurons are active at the same time. There is no restriction of a processor that necessarily works one instruction after another (FINOCHIO, 2014). The above characteristics make neural networks especially attractive for applications in nonlinear systems that have a large volume of parameters and can undergo temporal variations, as is the case for textile processing. ANNs can also be classified according to their architecture, for example: 1. Perceptron 2. Feed Forward Neural Network 3. Multilayer Perceptron 4. Convolutional Neural Network 5. Radial Basis Functional Neural Network 6. Recurrent Neural Network 7. LSTM – Long short-term memory 8. Sequence-to-Sequence Models 9. Modular Neural Network The classifications were cited in English on purpose, as this is the way they are usually referenced. In addition, neural networks can still be combined with each other, fed back and still have prior knowledge. Neural networks that have structures such as 3 layers, the first layer, the input layer, the second layer (intermediate layer), the processing layer, and the third layer, the output layer, are usually called black box neural networks. Figure 5 shows this type of structure. A neural network has elements called neurons, which can be completely or partially interconnected. Figure 6 shows the structure of an artificial neuron. Figure 6 - Structure of an artificial neuron, where p = process values or input data; w = is called weight and represents the relevance of a given input variable in the behavior of the process; b = bias, which are the adjustments that are necessary to correct some deviation; F = is the activation function of the neuron; and a = is the value predicted by the neuron. Equation 1 represents the activation function of the neuron, F: a = F(w.p + b) (1) where: a = output data or predicted values p = input data w = weights b = bias F = activation function The w and b parameters should be adjusted so that they represent the process data with relative accuracy. In the tuning phase, we say that the neural network is being trained, which is nothing more than the employment of an optimization algorithm to adjust the values of w and b in such a way as to predict the output data (a) with the least possible deviation from the input values. This is where machine learning terms come in. Neurons can also have multiple inputs, as shown in Figure 7. If we add several neurons to each layer, we consequently increase the "intelligence" of the neural network, as there are several activation functions and several parameters to be estimated. Figure 8 shows a schematic of an artificial neural network with several layers. However, it is necessary to keep in mind that when referring to industrial processes, we are dealing with numerical information (temperature, pressure, molar fractions, etc.) and subjective information (good, bad, more or less, etc.). Then, it is necessary to create a scale so that the subjective information can be quantified numerically. The numerical, quantitative information must have the same scale; otherwise, an adjustment considered excellent for the set of temperatures, which presents only 0.5 K of deviation, is a tremendous error when we evaluate molar fractions. In addition, there is the problem of properly handling information before processing it. For the sake of objectivity, we will deal in this document only with black box artificial neural networks because we do not yet have enough information about the process to be able to insert prior knowledge into the structure we are going to use. Returning to the issue of data processing, it is initially necessary to define the type of activation function to be used. The most common are described in Chart 1. Table 1 - Types of activation functions Due to its versatility and nonlinear characteristics, which enable the representation of extremely complex data, the tangential sigmoid activation function is preferentially used. Especially when you have sparse systems. In textile processing, we find both nonlinear and sparse systems. Therefore, the tangential sigmoid activation function is the first to be employed. However, we still need to take care of the data before we start training the network. We have already defined the activation function. Looking at their behavior, the limits are -1 and +1. That is, we need to scale the data so that they are all between -1 and +1. Previous studies have shown that the output of the activation function should not be saturated. In other words, we should not work at the lower and upper limits. Haykin (1999) recommends, in the case of the tangential sigmoid activation function, that scaling be performed between -0.8 and +0.8, according to Equations 2 and 3. For the input data: With the elements of the input and output datasets ranging from -0.8 to +0.8, we can then move on to the training phase. In terms of training, various optimization algorithms can be used, with some being faster and others being slower. The most common is the use of step-down algorithms. Nasra et al. (2016) explored different training algorithms and concluded that the Levenberg–Marquartd method has a higher convergence speed, which is why it is used in this work. Other optimization algorithms can be found in Yin et al. (2003). Figure 9 shows the flowchart of the training phase (machine learning). In the next subitem, we will show some uses of artificial intelligence in textile processing. 1.2.2 Artificial intelligence and textile processing The application of AI in the textile processing sector represents a significant evolution in the modern industry landscape. The integration of intelligent algorithms, machine learning, and automation has revolutionized processes from design conception to final production. This fusion of AI and textile beneficiation offers unprecedented opportunities to optimize efficiency, accuracy, and sustainability at every stage of the process, including printing, dyeing, finishing, and quality control. This innovative convergence not only reshapes the way fabrics are produced but also redefines standards for quality, customization, and speed of response to market demands (SIKKA et al., 2022; CHATTOPADHYAY; GUHA, 2004). Specifically, in the case of ANNs, notable advances have been identified in the field of textile processing. Currently, these nets are applied in a significant way in the quality control of fabrics, allowing the identification of defects, dyeing patterns and finishing failures in a much more accurate and efficient way than conventional methods. In addition, ANNs have been used in the optimization of process parameters such as temperature, dyeing time and chemical composition. In addition, RNA-based models play a crucial role in predicting the final properties of tissues, including strength, durability, and shrinkage behavior. This enables precise adjustments in manufacturing processes, contributing to improving the quality of textile products (SIKKA et al., 2022). Table 1 presents a general summary of the contributions of ANNs to the textile industry. Table 1 - Synthesis of studies developed with RNA and the textile industry. Authors Objective of the work Bahlmann et al. (1999) Develop RNA to establish an automated control in textile seams. Hui et al. (2007) Develop RNA for textile seam performance. Tiwari et al. (2023) Application of RNA to evaluate the performance of produced tissues. Doran et al. (2019) Develop RNA for prediction of cotton and spandex yarns. Jeyaraj et al. (2019) Use of RNA for detection of defects in the tissue manufacturing process Li et al. (2021) Identification of waste generated by the textile industry using RNA Source: The author, (2023). Despite these advances, some challenges remain, such as the need for more robust and representative datasets, as well as the interpretability of models for industrial application. Furthermore, the potential of neural networks to simulate complex processes is still evolving (SIKKA et al., 2022). MATERIALS AND METHODS This research is being developed in partnership with a textile company located in Indaial/SC and is composed of different stages linked to data collection and preparation, development of the ANN, and training and validation of the developed ANN. 2.1 Data collection and preparation The industrial park of the company in question is composed of knitting, dyeing and finishing, but the research is being carried out in the dyeing sector due to its high percentage of reprocessing. A survey of the process variables was carried out for the selection of ANN input data, collecting a wide variety of data and samples of polyamide mesh fabrics and identifying different variations, including products with and without reprocessing. Then, the dataset was treated according to the datasets obtained, and an initial architecture for the artificial neural network was defined, which was trained until it could represent the real data with relative accuracy. After the training phase, the network was validated with a new set of data (a step called cross-validation) to predict possible deviations in the quality of the finished product and, consequently, reprocessing. The software used was MATLAB. 2.2 DEVELOPMENT OF THE ARTIFICIAL NEURAL NETWORK The network architecture is determined by determining the number of layers, the number of neurons in each layer, and the proper activation function. Although there are some approaches in the literature to design the architecture of a neural network, we try not to delve into any of them because they have no theoretical foundation. Therefore, the network was designed by trial and error. 2.3 ARTICAL NEURAL NETWORK TRAINING AND VALIDATION 2.3.1 Data Preparation The training, validation, and testing datasets were generated in MATLAB. 2.3.2 Adjustments and Optimization Based on the results of the assessment, adjustments were made to the network architecture as needed. Weights and bias are evaluated by the principal component analysis technique. Those neurons that have weights and bias with negligible values are excluded from the network architecture, and the network is retrained. The procedure is repeated until the result is reproduced. When this occurs, we have defined the number of layers and neurons sufficient for the network to adequately represent the process. 2.3.3 Implementation in a Production Environment After validating the performance of the network, it will be implemented in the production environment to demonstrate its operation in the plant. 2.3.4 Maintenance and Upgrading Network performance in production is monitored and adjusted as needed. If necessary, the network will be retrained with newer data to improve performance over time. RESULTS AND DISCUSSION The following are preliminary results obtained regarding the data collected from the reprocessing of the textile processing sector of the company studied. 3.1 Data collection Table 2 shows the fiber production and reprocessing conducted for each type of fiber. The percentages of reprocessed polyamide, PES/TXT and viscose fibers were on the order of 75%, 12% and 5%, respectively (Figure 10). In general, the higher reprocessing rates observed for the polyamide, PES/TXT and viscose fibers can be attributed to the inherent complexities of the production processes and the intrinsic properties of these materials. Moreover, this behavior can be attributed to the complex interaction between the intrinsic properties of these materials and the production procedures. Polyamide, known for its strength, can face challenges during production due to its nature, while PES/TXT, possibly a combination of polyester with other textile fibers, can be susceptible to specific manufacturing flaws. Viscose, despite its favorable properties, can be sensitive to certain production processes. The complexity of manufacturing processes and the sensitivity of these materials to variations in processing conditions also contribute to these higher reprocessing rates. Table 2 - Amount of production and reprocessing of fibers produced from January to July 2023. Fiber Production (kg) Reprocessing (kg) Total Produced 3.332.607,900 392.621,371 Polyamide 2.489.445,282 295.008,101 PES/TXT 484.226,291 49.206,986 Viscose 102.047,540 19.034,677 Cotton 99.666,517 17.067,888 Polyester 79.544,618 6.376,200 Textured PES 69.155,484 5.927,519 Special polyamide 8.491,068 - Differentiated 31,100 - Source: The author, (2023). Specifically, for the polyamide fibers, Table 3 shows the causes of the reprocesses generated. Notably, approximately 95% of these occurrences are classified as undefined, indicating a lack of clarity about their specific causes, which compromises the ability to identify and solve such problems. This lack of precision in identifying the causes prevents a targeted and effective intervention to address the issues related to polyamide reprocessing. The high rate of undefined reprocessing in polyamide fibers could be due to multiple factors. The complexity of the production processes of this fiber, which involves delicate chemical and mechanical steps, can lead to a wide range of possible failures, from variations in chemical composition to inadequacies in temperature or pressure during the process. In addition, the difficulty of monitoring all the variables involved can contribute to inaccuracies in identifying the causes, creating gaps in the understanding of the events that lead to reprocessing. This lack of clarity can be compounded by the absence of robust tracking systems or detailed recording methods, making it difficult to accurately attribute the causes of quality issues, thus resulting in this high percentage of undefined occurrences. Table 3 - Causes of reprocesses generated in polyamide production. Causes Reprocessing (kg) Reprocessing (%) Grand total 295.008,101 - Indefinite 281.531,066 95 Operational failure 3.452,581 1 Technical Area Test 3.438,815 1 Wrong recipe 2.496,690 1 Raw Material 2.296,310 1 Electrical problem 643,540 0 Mechanical problem 394,680 0 Raw material-yarn 377,520 0 Wrong concentration 147,909 0 Test new process 32,610 0 Source: The author, (2023). The company has several dyeing lines in such a way that only one machine will be selected for the purpose of concluding this dissertation. The inclusion of all production lines, with their variables, would make it impossible to complete the present work within the regimental deadline. However, this work has an academic and didactic nature and can be explored by companies with enormous benefits. In the present work, the interest is to evaluate the input variables and the finished product in a static neural network. Intermediate steps, which can be performed by the company, including moving the architecture to a dynamic network, are not included. For the neural network to be applied, each qualification is assigned a numerical code. The code must be a number with a certain variation. For example, if the product should have a Ferrari red color coded, for example, as a number 5, and the result is 5.1, it will be essential that a tolerance index be defined by the company. The color is visually evaluated by the quality inspector, and in case of doubt, the spectrophotometer is read to verify that it is within the range of the panettone color chosen by the customer. 3.2 IMPLEMENTATION OF THE ARTIFICIAL NEURAL NETWORK Initially, an ANN was implemented with data from MATLAB itself to test the approach it intended to employ. Figure 11 shows the structure of the network, whose structure was created randomly. There are 13 input data points, an intermediate layer with ten neurons, a tangential sigmoid activation function, and an output layer. In addition to the MATLAB software allowing the insertion of scripts, it is possible to visualize all events. Figure 12 shows the indexes related to network training. Figure 12 - Evolution of network performance indexes. (a) Supervision of learning. (b) Histogram of error in relation to the predicted value and the wrong values, (c) evolution of the correlation coefficient, and (d) better performance obtained in relation to the actual data. We can conclude that the selected computational tool allows us to implement the proposed neural network with relative agility and the achievement of fast and visual results. To improve the visualization of the results, it is possible to increase the number of variables to be monitored and make the necessary adjustments without having to wait for the completion of the complete training of the network (in this case, the training is called epochs or epochs, which is the number of times the optimization algorithm was used to determine the parameters of the network (bias and weights). The decision variables are listed in Figure 13. Figure 13 - Variables for decision-making about the network architecture and its performance during the training phase. By analyzing the results, we can conclude that the selected platform as well as the approach are suitable for the completion of the work. CONCLUSIONS The higher rates of reprocessing identified in polyamide fibers suggest distinct challenges in their respective production chains. As synthetic fibers, they undergo complex processes and can generate out-of-specification products due to variations in chemical and physical parameters. These findings also indicate that the high percentage of undefined occurrences of reprocessing in polyamide fibers reveals challenges in accurately identifying the causes, possibly due to the complexity of the processes and the difficulty in monitoring all the variables involved. The integration of advanced technologies, such as ANNs, could play a crucial role in reducing reprocessing in polyamide manufacturing. The implementation of RNA can offer an innovative approach to analyzing vast datasets sourced from production processes, identifying subtle patterns and correlations between variables that can be difficult to detect by conventional methods. By training neural networks with detailed information about the polyamide manufacturing process, it is possible to develop predictive models capable of anticipating possible failures, minimizing the occurrence of out-of-specification products. This proactive approach could significantly contribute to reducing reprocessing, optimizing process efficiency and improving the quality of final polyamide products. Therefore, incorporating artificial intelligence technologies such as artificial neural networks has emerged as a vital necessity to propel the PA manufacturing industry toward more efficient and lower-waste practices. As previously discussed, the selected tool proved to be feasible for the approach intended in the present study. Declarations CONFLICT OF INTEREST On behalf of all the authors, the corresponding author declares that there are no conflicts of interest. References ABIT- Associação (2023) Brasileira da Indústria Têxtil e de Confecção. Perfil do Setor 2022. Disponível em: https://www.abit.org.br/cont/perfil-do-setor . Acesso em: 12 set ABIT- Associação Brasileira da Indústria Têxtil e de Confecção. Indústria têxtil brasileira (2018) Disponível em: https://www.abit.org.br/ . Acesso em: 12 set. 2023 ADANUR, Sabit (2017) Handbook of industrial textiles. Lancaster: Wellingt Sears Co, 839 p BAHLMANN C, HEIDEMANN G (1999) ; RITTER, Helge. Artificial neural networks for automated quality control of textile seams. Pattern Recognition, [S.L.], v. 32, n. 6, p. 1049–1060, jun. Elsevier BV. http://dx.doi.org/10.1016/s0031-3203(98)00128-9 BARANI H, MONTAZER, Majid A Review on Applications of Liposomes in Textile Processing. Journal Of Liposome Research, [S.L.], v. 18, n. 3, p. 249–262, jan. 2008. Informa UK Limited. http://dx.doi.org/10.1080/08982100802354665 BEALE R, JACKSON T (1990) Neural Computing - An Introduction. Crc, Eua, p 256 Arthur BROADBENT (2011) D. Basic Principles of Textile Coloration. Society Of Dyers And Colorizts, West Yorkshire, p 579 CHATTOPADHYAY R, GUHA A ARTIFICIAL NEURAL NETWORKS: applications to textiles. Textile Progress, [S.L.], v. 35, n. 1, p. 1–46, mar. 2004. Informa UK Limited. http://dx.doi.org/10.1080/00405160408688961 CHEN C-L, 3 abr (2019) Value Creation by SMEs Participating in Global Value Chains under Industry 4.0 Trend: case study of textile industry in taiwan. Journal Of Global Information Technology Management, [S.L.], v. 22, n. 2, p. 120–145,. Informa UK Limited. http://dx.doi.org/10.1080/1097198x.2019.1603512 DORAN EC, SAHIN, Cenk (2019) The prediction of quality characteristics of cotton/elastane core yarn using artificial neural networks and support vector machines. Textile Research Journal, [S.L.], v. 90, n. 13–14, p. 1558–1580, 27 dez. SAGE Publications. http://dx.doi.org/10.1177/0040517519896761 DUARTE, Adriana Yumi Sato. Proposta de integração entre ferramentas de avaliação de ciclo de vida do produto e Indústria 4.0 (Industrie 4.0): estudo de caso da indústria têxtil e de confecção brasileira (2017) 125 f. Tese (Doutorado) - Curso de Engenharia Mecänica, Centro Tecnológico, Universidade Estadual de Campinas, Campinas, 2017 FALANI LAraújo (2020) Catia Rosana Lange de; FORNO, Ana Julia dal. Mapeamento da literatura sobre as tecnologias da indústria 4.0 no segmento têxtil brasileiro. Brazilian J Dev 42437–42452 Brazilian Journal of Development. http://dx.doi.org/10.34117/bjdv6n7-019 FINOCCHIO MA (2014) F. Noções de redes neurais artificiais, Universidade Tecnológica Federal do Paraná FUJITA, Renata Mayumi Lopes; JORENTE, Maria José. A indústria têxtil no Brasil: uma perspectiva histórica e cultural. Modapalavra E-Periódico, [S.L.], v. 8, n. 15, p. 153–174, 1 jan. 2015. Universidade do Estado de Santa Catarina. http://dx.doi.org/10.5965/1982615x08152015153 GONDIM, Mauricio Lopes. Estudos comparativo do ângulo de contato em tecido de malha de poliamida tratados com plasma. (2016) 68 f. Dissertação (Mestrado) - Curso de Têxtil e Moda, Escola de Artes Ciências e Humanidades, Universidade de São Paulo, São Paulo, 2016 HASSABIS D, KUMARAN, Dharshan; SUMMERFIELD C (2017) ; BOTVINICK, Matthew. Neuroscience-Inspired Artificial Intelligence. Neuron, [S.L.], v. 95, n. 2, p. 245–258, jul. Elsevier BV. http://dx.doi.org/10.1016/j.neuron.2017.06.011 Patrick HUI, CHAN CL, Keith CC, YEUNG KW, NG, Frency SF (2007) Application of artificial neural networks to the prediction of sewing performance of fabrics. International Journal Of Clothing Science And Technology, [S.L.], v. 19, n. 5, p. 291–318, 9 out. Emerald. http://dx.doi.org/10.1108/09556220710819500 IEMI- Instituto de Estudos e Marketing. O Panorama Setorial da Indústria Têxtil Brasileira (2022) IEMI. Disponível em: https://www.iemi.com.br/brasil-textil-2022/ . Acesso em: 26 out. 2023 JEYARAJ PR (2019) ; NADAR, Edward Rajan Samuel. Effective textile quality processing and an accurate inspection system using the advanced deep learning technique. Textile Research Journal, [S.L.], v. 90, n. 9–10, p. 971–980, 23 out. SAGE Publications. http://dx.doi.org/10.1177/0040517519884124 NERURKAR JOSHIM, Madhura BADHE, Pallavi, ADIVAREKAR, Ravindra. Scouring of cotton using marine pectinase. Journal Of Molecular Catalysis B:, Enzymatic [S.L.], v. 98, p. 106–113. Elsevier BV. http://dx.doi.org/10.1016/j.molcatb.2013.10.010 KANT, Rita (2012) Textile dyeing industry an environmental hazard. Natural Science, [S.L.], v. 04, n. 01, p. 22–26, Scientific Research Publishing, Inc. http://dx.doi.org/10.4236/ns.2012.41004 Kang-Ting KUC-Chun;CHIENC-Fu;MA Digital transformation to empower smart production for Industry 3.5 and an empirical study for textile dyeing. Computers & Industrial Engineering, [S.L.], v. 142, p. 106297, abr. 2020. Elsevier BV. http://dx.doi.org/10.1016/j.cie.2020.106297 KUMAR H, KAUR, Gagandeep PRIYA Influence of tetra ethyl ammonium bromide (C2H5)4NBr on the aggregation behavior of surface active ionic liquid 1-tetradecyl-3-methylimidazolium bromide [C14mim][Br]. Journal Of Molecular Liquids, [S.L.], v. 313, p. 113431, set. 2020. Elsevier BV. http://dx.doi.org/10.1016/j.molliq.2020.113431 LI W, WEI Z, LIU, Zhengdong DU, Yujun; ZHENG, Jiahui; WANG H, ZHANG (2021) Shuo. Qualitative identification of waste textiles based on near-infrared spectroscopy and the back propagation artificial neural network. Textile Research Journal, [S.L.], v. 91, n. 21–22, p. 2459–2467, 9 abr. SAGE Publications. http://dx.doi.org/10.1177/00405175211007516 CHANG LIUJ, Huihong FORREST, YANG JY-L Baohua. Influence of artificial intelligence on technological innovation: evidence from the panel data of China's manufacturing sectors. Technological Forecasting And Social Change, [S.L.], v. 158, p. 120142, set. 2020. Elsevier BV. http://dx.doi.org/10.1016/j.techfore.2020.120142 MADHU A, CHAKRABORTY JN (2017) mar. Developments in application of enzymes for textile processing. Journal Of Cleaner Production, [S.L.], v. 145, p. 114–133, Elsevier BV. http://dx.doi.org/10.1016/j.jclepro.2017.01.013 MALUF E, KOLBE W (2003) Dados Técnicos para an Indústria Têxtil. 2.ed. São Paulo: IPT – Instituto de Pesquisas Tecnológicas do Estado de São Paulo. ABIT – Associação Brasileira da Indústria Têxtil e de Confecção MATHIS MR, KHETERPAL S; NAJARIAN, Kayvan. Artificial Intelligence for Anesthesia: what the practicing clinician needs to know., Anesthesiology (2018) [S.L.], v. 129, n. 4, p. 619–622, 1 out. Ovid Technologies (Wolters Kluwer Health). http://dx.doi.org/10.1097/aln.0000000000002384 PIMENTEL, Renê C, Iran Siqueira (2011) ; LIMA,. Relação trimestral de longo prazo entre indicadores de liquidez e de rentabilidade: evidência de empresas do setor têxtil. Revista de Administração, [S.L.], v. 46, n. 3, p. 275–289, 14 set. Business Department, School of Economics, Business & Accounting USP. http://dx.doi.org/10.5700/rausp1012 RAMESH A, KAMBHAMPATI C, MONSON, Jrt, DREW (2004) Pj. Artificial intelligence in medicine. Annals of The Royal College Of Surgeons Of England, [S.L.], v. 86, n. 5, p. 334–338, 1 set. Royal College of Surgeons of England. http://dx.doi.org/10.1308/147870804290 RUSCHIONI PB, ALFIERI P (2010) ; Estudo da influência da temperatura de termofixação 76 nas propriedades físicas de malhas de poliamida 6.6. Química Têxtil n ° 99. São Paulo. Junho, Pg. 52–61 SALEM, Vidal (2010) Tingimento têxtil: fibras, conceitos e tecnologias. Edgard Blucher, São Paulo, p 297 DE SALEMV, MARCHI A (2005) MENEZES, Felipe Gonçalves de. O beneficiamento têxtil na prática, 1 edn. Golden Química do Brasil, São Paulo SAMSAMI S, MOHAMADIZANIANI M, SARRAFZADEH M-H, Eldon RENE, FIROOZBAHR R (2020) Meysam. Recent advances in the treatment of dye-containing wastewater from textile industries: overview and perspectives. Process Safety And Environmental Protection, [S.L.], v. 143, p. 138–163, nov. Elsevier BV. http://dx.doi.org/10.1016/j.psep.2020.05.034 SAMSAMI S, MOHAMADIZANIANI M, SARRAFZADEH M-H, Eldon RENE, FIROOZBAHR R (2020) Meysam. Recent advances in the treatment of dye-containing wastewater from textile industries: overview and perspectives. Process Safety And Environmental Protection, [S.L.], v. 143, p. 138–163, nov. Elsevier BV. http://dx.doi.org/10.1016/j.psep.2020.05.034 SANDIN G, PETERS, Greg M (2018) Environmental impact of textile reuse and recycling – A review. Journal Of Cleaner Production, [S.L.], v. 184, p. 353–365, maio. Elsevier BV. http://dx.doi.org/10.1016/j.jclepro.2018.02.266 KOVAčEVIć SCHWARZI (2017) Stana. Textile Application: from need to imagination. Textiles For Advanced Applications, [S.L.], v. 1, n. 5, p. 312–325, 20 set. InTech. http://dx.doi.org/10.5772/intechopen.68376 Alok SIKKAMPSARKAR, GARG Samridhi. Artificial intelligence (AI) in textile industry operational modernization. Research Journal Of Textile And Apparel, [S.L.], p. 258–268, 12 abr. 2022. Emerald. http://dx.doi.org/10.1108/rjta-04-2021-0046 SMIRNOV, Evgeny A, TIMOSHENKO, Denis M, ANDRIANOV, Serge N (2014) Comparison of Regularization Methods for ImageNet Classification with Deep Convolutional Neural Networks. Aasri Procedia, [S.L.], v. 6, p. 89–94, Elsevier BV. http://dx.doi.org/10.1016/j.aasri.2014.05.013 Chih-Ching SUK-Hua;LINJ-Horng;LIN (2007) Influence of reprocessing on the mechanical properties and structure of polyamide 6. Journal Of Materials Processing Technology, [S.L.], v. 192–193, p. 532–538, out. Elsevier BV. http://dx.doi.org/10.1016/j.jmatprotec.2007.04.056 UDDIN F (2021) Environmental hazard in textile dyeing wastewater from local textile industry. Cellulose, [S.L.], v. 28, n. 17, p. 10715–10739, 2 out. Springer Science and Business Media LLC. http://dx.doi.org/10.1007/s10570-021-04228-4 Tiwari VIBHAV, Garg SAMAR, Bhat SAMARTH, Kumar YSANCHIT, Prakash Bhat MANIKANDANJPMATHIAZHAGAN, Tamilselvan M, Periyasamy S (2023) Design of a Real-Time Warp and Weft Measurement System for Fabric Analysis in Textile Testing Using Machine Learning, International Conference on Network, Multimedia and Information Technology (NMITCON), p.1–7, 2023 Tyrone VIGO (2002) L. Textile processing and properties: preparation, dyeing, finishing and performance, 3 edn. Elsevier Science B.V., Amsterdam, p 486 ZHANG J, HE, Lin, CHENG, Longdi (1145) Is China’s Textile Industry Still a Labor-Intensive Industry? Fibers And Textiles In Eastern Europe, [S.L.], v. 29, n. p. 13–16, 28 fev. 2021. Walter de Gruyter GmbH. http://dx.doi.org/10.5604/01.3001.0014.5038 Table Table 1 is available in the Supplementary Files section Additional Declarations The authors declare no competing interests. Supplementary Files Table1.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3996611","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":275384284,"identity":"7f054021-d0d9-430c-8fe2-d73343b12522","order_by":0,"name":"Francis Dalponte Voigt","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABE0lEQVRIiWNgGAWjYBAC9gYQaQDEB5jBTAZ+EJFQgFsLzwG4FsYGMEsSRCUYENLCgKTFAG4ILi3Sh589+FFQx8B3/GDbg487/sgZn1+d+OGBAYM8v9gB7Fr40swNewwOM0ieSWw3nHnGwNjsxtvNEkCHGc6cnYBViz0Pg5k0yDEGBxLbpHnbDBK33Ti7AaQlweA2di08POzfgFrqGAzOPwRrqd884+zmH/i18IBsYWYwuAGxJcGAv3cbAVt4yiSBfuGRvPEQ6Jc2Y8MZN3i3WSQYSOD0C9Bh2yR+/KmT4zuffOzBxzY5ef7+s5tv/qiwkeeXxq4FrhWI2SBMCbBKCbzKYQCqhf8AUapHwSgYBaNg5AAA/I1cIn5SM/kAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0002-7422-8636","institution":"UFSC - Universidade Federal de Santa Catarina","correspondingAuthor":true,"prefix":"","firstName":"Francis","middleName":"Dalponte","lastName":"Voigt","suffix":""},{"id":275384285,"identity":"7e71367a-c1f9-4618-a7a6-d2511de4ec1e","order_by":1,"name":"Ricardo Antonio Francisco Machado","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA70lEQVRIiWNgGAWjYDACZhDBxgAl2RjkGJgPMDAwNpCgxZiBLYGAFphimJbEBkJazNuZH36uKLNh0G0//OzBjzKb9A3H2B9+YNxxD6cWmcNsxpJnzqUxmJ1JMzfsOZeWu+EYj7EE45linFokmHkYJBvbDjOYHchhk+BtO5y74X4PGwNjWwI+Lcw/wVrOv2GT/Nt2ON3gGPszQlrYILbcyGGTBtqSYHCMwYyAFjYzy4ZzaTxmN56ZScucSzOcCfJL4hk8WvgPP77ZUGYjZ3Y++ZnkmzIbeT5QiH3cgVsLDPCgcglrGAWjYBSMglGADwAANBBNf0DsQWAAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0003-1959-1456","institution":"UFSC - Universidade Federal de Santa Catarina","correspondingAuthor":true,"prefix":"","firstName":"Ricardo","middleName":"Antonio Francisco","lastName":"Machado","suffix":""}],"badges":[],"createdAt":"2024-02-28 12:17:20","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-3996611/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3996611/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":51796432,"identity":"7d941bc3-bb22-45d6-a973-a7f01dab2b30","added_by":"auto","created_at":"2024-02-29 07:50:15","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":344901,"visible":true,"origin":"","legend":"\u003cp\u003eProduction flowchart of the textile chain.\u003c/p\u003e\n\u003cp\u003eSource: Adapted from Pimentel; Lima (2011).\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/b323fc019ef7032917297cc4.png"},{"id":51796431,"identity":"e2c9d90d-9d40-409d-bdd1-766ed591a956","added_by":"auto","created_at":"2024-02-29 07:50:15","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":22978,"visible":true,"origin":"","legend":"\u003cp\u003eStages linked to textile processing.\u003c/p\u003e\n\u003cp\u003eSource: Adapted from Madhu; Chakraborty, (2017).\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/c3c928909e86c666774d47ae.png"},{"id":51796433,"identity":"85521317-9927-4a08-878d-f7171c622018","added_by":"auto","created_at":"2024-02-29 07:50:15","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":44282,"visible":true,"origin":"","legend":"\u003cp\u003eGeneral model of the artificial neuron.\u003c/p\u003e\n\u003cp\u003eSource: Adapted from Finochio (2014).\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/7871dd2f01c77d4663f3fe07.png"},{"id":51796419,"identity":"aa211f82-3c1f-46ac-9fe1-362a4addddbe","added_by":"auto","created_at":"2024-02-29 07:50:14","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":17462,"visible":true,"origin":"","legend":"\u003cp\u003eHopfield neural network.\u003c/p\u003e\n\u003cp\u003eSource: Adapted from Finochio (2014).\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/3523b23417b900b0bdccc2c4.png"},{"id":51796427,"identity":"4018df78-4efb-4c52-bce2-5f1de05ba455","added_by":"auto","created_at":"2024-02-29 07:50:15","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":141654,"visible":true,"origin":"","legend":"\u003cp\u003eStructure of a black box neural network.\u003c/p\u003e\n\u003cp\u003eSource: Adaptado de Finochio, (2014).\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/17a3e357da7467f0716ea583.png"},{"id":51796928,"identity":"ffe8793c-7f1d-45b4-9735-f9889824152f","added_by":"auto","created_at":"2024-02-29 07:58:15","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":11373,"visible":true,"origin":"","legend":"\u003cp\u003eStructure of an artificial neuron, where p = process values or input data; w = is called weight and represents the relevance of a given input variable in the behavior of the process; b = bias, which are the adjustments that are necessary to correct some deviation; F = is the activation function of the neuron; and a = is the value predicted by the neuron.\u003c/p\u003e\n\u003cp\u003eSource: Adapted from Finochio (2014).\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/91fdd86840b0c839401de559.png"},{"id":51796421,"identity":"3101c2b9-e47c-4683-903b-d3778b25bfaf","added_by":"auto","created_at":"2024-02-29 07:50:14","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":22717,"visible":true,"origin":"","legend":"\u003cp\u003eRepresentation of a neural network that has several inputs in an artificial\u003c/p\u003e\n\u003cp\u003eSource: Adaptado de Finochio, (2014).\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/922f14cb6b9a12d569353f97.png"},{"id":51796422,"identity":"dbbed9bb-bcbc-4779-bf75-3e04fc8930f8","added_by":"auto","created_at":"2024-02-29 07:50:14","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":40617,"visible":true,"origin":"","legend":"\u003cp\u003eRepresentation of a neural network with several layers of interconnected artificial neurons. The stimuli are the inputs of neurons.\u003c/p\u003e\n\u003cp\u003eSource: Adaptado de Finochio, (2014).\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/28b26a407f7bb97d19621d73.png"},{"id":51796424,"identity":"a835be3b-b2f1-41c6-b114-92684aef4f72","added_by":"auto","created_at":"2024-02-29 07:50:14","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":32176,"visible":true,"origin":"","legend":"\u003cp\u003eNeural Network Training Phase\u003c/p\u003e\n\u003cp\u003eSource: Adapted from Finochio (2014).\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/7941acd468dd676cc11e11c3.png"},{"id":51796426,"identity":"90647f09-1ac1-4421-8bdf-139833baddae","added_by":"auto","created_at":"2024-02-29 07:50:15","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":20555,"visible":true,"origin":"","legend":"\u003cp\u003ePercentage of reprocessing in the manufacture of different types of fibers.\u003c/p\u003e\n\u003cp\u003eSource: The author, (2023).\u003c/p\u003e","description":"","filename":"10.png","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/cf21aec3a584bf128e8f6396.png"},{"id":51796428,"identity":"6c13b98b-5391-48b4-b551-875afdd7d691","added_by":"auto","created_at":"2024-02-29 07:50:15","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":46532,"visible":true,"origin":"","legend":"\u003cp\u003eSchematic of the intended network.\u003c/p\u003e\n\u003cp\u003eSource: The author, (2023).\u003c/p\u003e","description":"","filename":"11.png","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/64829279a66ed3f93919eb02.png"},{"id":51796429,"identity":"0cd9a09f-aa7a-428c-a3e7-ed89fea73911","added_by":"auto","created_at":"2024-02-29 07:50:15","extension":"png","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":182825,"visible":true,"origin":"","legend":"\u003cp\u003eEvolution of network performance indexes. (a) Supervision of learning. (b) Histogram of error in relation to the predicted value and the wrong values, (c) evolution of the correlation coefficient, and (d) better performance obtained in relation to the actual data.\u003c/p\u003e","description":"","filename":"12.png","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/33d6458b396581403313b28e.png"},{"id":51796430,"identity":"9b2364f5-c9f2-4929-a4b6-8c132673b6d7","added_by":"auto","created_at":"2024-02-29 07:50:15","extension":"png","order_by":13,"title":"Figure 13","display":"","copyAsset":false,"role":"figure","size":55214,"visible":true,"origin":"","legend":"\u003cp\u003eVariables for decision-making about the network architecture and its performance during the training phase.\u003c/p\u003e","description":"","filename":"13.png","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/79808b4328c8a8b87f6a1d07.png"},{"id":51797583,"identity":"7d9ee8c9-1b1a-4aef-b33c-4a16a2a97498","added_by":"auto","created_at":"2024-02-29 08:06:16","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1375061,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/94e00926-5b2d-49ec-be16-ec92305a3496.pdf"},{"id":51796434,"identity":"1b080b11-94fe-47bd-8954-4b126f1eba21","added_by":"auto","created_at":"2024-02-29 07:50:16","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":46481,"visible":true,"origin":"","legend":"","description":"","filename":"Table1.docx","url":"https://assets-eu.researchsquare.com/files/rs-3996611/v1/290410849dc36e7625aa226d.docx"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eDevelopment of an artificial neural network to maximize the reproducibility of dyeing polyamide fabrics\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eThe Brazilian textile industry plays a key role in the country's economy, as Brazil stands out as the fifth largest textile hub in the world, being especially strong in the clothing segment, occupying the fourth position globally. In addition, the country is self-sufficient in cotton production, which contributes to the supply of raw material for this industry, which is one of the most complete chains in the West, ranging from the production of fiber to retail marketing. This vertical integration allows for greater efficiency and quality control throughout the production process. In addition, with the discovery of presalt, Brazil has the opportunity to become a major global exporter of synthetic fibers (ABIT, 2021).\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eThe textile and apparel sector played an important role in the world's manufacturing history, even before the industrial period, when products were manufactured by hand. However, with the course of industrial revolutions, the sector underwent transformations in the configuration of the business model, replacing manual work operations with factory operations (DUARTE, 2017). As a result, the company developed a productive infrastructure that was transformed into industrial parks to meet large-scale demands in domestic and foreign markets. This infrastructure currently constitutes a network of independent productive infrastructures, such as spinning, weaving, knitting, processing and clothing, constituting the textile chain (FUJITA; JORENTE, 2015).\u003c/p\u003e\u003cp\u003eWithin the textile infrastructure, the processing sector plays a key role in adding value and meeting a wide range of needs in this segment. However, it is important to note that this process faces significant challenges, which makes it a critical point throughout the production chain. This behavior is linked to several factors, including the long production cycle, the substantial need for water for the process, the release of toxic substances into the environment, and the use of chemical products such as dyes, heavy metals, acids, caustic soda, and sodium carbonate (CHEN., 2017; KU et al., 2020). These factors pose not only environmental challenges but also concerns about the safety and sustainability of the beneficiation process.\u003c/p\u003e\u003cp\u003eIn this context, on the one hand, the global demand for textile products has been steadily increasing, driven by the increase in population and economic development; on the other hand, it is evident that companies are increasingly aware of and concerned about the environmental impact resulting from their manufacturing operations (KUMAR et al., et al. 2020, ZHANG et al., 2021; SANDIN; PETERS, 2018). In response to this concern, several areas have adopted Cleaner Production (P\u0026thinsp;+\u0026thinsp;L) approaches, as well as Industry 4.0 concepts, and applied their technologies in production processes to provide innovation for their companies and make them more competitive (KUMAR et al., 2020).\u003c/p\u003e\u003cp\u003eIn addition to issues related to environmental impacts, textile companies face significant challenges in their production processes due to the need to reprocess the meshes. Reprocesses occur when the product varies in relation to the standards of color, touch, equalization, elongation and elasticity, touch and volume, compromising the efficiency of production and affecting customer service due to delays in the shipment of products and for presenting a different standard than desired at the time of purchase. Such a condition damages the image of a company and reduces its performance in the market, in addition to representing an increase in costs and environmental impact, by consuming more inputs and resources to obtain products.\u003c/p\u003e\u003cp\u003eIn this sense, the textile industry is generating increasing demands in the global market, and modern technologies have benefited this segment through the automation of long and complex processes, resulting in improvements in the speed, quality and cost of textile manufacturing (SCHWARZ; KOVAČEVIĆ, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). According to Falani et al. (2020), technological advances are one of the strategic factors for generating changes that contribute to the improvement of techniques, tools, and the use of new inputs. Fujita and Jorente (2015) point out that it is necessary to invest in technological innovation and in the generation of new knowledge through scientific development carried out to encourage research shared with the textile industry.\u003c/p\u003e\u003cp\u003eIn this scenario, artificial intelligence (AI), in essence, makes it possible to contribute to this need of the textile industry. AI ensures that systems make decisions independently, accurately, and supported by digital data. In an optimistic view, this multiplies the rational capacity of the human being to solve practical problems, simulate situations, think of answers or, more broadly, enhance the capacity to be intelligent.\u003c/p\u003e\u003cp\u003eIn view of the above, the present research aims to autonomously implement a neural network for prediction and correction, identifying reprocessing in the dyeing of a textile industry through preestablished patterns. In addition to the economic and social aspects, the success of this research increases its relevance by contributing to the saving of resources (water, inputs, energy), with a high environmental impact, reinforcing the company's existing commitments and completing the tripod of sustainability: social, economic and environmental.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e"},{"header":"THEORETICAL BACKGROUND","content":"\u003cp\u003eIn this chapter, theoretical assumptions are presented based on the national and international literature on the processing process of textile industries with an emphasis on polyamide fibers, reprocessing, and artificial intelligence tools.\u003c/p\u003e\n\u003ch2\u003e2.1 TEXTILE PRODUCTION CHAIN IN BRAZIL\u003c/h2\u003e\n\u003ch3\u003e2.1.1 Importance of the textile production chain in the Brazilian economy\u003c/h3\u003e\n\u003cp\u003eIn general, the textile production chain is a vital cog in the Brazilian economy, contributing significantly to employment, regional development, exports, synergy between different sectors and technological advancement. Its key role in building a robust and diversified economy makes it an inalienable pillar in Brazil\u0026apos;s economic landscape.\u003c/p\u003e\n\u003cp\u003eIn this context, Brazil is the 5th largest textile industry in the world and the 4th largest in the clothing segment, whose average production, in tons, was approximately 190 million in 2021 (IEMI, 2022). The textile industry promotes 1.36 million direct jobs, and in 2020, it represented 19.8% of the total number of workers allocated to industrial production and 5% of the total value of production in the Brazilian manufacturing industry (IEMI, 2022).\u003c/p\u003e\n\u003cp\u003eAccording to the Institute of Industrial Studies and Marketing (2022), the Brazilian textile chain covers more than 24.6 thousand factories installed in the country, with approximately 3,030 textile companies and the remaining clothing companies. Along with these data, Brazil is the only country in the West that has a fully verticalized textile chain (ABIT, 2022).\u003c/p\u003e\n\u003cp\u003eIn this sense, due to the high relevance of the textile industry in the national economy, as it demands the intensive use of labor, this sector is a great generator of income and jobs for the country. In 2022, the textile industry produced 1.34 million formal jobs and more than 8 million indirect jobs, 60% of which involved women (IEMI, 2022). This fact is linked to the large per capita production of textiles in Brazil, which reached 9.9 kg per inhabitant, with a consumption of approximately 13.3 kg/inhabitant, and this difference is met by the international market (ABIT, 2022).\u003c/p\u003e\n\u003ch3\u003e2.1.2 Stages of the production processes of the textile chain\u003c/h3\u003e\n\u003cp\u003eThe beginning of the production process of the textile chain is first constituted by the production process of the fibers. Subsequently, several large-scale spinning companies produce wires. These companies have a large volume of capital and technology, are automated, and have a low labor ratio. The next stage of the chain includes the weaving, knitting and processing industries, which have characteristics similar to those of spinning (FALANI et al., 2020). In the last link of the chain are the clothing industries, which have a characteristic low use of capital and technologies but are very intense in the use of labor. For these reasons, they are mostly formed by micro- and small enterprises, most of which are informal (FALANI et al., 2020). Figure 1 shows the representation in the form of a flowchart, with the main stages of the textile production process. For ease of understanding, the flowchart contains only the macro steps, without any detail.\u003c/p\u003e\n\u003cp\u003eWithin the flowchart of the textile chain, the beneficiation process usually occurs after the production stage of the fibers and before the stage of making the garments or final textile products. Textile beneficiation involves a series of processes that prepare textile materials for the manufacture of finished products. These processes can include several steps, such as (i) dyeing, where dye is applied to color the fabric or yarn according to the design specifications; (ii) finishing, which involves processes such as bleaching, stamping, calendering and others that can affect the texture, gloss and feel of the material; (iii) washing, where materials are washed to remove impurities and residual chemicals from previous processes; (iv) mechanical finishing, which may include processes such as cutting, sewing and finishing of garments or textile products; and (v) quality inspection to finalize the process.\u003c/p\u003e\n\u003ch2\u003e2.2 PROCESSING OF TEXTILE INDUSTRIES\u003c/h2\u003e\n\u003cp\u003eTextile processing, in general, can be defined as the process of aggregating several steps to improve the characteristics of fabrics, fibers and yarns, giving fabrics unique particularities. It is composed of different steps and processes (SAMSAMI et al., 2020). Textile processing consists of the stages of primary processing (preparation), secondary processing (dyeing) and tertiary processing (finishing). Figure 2 shows a flowchart of the main steps and processes involved in textile processing.\u003c/p\u003e\n\u003ch3\u003e2.2.1 Pretreatment\u003c/h3\u003e\n\u003cp\u003ePrimary processing, or pretreatment, is the initial stage of the processing process. All textile substrates need this treatment before moving on to downstream processes, such as dyeing, printing or finishing. Madhu and Chakraborty (2017) noted that this processing is the first treatment applied to a textile material (yarn or fabric) and aims to obtain a final product with better acceptance in the market. The operations to be carried out follow a sequence and must be preserved to ensure good results.\u003c/p\u003e\n\u003cp\u003eThis process is also known as purging or boiling and is responsible for the removal of oils from textile fibers; for this purpose, oil emulsifier products, also called surfactants or detergents, which can be products with nonionic or anionic characteristics, are used. This process is usually carried out at temperatures of 50 to 60\u0026deg;C (SALEM et al., 2005).\u003c/p\u003e\n\u003cp\u003eIn general, the effective removal of impurities occurs with the use of 3-6% sodium hydroxide, calcium hydroxide or sodium carbonate. In addition, the proper choice of textile auxiliaries in an alkaline bath is essential for good purging processing. These include sequestering or chelating agents, such as ethylenediaminetetraacetic acid (EDTA), to solubilize insoluble inorganic substances present in hard water, and surfactants, such as sodium lauryl sulfate, which serves as a detergent, dispersing agent, and emulsifying agent to remove unsaponifiable waxes (BARANI, MONTAZER, 2008).\u003c/p\u003e\n\u003cp\u003ePurging is also applied to synthetic fibers that take on a yellow tint due to overheating or the buildup of impurities during manufacturing. This process is gentler than that of natural fibers (BROADBENT, 2011). In synthetic fibers, washing is carried out with soap or detergents containing lower amounts of alkaline solution (0.1-0.2% sodium carbonate) (VIGO, 2002).\u003c/p\u003e\n\u003cp\u003eMoreover, cotton fibers usually require prebleaching unless they are dyed in the dark. The purpose of bleaching is to remove impurities that mask the natural whiteness of the fibers. Oxidizing agents, such as hydrogen peroxide, sodium hypochlorite, and sodium chlorite, are used for this purpose. On an industrial scale, hydrogen peroxide is the most common oxidizing agent. In addition, the process can be conducted by the depletion method or on a continuous basis (ADANUR, 2017). After the preparation stage, most of the textile substrates are subjected to dyeing, where they receive the desired color.\u003c/p\u003e\n\u003ch3\u003e\u003cbr\u003e\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e2.2.2\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eMercerization\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn this process, a physicochemical treatment involving the impregnation of the textile material under tension with alkaline solutions under strictly controlled temperature and concentration conditions occurs. The purpose of this processing is to increase the brightness of the product, as well as the absorption of water and dyes, as well as the tensile strength and dimensional stability (ADANUR, 2017).\u003c/p\u003e\n\u003cp\u003eMercerization is usually applied to yarns and flat fabrics, as well as cellulosic fibers, especially cotton, and can be performed with sodium hydroxide at temperatures ranging from 10 to 18 \u0026deg;C, followed by rinsing and neutralization (BROADBENT, 2011).\u003c/p\u003e\n\u003ch3\u003e2.2.3 Bleaching\u003c/h3\u003e\n\u003cp\u003eBleaching is the chemical process used in the bleaching of textile materials, the ultimate goal of which is bleaching. In addition, this process can also be used in articles that require optical brightening to enhance the degree of whiteness (BROADBENT, 2011).\u003c/p\u003e\n\u003cp\u003eThis process is employed in different textile products, such as yarns, flat fabrics and knitwear; oxidative processes, such as sodium hypochlorite, ozone and hydrogen peroxide; and reducing processes, such as sodium hydrosulfite, sodium sulfoxylated formaldehyde, sodium bisulfite and thiurea dioxide.\u003c/p\u003e\n\u003ch3\u003e1.1.1 Thermosetting\u003c/h3\u003e\n\u003cp\u003eThermosetting can be considered a pretreatment of the fabric, as it can be performed before dyeing to provide dimensional stability. This process is carried out only on synthetic fibers such as polyester and its blends and on articles with elastane (RUSCHIONI; ALFIERI, 2010).\u003c/p\u003e\n\u003ch3\u003e1.1.2 Dyeing\u003c/h3\u003e\n\u003cp\u003eDyeing is a process known for the homogeneous coloring of textile substrates with the use of dyes. This process is also known as secondary beneficiation. In general, this process is divided into three stages, mediated by the processes of migration, absorption and fixation of the dye. In the first step, the dye migrates from the medium in which it is diluted to the surface of the fiber. Subsequently, the process of adsorption occurs in the surface layers of the textile material, and then the dye diffuses inside the fiber and is fixed by different types of bonds (ionic, van der Waals, covalent) depending on the type of material used (ADANUR, 2017).\u003c/p\u003e\n\u003cp\u003eTemperature is a primary factor in these steps due to the influence of chemicals and the mechanical action caused by the agitation of the dyeing bath of the textile substrate during processing (SALEM et al., 2005). In addition to the temperature, the dyeing speed must be carefully controlled in the form of a curve, considering the substrate, products and machine. This curve should express the dyeing time as a function of the total percentage of dye that will assemble when it reaches equilibrium as well as the time required to reach half depletion. Many critical factors are important for good equalization and reproducibility (SALEM et al., 2005).\u003c/p\u003e\n\u003cp\u003eCompared to wool, the diffusion of acid dyes in polyamide is slower, and the limited number of sites loaded in polyamide can also cause problems in dyeing deep shade blends, where individual dyes compete for available sites. Under these conditions, the fastest diffusing dyes can block the entry of a second component, and a hue is not achieved (ADANUR, 2017).\u003c/p\u003e\n\u003cp\u003eThe greater or lesser saturation of the fiber depends on the percentage of terminal amine groups (AEGs), and these in polyamide fibers are limited in number. Saturation also depends on the number of sulfonic groups in the dye that will react with the terminal amino groups. The greater the number of sulfonic groups in a dye molecule is, the lower the saturation: a dye molecule will occupy more than one terminal amine group. Thus, a trisulfonic dye occupies three terminal amine groups. Therefore, when dyes are combined with mono- and trisulfonic agents in the same recipe, the monosulfonic dye tends to mount first on the fiber, occupy the amine groups (which are limited) and block the assembly of the trisulfonic dye, which remains in the bath (GONDIM, 2016; SALEM, et al., 2005).\u003c/p\u003e\n\u003cp\u003eThe permanence of the dye in the fiber is affected by several factors, among which we can highlight the (i) vibration of the molecular structure of the fiber, at each moment, taking new configurations; (ii) constant bombardment of the dye by water molecules during dyeing, making it difficult to fix it in the fiber; and (iii) as the temperature of the system increases, the vibration of the fiber molecules and the bombardment of the water molecules increase (SALEM et al., 2005).\u003c/p\u003e\n\u003cp\u003ePolyamide reprocessing in the textile industry mainly occurs through dyeing processes, where the goal is to change the color or physical properties of polyamide fabrics. This reprocessing may be necessary for a variety of reasons, such as color correction, improved dyeing quality, shade adjustments, or even to correct imperfections that occurred during the initial process (SU et al., 2007).\u003c/p\u003e\n\u003ch3\u003e1.1.3 Finishing\u003c/h3\u003e\n\u003cp\u003eFinishing is intended to provide desirable properties to the fabric, including soft touch, wrinkle resistance, and impermeability, among other functional characteristics. This stage involves procedures such as calendering, which uses heated cylinders to smooth and shine the fabric, as well as wire to remove lint and surface impurities.\u003c/p\u003e\n\u003cp\u003eChemical agents, such as water repellency, stain resistance, or antimicrobial agents, are commonly applied at the finish of a material to impart specific properties to the fabric. These chemical treatments are applied in a controlled manner to ensure the desired functionality without compromising the integrity of the tissue (SALEM et al., 2005).\u003c/p\u003e\n\u003cp\u003eThe efficient and accurate integration of these steps is crucial to achieving final textile products that meet the requirements for quality, aesthetics, and functionality. The combination of printing and finishing techniques in textile processing plays a decisive role in the variety and quality of textile products available on the market, reflecting the importance of these processes in the contemporary textile industry.\u003c/p\u003e\n\u003ch2\u003e1.2 ARTIFICIAL INTELLIGENCE\u003c/h2\u003e\n\u003cp\u003eAI represents the pinnacle of technological innovation, being a multidisciplinary field that seeks to replicate the human capacity for learning, reasoning, and decision-making through algorithms and computational systems. Based on concepts such as machine learning, neural networks, and natural language processing, AI has gained a central role in several spheres of modern life (SU et al., 1993). Its ability to analyze large volumes of data, identify complex patterns, and make autonomous decisions is redefining industries, driving automation, and transforming the way we interact with technology, from virtual assistants to advanced medical diagnostic systems. The rapid evolution and application of AI are shaping not only the age of technology but also redefining the boundaries of human knowledge and its practical applications (MUKHERJEE; BHA, 2023).\u003c/p\u003e\n\u003cp\u003eTheoretical assumptions about AI are critical to understanding the conceptual basis behind the development and application of this scenario. These assumptions form the theoretical foundation that guides the creation of intelligent systems and algorithms (Sch\u0026auml;dler and Wysotzki, 1999). They encompass areas such as machine learning, which focuses on the ability of systems to learn and improve from experience; computational logic, which aims to understand how systems can represent knowledge and reason logically; and neural networks, which are inspired by the functioning of the human brain to create highly efficient computational models (HIMMELBLAU, 2008). Understanding these assumptions is crucial for exploring not only the potential but also the limits and challenges of AI in diverse application fields, including textile beneficiation (SIKKA et al., 2022).\u003c/p\u003e\n\u003ch3\u003e1.2.1 Artificial neural networks (ANN)\u003c/h3\u003e\n\u003cp\u003eANNs are computational elements inspired by biological neurons, which, when networked, can reproduce some characteristics of intelligent processes (RAMESH et al., 2004). ANN learning has emerged as a dominant framework in the present day, generating advances in a wide range of applications, including computer vision, natural language processing, and strategic games (SMIRNOV et al., 2014). Some key ideas in this field can be traced back to the human brain, and there is currently a continuous exchange of ideas from neuroscience to the field of artificial intelligence (HASSABIS et al., 2017).\u003c/p\u003e\n\u003cp\u003eAt the same time, learning about AI offers powerful new tools for systems neuroscience. In fact, advances in computer vision, especially ANNs, have revolutionized image and video data processing. Uncontrolled behaviors over time, such as the micromovements of animals in laboratory experiments, can now be efficiently tracked and quantified with the help of this technology (MATHIS et al., 2018).\u003c/p\u003e\n\u003cp\u003eInspired by the ability of humans and other animals to perform functions such as processing sensory information and interacting with poorly defined environments, engineers are concerned with developing artificial systems capable of performing similar tasks. Skills such as incomplete or inaccurate information processing ability and generalization are desirable properties in such systems (LIU et al., 2020).\u003c/p\u003e\n\u003cp\u003eIn this sense, ANNs are computational techniques that have the ability to solve problems through simple circuits that simulate the functioning and behavior of the human brain. They present a model inspired by the neural structure of intelligent organisms, which acquire knowledge through experience, that is, by learning, making mistakes and making discoveries. An artificial neural network can have hundreds or even thousands of processing units, while a mammal\u0026apos;s brain can contain many billions of neurons (LIU et al., 2020)\u003c/p\u003e\n\u003cp\u003eHowever, it is important to point out that there are differences between human nerve cells and artificial neuron models. However, the principle of information transfer is the same. Analogous to the human brain, ANNs can interact with and adapt to the external environment. These characteristics give ANNs multidisciplinary importance, which is why this tool has been gaining prominence in different areas of knowledge, such as engineering, mathematics, physics, and computer science (LIU et al., 2020).\u003c/p\u003e\n\u003cp\u003eThe mathematical model of the neuron, which encompasses the main characteristics of a biological neural network, parallelism and high connectivity, was proposed in 1942 by McCulloch and Pitts, where the researchers designed the structure known as the first neural network known worldwide as the MCP model (McCulloch-Pitts). The MCP model is a simplification of the biological neuron model, which considers the neuron as a binary information processing unit with multiple binary inputs and a single binary output, showing that these units are capable of performing different logical operations.\u003c/p\u003e\n\u003cp\u003eFigure 3 shows the general model of the artificial neuron, where x1, x2, and xn represent the input signals; W1, W2, and wn are the weights or synaptic connections; BIAS represents the neuron\u0026apos;s activation threshold; u is the output of the linear combiner; (f) is the activation function (limits neuron output); and y is the neuron\u0026apos;s output signal.\u003c/p\u003e\n\u003cp\u003eThe operation of a cell in a neural network can generally be described as follows: (i) signals are presented to the input; (ii) each signal is multiplied by a weight, which indicates its influence on the cell\u0026apos;s output; (iii) the weighted sum of the signals is executed, which produces a level of activity; and (iv) when this level exceeds a threshold, the unit produces an output (FINOCHIO, 2014).\u003c/p\u003e\n\u003cp\u003eOver time, ANNs have undergone a process of evolution. First, in 1948, N. Wiener coined the word cybernetics to describe, in a unified way, control and communication in living organisms and machines. Subsequently, in 1949, D. O. HEBB proposed a hypothesis about how the strength of synapses in the brain changes in response to experience. In particular, he suggested that connections between cells that are activated at the same time tend to strengthen, while other connections tend to weaken. This hypothesis had a decisive influence on the evolution of the theory of learning in artificial neural networks (HOPFIELD, 1984).\u003c/p\u003e\n\u003cp\u003eLater, in 1957, Rosenblatt introduced a new approach to the pattern recognition problem with the development of the perceptron (Figure 4). This same researcher also proposed an algorithm for adjusting the perceptron weights and proved their convergence when the patterns are linearly separable. At approximately the same time, B. WIDROW (1965) and his collaborators developed the adaptive linear element (Adaline) (STEINBUCH and WIDROW, 1965).\u003c/p\u003e\n\u003cp\u003eThe interesting conclusion adopted by Hopfield was that such equilibrium states can be used as memory devices. Unlike that used by conventional computers, in which access to stored information occurs through an address, access to the contents of the memory of a Hopfield network is given by allowing the network to evolve over time to one of its equilibrium states. Such memory models are called content-addressable memories (HOPFIELD, 1984).\u003c/p\u003e\n\u003cp\u003eHowever, effectively placing the RNA area as one of the priorities in obtaining resources was the priority of developing a method for adjusting the parameters of nonrecurrent multilayer networks. Therefore, one of the most commonly used architectures is the multilayer perceptron network, which uses the back propagation algorithm, proposed in 1986 by Rumelhart, McClelland and Williams (BEALE; JACKSON, 1990), as a learning rule.\u003c/p\u003e\n\u003ch4\u003e1.2.1.1 Characteristics of artificial neural networks\u003c/h4\u003e\n\u003cp\u003eThe most important characteristics of artificial neural networks are numerous: (i) the ability to learn the relationships between a set of input data, also called training examples, and thus improve their performance. This ability is due to the training or learning algorithm, which will be discussed below; (ii) the ability to generalize learning to new examples. Networks can provide responses similar to those for which they were trained, for examples not presented in the training.\u003c/p\u003e\n\u003cp\u003eThe ability to extract the essence of a dataset and learn from incomplete information; (iii) robustness and fault tolerance: the elimination of some neurons does not substantially affect their overall performance; (iv) flexibility: can be adjusted to new environments through a learning process, being able to learn new actions based on the information contained in the training data; (v) processing of uncertain information: even if the information provided is incomplete and affected by noise, it is still possible to obtain correct reasoning; and (vi) parallelism: an immense number of neurons are active at the same time. There is no restriction of a processor that necessarily works one instruction after another (FINOCHIO, 2014).\u003c/p\u003e\n\u003cp\u003eThe above characteristics make neural networks especially attractive for applications in nonlinear systems that have a large volume of parameters and can undergo temporal variations, as is the case for textile processing. ANNs can also be classified according to their architecture, for example:\u003c/p\u003e\n\u003cp\u003e1. Perceptron\u003c/p\u003e\n\u003cp\u003e2. Feed Forward Neural Network\u003c/p\u003e\n\u003cp\u003e3. Multilayer Perceptron\u003c/p\u003e\n\u003cp\u003e4. Convolutional Neural Network\u003c/p\u003e\n\u003cp\u003e5. Radial Basis Functional Neural Network\u003c/p\u003e\n\u003cp\u003e6. Recurrent Neural Network\u003c/p\u003e\n\u003cp\u003e7. LSTM \u0026ndash; Long short-term memory\u003c/p\u003e\n\u003cp\u003e8. Sequence-to-Sequence Models\u003c/p\u003e\n\u003cp\u003e9. Modular Neural Network\u003c/p\u003e\n\u003cp\u003eThe classifications were cited in English on purpose, as this is the way they are usually referenced. In addition, neural networks can still be combined with each other, fed back and still have prior knowledge.\u003c/p\u003e\n\u003cp\u003eNeural networks that have structures such as 3 layers, the first layer, the input layer, the second layer (intermediate layer), the processing layer, and the third layer, the output layer, are usually called black box neural networks. Figure 5 shows this type of structure.\u003c/p\u003e\n\u003cp\u003eA neural network has elements called neurons, which can be completely or partially interconnected. Figure 6 shows the structure of an artificial neuron.\u003c/p\u003e\n\u003cp\u003eFigure 6 - Structure of an artificial neuron, where p = process values or input data; w = is called weight and represents the relevance of a given input variable in the behavior of the process; b = bias, which are the adjustments that are necessary to correct some deviation; F = is the activation function of the neuron; and a = is the value predicted by the neuron.\u003c/p\u003e\n\u003cp\u003eEquation 1 represents the activation function of the neuron, F:\u003c/p\u003e\n\u003cp\u003ea = F(w.p + b) (1)\u003c/p\u003e\n\u003cp\u003ewhere:\u003c/p\u003e\n\u003cp\u003ea = output data or predicted values\u003c/p\u003e\n\u003cp\u003ep = input data\u003c/p\u003e\n\u003cp\u003ew = weights\u003c/p\u003e\n\u003cp\u003eb = bias\u003c/p\u003e\n\u003cp\u003eF = activation function\u003c/p\u003e\n\u003cp\u003eThe w and b parameters should be adjusted so that they represent the process data with relative accuracy. In the tuning phase, we say that the neural network is being trained, which is nothing more than the employment of an optimization algorithm to adjust the values of w and b in such a way as to predict the output data (a) with the least possible deviation from the input values. This is where machine learning terms come in. Neurons can also have multiple inputs, as shown in Figure 7.\u003c/p\u003e\n\u003cp\u003eIf we add several neurons to each layer, we consequently increase the \u0026quot;intelligence\u0026quot; of the neural network, as there are several activation functions and several parameters to be estimated. Figure 8 shows a schematic of an artificial neural network with several layers.\u003c/p\u003e\n\u003cp\u003eHowever, it is necessary to keep in mind that when referring to industrial processes, we are dealing with numerical information (temperature, pressure, molar fractions, etc.) and subjective information (good, bad, more or less, etc.). Then, it is necessary to create a scale so that the subjective information can be quantified numerically.\u003c/p\u003e\n\u003cp\u003eThe numerical, quantitative information must have the same scale; otherwise, an adjustment considered excellent for the set of temperatures, which presents only 0.5 K of deviation, is a tremendous error when we evaluate molar fractions. In addition, there is the problem of properly handling information before processing it.\u003c/p\u003e\n\u003cp\u003eFor the sake of objectivity, we will deal in this document only with black box artificial neural networks because we do not yet have enough information about the process to be able to insert prior knowledge into the structure we are going to use.\u003c/p\u003e\n\u003cp\u003eReturning to the issue of data processing, it is initially necessary to define the type of activation function to be used. The most common are described in Chart 1.\u003c/p\u003e\n\u003cp\u003eTable 1 - Types of activation functions\u003c/p\u003e\n\u003cp\u003eDue to its versatility and nonlinear characteristics, which enable the representation of extremely complex data, the tangential sigmoid activation function is preferentially used. Especially when you have sparse systems.\u003c/p\u003e\n\u003cp\u003eIn textile processing, we find both nonlinear and sparse systems. Therefore, the tangential sigmoid activation function is the first to be employed.\u003c/p\u003e\n\u003cp\u003eHowever, we still need to take care of the data before we start training the network. We have already defined the activation function. Looking at their behavior, the limits are -1 and +1. That is, we need to scale the data so that they are all between -1 and +1.\u003c/p\u003e\n\u003cp\u003ePrevious studies have shown that the output of the activation function should not be saturated. In other words, we should not work at the lower and upper limits.\u003c/p\u003e\n\u003cp\u003eHaykin (1999) recommends, in the case of the tangential sigmoid activation function, that scaling be performed between -0.8 and +0.8, according to Equations 2 and 3.\u003c/p\u003e\n\u003cp\u003eFor the input data:\u003c/p\u003e\n\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\"\u003e\u003cbr\u003e\u003c/p\u003e\n\u003cp\u003eWith the elements of the input and output datasets ranging from -0.8 to +0.8, we can then move on to the training phase.\u003c/p\u003e\n\u003cp\u003eIn terms of training, various optimization algorithms can be used, with some being faster and others being slower. The most common is the use of step-down algorithms. Nasra et al. (2016) explored different training algorithms and concluded that the Levenberg\u0026ndash;Marquartd method has a higher convergence speed, which is why it is used in this work. Other optimization algorithms can be found in Yin et al. (2003). Figure 9 shows the flowchart of the training phase (machine learning).\u003c/p\u003e\n\u003cp\u003eIn the next subitem, we will show some uses of artificial intelligence in textile processing.\u003c/p\u003e\n\u003ch3\u003e1.2.2 \u0026nbsp; Artificial intelligence and textile processing\u003c/h3\u003e\n\u003cp\u003eThe application of AI in the textile processing sector represents a significant evolution in the modern industry landscape. The integration of intelligent algorithms, machine learning, and automation has revolutionized processes from design conception to final production. This fusion of AI and textile beneficiation offers unprecedented opportunities to optimize efficiency, accuracy, and sustainability at every stage of the process, including printing, dyeing, finishing, and quality control. This innovative convergence not only reshapes the way fabrics are produced but also redefines standards for quality, customization, and speed of response to market demands (SIKKA et al., 2022; CHATTOPADHYAY; GUHA, 2004).\u003c/p\u003e\n\u003cp\u003eSpecifically, in the case of ANNs, notable advances have been identified in the field of textile processing. Currently, these nets are applied in a significant way in the quality control of fabrics, allowing the identification of defects, dyeing patterns and finishing failures in a much more accurate and efficient way than conventional methods. In addition, ANNs have been used in the optimization of process parameters such as temperature, dyeing time and chemical composition. In addition, RNA-based models play a crucial role in predicting the final properties of tissues, including strength, durability, and shrinkage behavior. This enables precise adjustments in manufacturing processes, contributing to improving the quality of textile products (SIKKA et al., 2022). Table 1 presents a general summary of the contributions of ANNs to the textile industry.\u003c/p\u003e\n\u003cp\u003eTable 1 - Synthesis of studies developed with RNA and the textile industry.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"32.78145695364238%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eAuthors\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"67.21854304635761%\" valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eObjective of the work\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"32.78145695364238%\" valign=\"top\"\u003e\n \u003cp\u003eBahlmann et al. (1999)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"67.21854304635761%\" valign=\"top\"\u003e\n \u003cp\u003eDevelop RNA to establish an automated control in textile seams.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"32.78145695364238%\" valign=\"top\"\u003e\n \u003cp\u003eHui et al. (2007)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"67.21854304635761%\" valign=\"top\"\u003e\n \u003cp\u003eDevelop RNA for textile seam performance.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"32.78145695364238%\" valign=\"top\"\u003e\n \u003cp\u003eTiwari et al. (2023)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"67.21854304635761%\" valign=\"top\"\u003e\n \u003cp\u003eApplication of RNA to evaluate the performance of produced tissues.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"32.78145695364238%\" valign=\"top\"\u003e\n \u003cp\u003eDoran et al. (2019)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"67.21854304635761%\" valign=\"top\"\u003e\n \u003cp\u003eDevelop RNA for prediction of cotton and spandex yarns.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"32.78145695364238%\" valign=\"top\"\u003e\n \u003cp\u003eJeyaraj et al. (2019)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"67.21854304635761%\" valign=\"top\"\u003e\n \u003cp\u003eUse of RNA for detection of defects in the tissue manufacturing process\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"32.78145695364238%\" valign=\"top\"\u003e\n \u003cp\u003eLi et al. (2021)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"67.21854304635761%\" valign=\"top\"\u003e\n \u003cp\u003eIdentification of waste generated by the textile industry using RNA\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eSource: The author, (2023).\u003c/p\u003e\n\u003cp\u003eDespite these advances, some challenges remain, such as the need for more robust and representative datasets, as well as the interpretability of models for industrial application. Furthermore, the potential of neural networks to simulate complex processes is still evolving (SIKKA et al., 2022).\u003c/p\u003e"},{"header":"MATERIALS AND METHODS","content":"\u003cp\u003eThis research is being developed in partnership with a textile company located in Indaial/SC and is composed of different stages linked to data collection and preparation, development of the ANN, and training and validation of the developed ANN.\u003c/p\u003e\n\u003ch2\u003e2.1 Data collection and preparation\u003c/h2\u003e\n\u003cp\u003eThe industrial park of the company in question is composed of knitting, dyeing and finishing, but the research is being carried out in the dyeing sector due to its high percentage of reprocessing.\u003c/p\u003e\n\u003cp\u003eA survey of the process variables was carried out for the selection of ANN input data, collecting a wide variety of data and samples of polyamide mesh fabrics and identifying different variations, including products with and without reprocessing.\u003c/p\u003e\n\u003cp\u003eThen, the dataset was treated according to the datasets obtained, and an initial architecture for the artificial neural network was defined, which was trained until it could represent the real data with relative accuracy. After the training phase, the network was validated with a new set of data (a step called cross-validation) to predict possible deviations in the quality of the finished product and, consequently, reprocessing. The software used was MATLAB.\u003c/p\u003e\n\u003ch2\u003e2.2 DEVELOPMENT OF THE ARTIFICIAL NEURAL NETWORK\u003c/h2\u003e\n\u003cp\u003eThe network architecture is determined by determining the number of layers, the number of neurons in each layer, and the proper activation function. Although there are some approaches in the literature to design the architecture of a neural network, we try not to delve into any of them because they have no theoretical foundation. Therefore, the network was designed by trial and error.\u003c/p\u003e\n\u003ch2\u003e2.3 ARTICAL NEURAL NETWORK TRAINING AND VALIDATION\u003c/h2\u003e\n\u003ch3\u003e2.3.1 Data Preparation\u003c/h3\u003e\n\u003cp\u003eThe training, validation, and testing datasets were generated in MATLAB.\u003c/p\u003e\n\u003ch3\u003e2.3.2 Adjustments and Optimization\u003c/h3\u003e\n\u003cp\u003eBased on the results of the assessment, adjustments were made to the network architecture as needed. Weights and bias are evaluated by the principal component analysis technique. Those neurons that have weights and bias with negligible values are excluded from the network architecture, and the network is retrained. The procedure is repeated until the result is reproduced. When this occurs, we have defined the number of layers and neurons sufficient for the network to adequately represent the process.\u003c/p\u003e\n\u003ch3\u003e2.3.3 Implementation in a Production Environment\u003c/h3\u003e\n\u003cp\u003eAfter validating the performance of the network, it will be implemented in the production environment to demonstrate its operation in the plant.\u003c/p\u003e\n\u003ch3\u003e2.3.4 Maintenance and Upgrading\u003c/h3\u003e\n\u003cp\u003eNetwork performance in production is monitored and adjusted as needed. If necessary, the network will be retrained with newer data to improve performance over time.\u003c/p\u003e"},{"header":"RESULTS AND DISCUSSION","content":"\u003cp\u003eThe following are preliminary results obtained regarding the data collected from the reprocessing of the textile processing sector of the company studied.\u003c/p\u003e\n\u003ch2\u003e3.1 Data collection\u003c/h2\u003e\n\u003cp\u003eTable 2 shows the fiber production and reprocessing conducted for each type of fiber. The percentages of reprocessed polyamide, PES/TXT and viscose fibers were on the order of 75%, 12% and 5%, respectively (Figure 10). In general, the higher reprocessing rates observed for the polyamide, PES/TXT and viscose fibers can be attributed to the inherent complexities of the production processes and the intrinsic properties of these materials. Moreover, this behavior can be attributed to the complex interaction between the intrinsic properties of these materials and the production procedures. Polyamide, known for its strength, can face challenges during production due to its nature, while PES/TXT, possibly a combination of polyester with other textile fibers, can be susceptible to specific manufacturing flaws. Viscose, despite its favorable properties, can be sensitive to certain production processes. The complexity of manufacturing processes and the sensitivity of these materials to variations in processing conditions also contribute to these higher reprocessing rates.\u003c/p\u003e\n\u003cp\u003eTable 2 - Amount of production and reprocessing of fibers produced from January to July 2023.\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"35.714285714285715%\" valign=\"bottom\"\u003e\n \u003cp\u003e\u003cstrong\u003eFiber\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.591836734693878%\" valign=\"bottom\"\u003e\n \u003cp\u003e\u003cstrong\u003eProduction (kg)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"bottom\"\u003e\n \u003cp\u003e\u003cstrong\u003eReprocessing (kg)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"35.714285714285715%\" valign=\"top\"\u003e\n \u003cp\u003eTotal Produced\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.591836734693878%\"\u003e\n \u003cp\u003e3.332.607,900\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e392.621,371\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"35.714285714285715%\" valign=\"top\"\u003e\n \u003cp\u003ePolyamide\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.591836734693878%\"\u003e\n \u003cp\u003e2.489.445,282\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e295.008,101\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"35.714285714285715%\" valign=\"top\"\u003e\n \u003cp\u003ePES/TXT\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.591836734693878%\"\u003e\n \u003cp\u003e484.226,291\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e49.206,986\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"35.714285714285715%\" valign=\"top\"\u003e\n \u003cp\u003eViscose\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.591836734693878%\"\u003e\n \u003cp\u003e102.047,540\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e19.034,677\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"35.714285714285715%\" valign=\"top\"\u003e\n \u003cp\u003eCotton\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.591836734693878%\"\u003e\n \u003cp\u003e99.666,517\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e17.067,888\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"35.714285714285715%\" valign=\"top\"\u003e\n \u003cp\u003ePolyester\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.591836734693878%\"\u003e\n \u003cp\u003e79.544,618\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e6.376,200\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"35.714285714285715%\" valign=\"top\"\u003e\n \u003cp\u003eTextured PES\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.591836734693878%\"\u003e\n \u003cp\u003e69.155,484\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e5.927,519\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"35.714285714285715%\" valign=\"top\"\u003e\n \u003cp\u003eSpecial polyamide\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.591836734693878%\"\u003e\n \u003cp\u003e8.491,068\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e\u0026nbsp;-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"35.714285714285715%\" valign=\"top\"\u003e\n \u003cp\u003eDifferentiated\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"29.591836734693878%\"\u003e\n \u003cp\u003e31,100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e\u0026nbsp;-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eSource: The author, (2023).\u003c/p\u003e\n\u003cp\u003eSpecifically, for the polyamide fibers, Table 3 shows the causes of the reprocesses generated. Notably, approximately 95% of these occurrences are classified as undefined, indicating a lack of clarity about their specific causes, which compromises the ability to identify and solve such problems. This lack of precision in identifying the causes prevents a targeted and effective intervention to address the issues related to polyamide reprocessing.\u003c/p\u003e\n\u003cp\u003eThe high rate of undefined reprocessing in polyamide fibers could be due to multiple factors. The complexity of the production processes of this fiber, which involves delicate chemical and mechanical steps, can lead to a wide range of possible failures, from variations in chemical composition to inadequacies in temperature or pressure during the process. In addition, the difficulty of monitoring all the variables involved can contribute to inaccuracies in identifying the causes, creating gaps in the understanding of the events that lead to reprocessing. This lack of clarity can be compounded by the absence of robust tracking systems or detailed recording methods, making it difficult to accurately attribute the causes of quality issues, thus resulting in this high percentage of undefined occurrences.\u003c/p\u003e\n\u003cp\u003eTable 3 - Causes of reprocesses generated in polyamide production.\u003c/p\u003e\n\u003cdiv align=\"Left\"\u003e\n \u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"79%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"bottom\"\u003e\n \u003cp\u003e\u003cstrong\u003eCauses\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.612244897959183%\" valign=\"bottom\"\u003e\n \u003cp\u003e\u003cstrong\u003eReprocessing (kg)\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"bottom\"\u003e\n \u003cp\u003e\u003cstrong\u003eReprocessing (%)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"top\"\u003e\n \u003cp\u003eGrand total\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.612244897959183%\"\u003e\n \u003cp\u003e295.008,101\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"top\"\u003e\n \u003cp\u003eIndefinite\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.612244897959183%\"\u003e\n \u003cp\u003e281.531,066\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e95\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"top\"\u003e\n \u003cp\u003eOperational failure\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.612244897959183%\"\u003e\n \u003cp\u003e3.452,581\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"top\"\u003e\n \u003cp\u003eTechnical Area Test\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.612244897959183%\"\u003e\n \u003cp\u003e3.438,815\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"top\"\u003e\n \u003cp\u003eWrong recipe\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.612244897959183%\"\u003e\n \u003cp\u003e2.496,690\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"top\"\u003e\n \u003cp\u003eRaw Material\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.612244897959183%\"\u003e\n \u003cp\u003e2.296,310\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"top\"\u003e\n \u003cp\u003eElectrical problem\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.612244897959183%\"\u003e\n \u003cp\u003e643,540\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"top\"\u003e\n \u003cp\u003eMechanical problem\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.612244897959183%\"\u003e\n \u003cp\u003e394,680\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"top\"\u003e\n \u003cp\u003eRaw material-yarn\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.612244897959183%\"\u003e\n \u003cp\u003e377,520\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"top\"\u003e\n \u003cp\u003eWrong concentration\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.612244897959183%\"\u003e\n \u003cp\u003e147,909\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd width=\"34.69387755102041%\" valign=\"top\"\u003e\n \u003cp\u003eTest new process\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"30.612244897959183%\"\u003e\n \u003cp\u003e32,610\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd width=\"34.69387755102041%\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003eSource: The author, (2023).\u003c/p\u003e\n\u003cp\u003eThe company has several dyeing lines in such a way that only one machine will be selected for the purpose of concluding this dissertation. The inclusion of all production lines, with their variables, would make it impossible to complete the present work within the regimental deadline. However, this work has an academic and didactic nature and can be explored by companies with enormous benefits.\u003c/p\u003e\n\u003cp\u003eIn the present work, the interest is to evaluate the input variables and the finished product in a static neural network. Intermediate steps, which can be performed by the company, including moving the architecture to a dynamic network, are not included.\u003c/p\u003e\n\u003cp\u003eFor the neural network to be applied, each qualification is assigned a numerical code. The code must be a number with a certain variation. For example, if the product should have a Ferrari red color coded, for example, as a number 5, and the result is 5.1, it will be essential that a tolerance index be defined by the company. The color is visually evaluated by the quality inspector, and in case of doubt, the spectrophotometer is read to verify that it is within the range of the panettone color chosen by the customer.\u003c/p\u003e\n\u003ch2\u003e3.2 IMPLEMENTATION OF THE ARTIFICIAL NEURAL NETWORK\u003c/h2\u003e\n\u003cp\u003eInitially, an ANN was implemented with data from MATLAB itself to test the approach it intended to employ. Figure 11 shows the structure of the network, whose structure was created randomly. There are 13 input data points, an intermediate layer with ten neurons, a tangential sigmoid activation function, and an output layer.\u003c/p\u003e\n\u003cp\u003eIn addition to the MATLAB software allowing the insertion of scripts, it is possible to visualize all events. Figure 12 shows the indexes related to network training.\u003c/p\u003e\n\u003cp\u003eFigure 12 - Evolution of network performance indexes. (a) Supervision of learning. (b) Histogram of error in relation to the predicted value and the wrong values, (c) evolution of the correlation coefficient, and (d) better performance obtained in relation to the actual data.\u003c/p\u003e\n\u003cp\u003eWe can conclude that the selected computational tool allows us to implement the proposed neural network with relative agility and the achievement of fast and visual results. To improve the visualization of the results, it is possible to increase the number of variables to be monitored and make the necessary adjustments without having to wait for the completion of the complete training of the network (in this case, the training is called epochs or epochs, which is the number of times the optimization algorithm was used to determine the parameters of the network (bias and weights). The decision variables are listed in Figure 13.\u003c/p\u003e\n\u003cp\u003eFigure 13 - Variables for decision-making about the network architecture and its performance during the training phase.\u003c/p\u003e\n\u003cp\u003eBy analyzing the results, we can conclude that the selected platform as well as the approach are suitable for the completion of the work.\u003c/p\u003e"},{"header":"CONCLUSIONS","content":"\u003cp\u003eThe higher rates of reprocessing identified in polyamide fibers suggest distinct challenges in their respective production chains. As synthetic fibers, they undergo complex processes and can generate out-of-specification products due to variations in chemical and physical parameters. These findings also indicate that the high percentage of undefined occurrences of reprocessing in polyamide fibers reveals challenges in accurately identifying the causes, possibly due to the complexity of the processes and the difficulty in monitoring all the variables involved.\u003c/p\u003e \u003cp\u003eThe integration of advanced technologies, such as ANNs, could play a crucial role in reducing reprocessing in polyamide manufacturing. The implementation of RNA can offer an innovative approach to analyzing vast datasets sourced from production processes, identifying subtle patterns and correlations between variables that can be difficult to detect by conventional methods. By training neural networks with detailed information about the polyamide manufacturing process, it is possible to develop predictive models capable of anticipating possible failures, minimizing the occurrence of out-of-specification products. This proactive approach could significantly contribute to reducing reprocessing, optimizing process efficiency and improving the quality of final polyamide products. Therefore, incorporating artificial intelligence technologies such as artificial neural networks has emerged as a vital necessity to propel the PA manufacturing industry toward more efficient and lower-waste practices.\u003c/p\u003e \u003cp\u003eAs previously discussed, the selected tool proved to be feasible for the approach intended in the present study.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eCONFLICT OF INTEREST\u003c/h2\u003e \u003cp\u003eOn behalf of all the authors, the corresponding author declares that there are no conflicts of interest.\u003c/p\u003e \u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003e\u003cspan\u003eABIT- Associa\u0026ccedil;\u0026atilde;o (2023) Brasileira da Ind\u0026uacute;stria T\u0026ecirc;xtil e de Confec\u0026ccedil;\u0026atilde;o. Perfil do Setor 2022. Dispon\u0026iacute;vel em: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.abit.org.br/cont/perfil-do-setor\u003c/span\u003e\u003c/span\u003e. Acesso em: 12 set\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eABIT- Associa\u0026ccedil;\u0026atilde;o Brasileira da Ind\u0026uacute;stria T\u0026ecirc;xtil e de Confec\u0026ccedil;\u0026atilde;o. Ind\u0026uacute;stria t\u0026ecirc;xtil brasileira (2018) Dispon\u0026iacute;vel em: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.abit.org.br/\u003c/span\u003e\u003c/span\u003e. Acesso em: 12 set. 2023\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eADANUR, Sabit (2017) Handbook of industrial textiles. Lancaster: Wellingt Sears Co, 839 p\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eBAHLMANN C, HEIDEMANN G (1999) ; RITTER, Helge. Artificial neural networks for automated quality control of textile seams. Pattern Recognition, [S.L.], v. 32, n. 6, p. 1049\u0026ndash;1060, jun. Elsevier BV. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1016/s0031-3203(98)00128-9\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eBARANI H, MONTAZER, Majid A Review on Applications of Liposomes in Textile Processing. Journal Of Liposome Research, [S.L.], v. 18, n. 3, p. 249\u0026ndash;262, jan. 2008. Informa UK Limited. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1080/08982100802354665\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eBEALE R, JACKSON T (1990) Neural Computing - An Introduction. Crc, Eua, p 256\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eArthur BROADBENT (2011) D. Basic Principles of Textile Coloration. Society Of Dyers And Colorizts, West Yorkshire, p 579\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eCHATTOPADHYAY R, GUHA A ARTIFICIAL NEURAL NETWORKS: applications to textiles. Textile Progress, [S.L.], v. 35, n. 1, p. 1\u0026ndash;46, mar. 2004. Informa UK Limited. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1080/00405160408688961\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eCHEN C-L, 3 abr (2019) Value Creation by SMEs Participating in Global Value Chains under Industry 4.0 Trend: case study of textile industry in taiwan. Journal Of Global Information Technology Management, [S.L.], v. 22, n. 2, p. 120\u0026ndash;145,. Informa UK Limited. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1080/1097198x.2019.1603512\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eDORAN EC, SAHIN, Cenk (2019) The prediction of quality characteristics of cotton/elastane core yarn using artificial neural networks and support vector machines. Textile Research Journal, [S.L.], v. 90, n. 13\u0026ndash;14, p. 1558\u0026ndash;1580, 27 dez. SAGE Publications. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1177/0040517519896761\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eDUARTE, Adriana Yumi Sato. Proposta de integra\u0026ccedil;\u0026atilde;o entre ferramentas de avalia\u0026ccedil;\u0026atilde;o de ciclo de vida do produto e Ind\u0026uacute;stria 4.0 (Industrie 4.0): estudo de caso da ind\u0026uacute;stria t\u0026ecirc;xtil e de confec\u0026ccedil;\u0026atilde;o brasileira (2017) 125 f. Tese (Doutorado) - Curso de Engenharia Mec\u0026auml;nica, Centro Tecnol\u0026oacute;gico, Universidade Estadual de Campinas, Campinas, 2017\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eFALANI LAra\u0026uacute;jo (2020) Catia Rosana Lange de; FORNO, Ana Julia dal. Mapeamento da literatura sobre as tecnologias da ind\u0026uacute;stria 4.0 no segmento t\u0026ecirc;xtil brasileiro. Brazilian J Dev 42437\u0026ndash;42452 Brazilian Journal of Development. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.34117/bjdv6n7-019\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eFINOCCHIO MA (2014) F. No\u0026ccedil;\u0026otilde;es de redes neurais artificiais, Universidade Tecnol\u0026oacute;gica Federal do Paran\u0026amp;#225\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eFUJITA, Renata Mayumi Lopes; JORENTE, Maria Jos\u0026eacute;. A ind\u0026uacute;stria t\u0026ecirc;xtil no Brasil: uma perspectiva hist\u0026oacute;rica e cultural. Modapalavra E-Peri\u0026oacute;dico, [S.L.], v. 8, n. 15, p. 153\u0026ndash;174, 1 jan. 2015. Universidade do Estado de Santa Catarina. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.5965/1982615x08152015153\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eGONDIM, Mauricio Lopes. Estudos comparativo do \u0026acirc;ngulo de contato em tecido de malha de poliamida tratados com plasma. (2016) 68 f. Disserta\u0026ccedil;\u0026atilde;o (Mestrado) - Curso de T\u0026ecirc;xtil e Moda, Escola de Artes Ci\u0026ecirc;ncias e Humanidades, Universidade de S\u0026atilde;o Paulo, S\u0026atilde;o Paulo, 2016\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eHASSABIS D, KUMARAN, Dharshan; SUMMERFIELD C (2017) ; BOTVINICK, Matthew. Neuroscience-Inspired Artificial Intelligence. Neuron, [S.L.], v. 95, n. 2, p. 245\u0026ndash;258, jul. Elsevier BV. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1016/j.neuron.2017.06.011\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003ePatrick HUI, CHAN CL, Keith CC, YEUNG KW, NG, Frency SF (2007) Application of artificial neural networks to the prediction of sewing performance of fabrics. International Journal Of Clothing Science And Technology, [S.L.], v. 19, n. 5, p. 291\u0026ndash;318, 9 out. Emerald. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1108/09556220710819500\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eIEMI- Instituto de Estudos e Marketing. O Panorama Setorial da Ind\u0026uacute;stria T\u0026ecirc;xtil Brasileira (2022) IEMI. Dispon\u0026iacute;vel em: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.iemi.com.br/brasil-textil-2022/\u003c/span\u003e\u003c/span\u003e. Acesso em: 26 out. 2023\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eJEYARAJ PR (2019) ; NADAR, Edward Rajan Samuel. Effective textile quality processing and an accurate inspection system using the advanced deep learning technique. Textile Research Journal, [S.L.], v. 90, n. 9\u0026ndash;10, p. 971\u0026ndash;980, 23 out. SAGE Publications. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1177/0040517519884124\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eNERURKAR JOSHIM, Madhura BADHE, Pallavi, ADIVAREKAR, Ravindra. Scouring of cotton using marine pectinase. Journal Of Molecular Catalysis B:, Enzymatic [S.L.], v. 98, p. 106\u0026ndash;113. Elsevier BV. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1016/j.molcatb.2013.10.010\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eKANT, Rita (2012) Textile dyeing industry an environmental hazard. Natural Science, [S.L.], v. 04, n. 01, p. 22\u0026ndash;26, Scientific Research Publishing, Inc. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.4236/ns.2012.41004\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eKang-Ting KUC-Chun;CHIENC-Fu;MA Digital transformation to empower smart production for Industry 3.5 and an empirical study for textile dyeing. Computers \u0026amp; Industrial Engineering, [S.L.], v. 142, p. 106297, abr. 2020. Elsevier BV. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1016/j.cie.2020.106297\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eKUMAR H, KAUR, Gagandeep PRIYA Influence of tetra ethyl ammonium bromide (C2H5)4NBr on the aggregation behavior of surface active ionic liquid 1-tetradecyl-3-methylimidazolium bromide [C14mim][Br]. Journal Of Molecular Liquids, [S.L.], v. 313, p. 113431, set. 2020. Elsevier BV. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1016/j.molliq.2020.113431\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eLI W, WEI Z, LIU, Zhengdong DU, Yujun; ZHENG, Jiahui; WANG H, ZHANG (2021) Shuo. Qualitative identification of waste textiles based on near-infrared spectroscopy and the back propagation artificial neural network. Textile Research Journal, [S.L.], v. 91, n. 21\u0026ndash;22, p. 2459\u0026ndash;2467, 9 abr. SAGE Publications. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1177/00405175211007516\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eCHANG LIUJ, Huihong FORREST, YANG JY-L Baohua. Influence of artificial intelligence on technological innovation: evidence from the panel data of China\u0026apos;s manufacturing sectors. Technological Forecasting And Social Change, [S.L.], v. 158, p. 120142, set. 2020. Elsevier BV. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1016/j.techfore.2020.120142\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eMADHU A, CHAKRABORTY JN (2017) mar. Developments in application of enzymes for textile processing. Journal Of Cleaner Production, [S.L.], v. 145, p. 114\u0026ndash;133, Elsevier BV. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1016/j.jclepro.2017.01.013\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eMALUF E, KOLBE W (2003) Dados T\u0026eacute;cnicos para an Ind\u0026uacute;stria T\u0026ecirc;xtil. 2.ed. S\u0026atilde;o Paulo: IPT \u0026ndash; Instituto de Pesquisas Tecnol\u0026oacute;gicas do Estado de S\u0026atilde;o Paulo. ABIT \u0026ndash; Associa\u0026ccedil;\u0026atilde;o Brasileira da Ind\u0026uacute;stria T\u0026ecirc;xtil e de Confec\u0026ccedil;\u0026atilde;o\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eMATHIS MR, KHETERPAL S; NAJARIAN, Kayvan. Artificial Intelligence for Anesthesia: what the practicing clinician needs to know., Anesthesiology (2018) [S.L.], v. 129, n. 4, p. 619\u0026ndash;622, 1 out. Ovid Technologies (Wolters Kluwer Health). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1097/aln.0000000000002384\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003ePIMENTEL, Ren\u0026ecirc; C, Iran Siqueira (2011) ; LIMA,. Rela\u0026ccedil;\u0026atilde;o trimestral de longo prazo entre indicadores de liquidez e de rentabilidade: evid\u0026ecirc;ncia de empresas do setor t\u0026ecirc;xtil. Revista de Administra\u0026ccedil;\u0026atilde;o, [S.L.], v. 46, n. 3, p. 275\u0026ndash;289, 14 set. Business Department, School of Economics, Business \u0026amp; Accounting USP. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.5700/rausp1012\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eRAMESH A, KAMBHAMPATI C, MONSON, Jrt, DREW (2004) Pj. Artificial intelligence in medicine. Annals of The Royal College Of Surgeons Of England, [S.L.], v. 86, n. 5, p. 334\u0026ndash;338, 1 set. Royal College of Surgeons of England. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1308/147870804290\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eRUSCHIONI PB, ALFIERI P (2010) ; Estudo da influ\u0026ecirc;ncia da temperatura de termofixa\u0026ccedil;\u0026atilde;o 76 nas propriedades f\u0026iacute;sicas de malhas de poliamida 6.6. Qu\u0026iacute;mica T\u0026ecirc;xtil n \u0026deg; 99. S\u0026atilde;o Paulo. Junho, Pg. 52\u0026ndash;61\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eSALEM, Vidal (2010) Tingimento t\u0026ecirc;xtil: fibras, conceitos e tecnologias. Edgard Blucher, S\u0026atilde;o Paulo, p 297\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eDE SALEMV, MARCHI A (2005) MENEZES, Felipe Gon\u0026ccedil;alves de. O beneficiamento t\u0026ecirc;xtil na pr\u0026aacute;tica, 1 edn. Golden Qu\u0026iacute;mica do Brasil, S\u0026atilde;o Paulo\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eSAMSAMI S, MOHAMADIZANIANI M, SARRAFZADEH M-H, Eldon RENE, FIROOZBAHR R (2020) Meysam. Recent advances in the treatment of dye-containing wastewater from textile industries: overview and perspectives. Process Safety And Environmental Protection, [S.L.], v. 143, p. 138\u0026ndash;163, nov. Elsevier BV. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1016/j.psep.2020.05.034\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eSAMSAMI S, MOHAMADIZANIANI M, SARRAFZADEH M-H, Eldon RENE, FIROOZBAHR R (2020) Meysam. Recent advances in the treatment of dye-containing wastewater from textile industries: overview and perspectives. Process Safety And Environmental Protection, [S.L.], v. 143, p. 138\u0026ndash;163, nov. Elsevier BV. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1016/j.psep.2020.05.034\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eSANDIN G, PETERS, Greg M (2018) Environmental impact of textile reuse and recycling \u0026ndash; A review. Journal Of Cleaner Production, [S.L.], v. 184, p. 353\u0026ndash;365, maio. Elsevier BV. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1016/j.jclepro.2018.02.266\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eKOVAčEVIć SCHWARZI (2017) Stana. Textile Application: from need to imagination. Textiles For Advanced Applications, [S.L.], v. 1, n. 5, p. 312\u0026ndash;325, 20 set. InTech. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.5772/intechopen.68376\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eAlok SIKKAMPSARKAR, GARG Samridhi. Artificial intelligence (AI) in textile industry operational modernization. Research Journal Of Textile And Apparel, [S.L.], p. 258\u0026ndash;268, 12 abr. 2022. Emerald. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1108/rjta-04-2021-0046\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eSMIRNOV, Evgeny A, TIMOSHENKO, Denis M, ANDRIANOV, Serge N (2014) Comparison of Regularization Methods for ImageNet Classification with Deep Convolutional Neural Networks. Aasri Procedia, [S.L.], v. 6, p. 89\u0026ndash;94, Elsevier BV. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1016/j.aasri.2014.05.013\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eChih-Ching SUK-Hua;LINJ-Horng;LIN (2007) Influence of reprocessing on the mechanical properties and structure of polyamide 6. Journal Of Materials Processing Technology, [S.L.], v. 192\u0026ndash;193, p. 532\u0026ndash;538, out. Elsevier BV. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1016/j.jmatprotec.2007.04.056\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eUDDIN F (2021) Environmental hazard in textile dyeing wastewater from local textile industry. Cellulose, [S.L.], v. 28, n. 17, p. 10715\u0026ndash;10739, 2 out. Springer Science and Business Media LLC. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.1007/s10570-021-04228-4\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eTiwari VIBHAV, Garg SAMAR, Bhat SAMARTH, Kumar YSANCHIT, Prakash Bhat MANIKANDANJPMATHIAZHAGAN, Tamilselvan M, Periyasamy S (2023) Design of a Real-Time Warp and Weft Measurement System for Fabric Analysis in Textile Testing Using Machine Learning, International Conference on Network, Multimedia and Information Technology (NMITCON), p.1\u0026ndash;7, 2023\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eTyrone VIGO (2002) L. Textile processing and properties: preparation, dyeing, finishing and performance, 3 edn. Elsevier Science B.V., Amsterdam, p 486\u003c/span\u003e\u003c/li\u003e\n \u003cli\u003e\u003cspan\u003eZHANG J, HE, Lin, CHENG, Longdi (1145) Is China\u0026rsquo;s Textile Industry Still a Labor-Intensive Industry? Fibers And Textiles In Eastern Europe, [S.L.], v. 29, n. p. 13\u0026ndash;16, 28 fev. 2021. Walter de Gruyter GmbH. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://dx.doi.org/10.5604/01.3001.0014.5038\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Table","content":"\u003cp\u003eTable 1 is available in the Supplementary Files section\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Artificial intelligence, Textile processing, Polyamide.","lastPublishedDoi":"10.21203/rs.3.rs-3996611/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3996611/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe Brazilian textile industry is an essential pillar of the country's economy, standing out globally as the fifth-largest textile hub and the fourth-largest in the clothing segment. However, one of the critical challenges faced by this sector is the reprocessing of fabrics, which leads to delivery delays, quality impacts, increased costs, and environmental impacts. Therefore, the aim of this study is to identify reprocessing in the dyeing process of a textile industry through preestablished patterns using a neural network. To achieve this goal, this research is being conducted in partnership with a company in the sector, focusing on data collection, preparation, processing, training and validating the neural network. Specifically, the focus is on the data collected from the production of polyamide, where approximately 95% of the reprocessing is classified as undefined, making the identification and precise resolution of these issues challenging. Thus, this research aims not only to enhance the efficiency of polyamide production but also to contribute to resource savings and compliance with environmental commitments, consolidating the concept of sustainability in the textile industry. The incorporation of artificial intelligence, such as neural networks, has emerged as an essential strategy to drive the textile industry toward more efficient and less impactful practices.\u003c/p\u003e","manuscriptTitle":"Development of an artificial neural network to maximize the reproducibility of dyeing polyamide fabrics","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-02-29 07:50:09","doi":"10.21203/rs.3.rs-3996611/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"784ac006-2207-422b-9da4-a7f023183f5d","owner":[],"postedDate":"February 29th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":29029185,"name":"Industrial Engineering"}],"tags":[],"updatedAt":"2024-02-29T07:50:09+00:00","versionOfRecord":[],"versionCreatedAt":"2024-02-29 07:50:09","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-3996611","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3996611","identity":"rs-3996611","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}