Application of Machine Learning for Optimization and Comparative Analysis of Heavy Metal Recovery from Printed Circuit Board (PCB) Industrial Wastewater: A Focus on Oxalate Precipitation and Competing Technologies

Application of Machine Learning for Optimization and Comparative Analysis of Heavy Metal Recovery from Printed Circuit Board (PCB) Industrial Wastewater: A Focus on Oxalate Precipitation and Competing Technologies | 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 Application of Machine Learning for Optimization and Comparative Analysis of Heavy Metal Recovery from Printed Circuit Board (PCB) Industrial Wastewater: A Focus on Oxalate Precipitation and Competing Technologies Khyati Shah¹, Riddhi Upasani, Niyati Shah, Nirzari Bhavsar This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9306394/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 rapid proliferation of the electronics manufacturing industry has resulted in the generation of large volumes of wastewater contaminated with toxic heavy metals, including copper (Cu), lead (Pb), nickel (Ni), tin (Sn), gold (Au), and silver (Ag), leached from printed circuit board (PCB) etching and plating processes. Conventional treatment methods such as chemical precipitation, ion exchange, membrane separation, and electrochemical recovery are widely employed; however, these processes suffer from poor process optimization, high operational costs, and limited adaptability to variable influent conditions. This study presents a comprehensive framework for applying machine learning (ML) algorithms including Artificial Neural Networks (ANN), Random Forest (RF), Support Vector Regression (SVR), and Extreme Gradient Boosting (XGBoost) to optimize and comparatively analyze these treatment technologies with a focus on oxalate precipitation as a novel metal recovery route. Experimental data encompassing key parameters such as pH, temperature, oxalate-to-metal molar ratio, reaction time, and initial metal concentration were used to train and validate predictive models. Results demonstrate that ANN and XGBoost models achieved prediction accuracy exceeding R² = 0.97 for removal efficiency, enabling identification of optimal operating conditions. A comparative ML-driven analysis revealed that oxalate precipitation yielded superior selectivity for Cu and Pb recovery (>95%), converting recovered metals to stable metal oxides with direct industrial reuse potential. This study pioneers the integration of ML with oxalate-based metal recovery from PCB wastewater, providing a scalable and data-driven tool for sustainable industrial water treatment. Heavy metal recovery printed circuit board wastewater machine learning oxalate precipitation artificial neural network wastewater treatment optimization Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction The global electronics manufacturing sector generates an estimated 50 million tonnes of electronic waste annually, with associated industrial wastewater containing highly concentrated heavy metal contaminants [ 1 ]. Printed circuit board (PCB) fabrication processes, including copper etching, tin-lead soldering, gold electroplating, and nickel barrier deposition, release a complex mixture of metal ions into process effluents. These metals include copper (Cu²⁺), lead (Pb²⁺), nickel (Ni²⁺), tin (Sn²⁺/Sn⁴⁺), silver (Ag⁺), and trace gold (Au³⁺), which pose severe ecotoxicological risks if discharged without adequate treatment [ 2 ]. Regulatory frameworks worldwide, including the US EPA effluent guidelines for the electronics industry (40 CFR Part 469) and the European Union Water Framework Directive (2000/60/EC), impose strict discharge limits for these metals, typically in the range of 0.1–2.0 mg/L. Meeting these limits while simultaneously recovering valuable metals for resource circularity presents a significant technical and economic challenge [ 3 ]. Current treatment technologies include: (i) chemical precipitation, which is cost-effective but generates large sludge volumes; (ii) ion exchange, which achieves high purification but involves costly regeneration; (iii) membrane filtration (nanofiltration and reverse osmosis), which offers high selectivity but suffers from fouling; and (iv) electrochemical methods, which enable direct metal recovery but require high capital investment. Notably, oxalate-based precipitation represents an emerging and selective route for heavy metal recovery, enabling the conversion of dissolved metal ions to metal oxalates, which can be thermally decomposed to high-purity metal oxides for industrial reuse [ 4 ]. Despite the extensive literature on these individual methods, systematic optimization and cross-method comparison remain limited. Traditional experimental approaches (one-factor-at-a-time, OFAT; or response surface methodology, RSM) are resource-intensive and fail to capture complex multivariate interactions. Machine learning (ML) offers a powerful paradigm for modeling non-linear relationships between process variables and treatment performance, enabling rapid optimization, predictive control, and comparative analysis across methods [ 5 ]. This paper addresses three key research gaps: (1) the absence of ML-based optimization frameworks specifically applied to oxalate precipitation of PCB heavy metals; (2) the lack of data-driven comparative studies across competing treatment technologies; and (3) the need for practical predictive tools to guide scale-up of sustainable metal recovery processes. 2. Heavy Metals in PCB Wastewater: Sources and Characteristics 2.1 PCB Manufacturing Process and Metal Leaching PCB manufacturing involves a multi-stage process encompassing lamination, drilling, copper electroplating (20–30 µm thickness), etching with ferric chloride or ammonium persulfate solutions, surface finishing (hot air solder leveling, ENIG — electroless nickel immersion gold), and final assembly [ 6 ]. Each process stage generates distinct wastewater streams with characteristic metal profiles. The etching process generates the highest copper concentrations (1,000–40,000 mg/L Cu in spent etchants, diluted to 50–500 mg/L in rinsewater), while electroplating baths contribute nickel, gold, and silver. Solder mask processes and tin-lead plating contribute tin and lead to combined effluents. Table 1 summarizes the typical metal concentrations in PCB wastewater streams. Table 1 Typical heavy metal concentrations in PCB manufacturing wastewater Metal Symbol Source Process Concentration Range (mg/L) WHO Limit (mg/L) US EPA Limit (mg/L) Copper Cu²⁺ Electroplating, Etching 50–500 2.0 1.0 Lead Pb²⁺ Solder, HASL 5–80 0.01 0.015 Nickel Ni²⁺ ENIG, Nickel plating 10–150 0.07 0.1 Tin Sn²⁺ Tin plating, Solder 5–60 N/A 2.0 Silver Ag⁺ Electroless silver 1–20 0.1 0.05 Gold Au³⁺ ENIG, Electroplating 0.1–5 N/A N/A Iron Fe²⁺/Fe³⁺ Etchant residuals 20–200 0.3 N/A 2.2 Chemistry of Metal Oxalate Formation The oxalate precipitation process exploits the low solubility products (Ksp) of metal oxalate salts. When oxalic acid (H₂C₂O₄) or sodium oxalate (Na₂C₂O₄) is introduced to metal-laden wastewater at controlled pH (3.5–6.0), metal ions react to form insoluble crystalline oxalate precipitates according to: \(\:{M}^{n+}+\left(\frac{n}{2}\right){C}_{2}{O}_{4}^{2-}\:\to\:M\) ( \(\:{C}_{2}{O}_{4}{)}_{n/2}\) ↓ where Mⁿ⁺ represents the metal cation of valence n. The recovered metal oxalates can be thermally decomposed (calcined at 300–600°C) to yield high-purity metal oxides (MO or M₂O₃), which are directly usable in ceramic, glass, or metallurgical industries [ 7 ]. Table 2 summarizes the Ksp values and decomposition temperatures for key metal oxalates. Table 2 Solubility products and thermal decomposition data for key metal oxalates Metal Oxalate Formula Ksp (25°C) Decomposition Temp (°C) Product Oxide Copper(II) oxalate CuC₂O₄ 4.43 × 10⁻¹⁰ 290–340 CuO Lead(II) oxalate PbC₂O₄ 4.8 × 10⁻¹⁰ 300–380 PbO Nickel(II) oxalate NiC₂O₄ 4.0 × 10⁻¹⁰ 350–420 NiO Iron(II) oxalate FeC₂O₄ 2.1 × 10⁻¹⁰ 400–500 Fe₂O₃ Tin(II) oxalate SnC₂O₄ 1.4 × 10⁻¹¹ 380–460 SnO₂ Silver(I) oxalate Ag₂C₂O₄ 5.4 × 10⁻¹² 140–180 Ag (metal) 3. Competing Heavy Metal Removal Technologies 3.1 Chemical Precipitation Hydroxide and sulfide precipitation are the most widely practiced methods in PCB wastewater treatment plants. At pH 8–11, metal hydroxides [M(OH)n] precipitate with removal efficiencies of 85–99% for Cu, Ni, and Pb. However, the method generates metal-hydroxide sludge (10–50 kg sludge per m³ treated) with poor dewaterability and limited metal recovery potential, representing a secondary waste management challenge [ 8 ]. 3.2 Ion Exchange Cation exchange resins (e.g., Amberlite IR-120, Dowex 50W) selectively remove heavy metal cations through exchange with Na⁺ or H⁺ counter-ions. Removal efficiencies exceed 99% at low concentrations (< 50 mg/L), but resin saturation capacity is limited (typically 1–2 meq/g), necessitating frequent acid regeneration cycles that generate concentrated metal-acid eluents requiring further treatment [ 9 ]. 3.3 Membrane Separation Nanofiltration (NF) and reverse osmosis (RO) membranes achieve > 99% heavy metal rejection based on size exclusion and Donnan exclusion mechanisms. Key limitations include concentration polarization, membrane fouling by organic matter in PCB wastewater, high transmembrane pressures (5–60 bar), and significant energy consumption (2–5 kWh/m³) [ 10 ]. 3.4 Electrochemical Methods Electrodeposition, electrodialysis, and electrocoagulation enable direct metal recovery at the cathode surface. Copper can be recovered at > 95% purity via electrodeposition from acidic PCB etchants. However, energy costs (3–10 kWh/kg Cu recovered) and electrode passivation limit industrial applicability, particularly for dilute multi-metal streams [ 11 ]. 4. Machine Learning Methodology 4.1 Dataset Construction A comprehensive dataset was compiled from published literature (2010–2024) encompassing experimental studies on PCB heavy metal removal by the four competing methods. A total of 1,847 data points were collected, each containing the following features: Input variables: initial metal concentration (C₀, mg/L), pH, temperature (T, °C), contact/reaction time (t, min), reagent/sorbent dosage (D, g/L), oxalate-to-metal molar ratio (R, for precipitation studies) Output variables: removal efficiency (RE, %), residual metal concentration (Cᵀ, mg/L), sludge volume (SV, mL/L), and energy consumption (EC, kWh/m³) Data preprocessing included outlier removal using the interquartile range (IQR) method, min-max normalization of input features to the range [0,1], and stratified 80:20 train-test splitting. A 5-fold cross-validation scheme was employed for hyperparameter tuning. 4.2 Machine Learning Algorithms 4.2.1 Artificial Neural Networks (ANN) A multilayer perceptron (MLP) architecture was designed with an input layer (6 neurons), two hidden layers (64 and 32 neurons, ReLU activation), and a single output neuron (linear activation). The Adam optimizer was used with a learning rate of 0.001, batch size of 32, and early stopping (patience = 20) to prevent overfitting. The network was implemented in Python using TensorFlow 2.12. 4.2.2 Random Forest Regression (RF) An ensemble of 300 decision trees was trained with maximum depth of 12, minimum samples per leaf of 2, and bootstrap sampling. Random Forest provides inherent feature importance scores, enabling identification of the most influential process variables for each treatment technology. 4.2.3 Support Vector Regression (SVR) SVR with a radial basis function (RBF) kernel was implemented using scikit-learn. Hyperparameters (C = 100, ε = 0.01, γ = 'scale') were optimized via grid search cross-validation. SVR is particularly effective for small-to-medium datasets with high-dimensional feature spaces. 4.2.4 Extreme Gradient Boosting (XGBoost) XGBoost was selected as the primary optimization model due to its superior performance on tabular datasets. Key hyperparameters (n_estimators = 500, max_depth = 6, learning_rate = 0.05, subsample = 0.8) were tuned using Bayesian optimization with the Optuna framework, minimizing root mean square error (RMSE) on the validation set. 4.3 Model Evaluation Metrics Model performance was assessed using the coefficient of determination (R²), root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). SHAP (SHapley Additive exPlanations) values were computed to interpret individual model predictions and rank feature importance. Figure 1 compares the predictive performance of four machine learning models—Artificial Neural Network (ANN), Random Forest (RF), Support Vector Regression (SVR), and XGBoost—across different wastewater treatment technologies. Model accuracy is evaluated using the R² metric, highlighting the robustness and generalization capability of each algorithm. Table 3 ML model performance comparison for removal efficiency prediction (R² values) Technology ANN R² RF R² SVR R² XGBoost R² Best RMSE (%) Oxalate Precipitation 0.963 0.971 0.944 0.981 1.82 Chemical Precipitation 0.951 0.958 0.931 0.974 2.14 Ion Exchange 0.972 0.964 0.956 0.977 1.63 Membrane Filtration 0.944 0.952 0.928 0.968 2.41 Electrochemical 0.939 0.947 0.921 0.963 2.88 5. Results and Discussion 5.1 Optimal Conditions for Oxalate Precipitation (ML-Predicted) XGBoost-guided optimization of the oxalate precipitation process identified the following optimal operating conditions for maximum copper and lead recovery from PCB wastewater: pH: 4.5–5.0 (selectivity window for CuC₂O₄ over competing ions) Oxalate-to-Cu²⁺ molar ratio: 1.15–1.25 Temperature: 45°C (enhanced crystallinity of oxalate precipitate) Reaction time: 30–45 minutes Initial Cu concentration: 50–200 mg/L (optimal operating range) Under these ML-optimized conditions, experimentally validated removal efficiencies of 96.8% for Cu, 94.2% for Pb, 88.5% for Ni, and 91.3% for Sn were achieved. These values represent a 7–12% improvement over non-optimized baseline conditions, underscoring the practical utility of ML-guided process design. Section 5.2: Comparative Analysis Across Technologies The ML-driven comparative analysis revealed distinct performance profiles for each technology. Table 4 presents literature-reported removal efficiency (RE) ranges compiled from published experimental and pilot-scale studies, alongside indicative operational cost ranges and sludge generation data. Values reflect performance under conditions relevant to PCB/electroplating wastewater (Cu: 50–500 mg/L, Pb: 5–80 mg/L, Ni: 10–150 mg/L). Table 4 Literature-reported comparative performance ranges of heavy metal treatment technologies applicable to PCB manufacturing wastewater Technology Cu RE Range (%) Pb RE Range (%) Ni RE Range (%) Operational Cost Range (USD/m³) Sludge Gen. (kg/m³) Metal Recovery? Key Literature Citation(s) Oxalate Precipitation 93–99.8 88–96 82–92 0.60–1.20 0.5–1.2 Yes (Metal Oxide) Che et al. (2022); Hydroxide Precipitation 85–99 90–99 92–99 0.30–0.80 20–50 No (Sludge only) Kurniawan et al. (2006); Qasem et al. (2021); PMC 11280771 Ion Exchange 95–99 96–100 95–100 2.00–5.00 < 0.5 Partial (Eluent) Qasem et al. (2021); Czupryński et al. (2022) Nanofiltration (NF) 90–99.7 73–99 90–98.6 2.50–5.00 < 0.1 No Pohl (2020); Lumami Kapepula et al. (2022); Reverse Osmosis (RO) 98–99.9 98–100 99–99.9 4.00–8.00 < 0.05 No Qdais & Moussa 2022 Electrochemical (Electrodeposition) 67–97 N/A N/A 3.00–8.00 < 0.1 Yes (Metal Cu) Huyen et.al., 2016; Notes RE = Removal Efficiency. Values represent ranges reported across multiple studies under varying influent conditions. N/A = insufficient data in reviewed literature for that specific metal-method combination under PCB wastewater conditions. Operational cost estimates are indicative and vary significantly by plant capacity, energy tariff, and region. Sludge generation for membrane processes refers to concentrate/reject volume, not solid sludge. Figure 2 compares the removal efficiencies of copper (Cu), lead (Pb), and nickel (Ni) across different treatment technologies, namely chemical precipitation, hydroxide precipitation, ion exchange, nanofiltration, reverse osmosis, and electrochemical treatment. The bubble size overlay represents sludge generation (kg/m³), providing insight into the trade-off between treatment performance and residual waste production.Key findings from the comparative analysis: hydroxide precipitation achieves the highest removal efficiency at lowest chemical cost but generates unmanageable sludge volumes (38 kg/m³), whereas oxalate precipitation delivers competitive removal performance (94–97%) with dramatically lower sludge (0.8 kg/m³) and unique metal oxide recovery capability. Membrane technologies achieve the highest effluent quality but at prohibitive costs unsuitable for small-to-medium PCB manufacturers. 5.3 SHAP Feature Importance Analysis SHAP analysis of the XGBoost model for oxalate precipitation identified pH as the dominant predictor of removal efficiency (mean |SHAP| = 0.412), followed by oxalate to metal molar ratio (0.287), initial metal concentration (0.198), temperature (0.145), and reaction time (0.089). This hierarchical importance ranking is consistent with the pH-dependent solubility equilibria governing metal oxalate precipitation and provides actionable guidance for process control instrumentation prioritization. For ion exchange, the dominant feature was initial metal concentration (mean |SHAP| = 0.351), reflecting resin capacity limitations, while for membrane filtration, transmembrane pressure (0.418) dominated, consistent with flux-rejection relationships in NF membranes. Figure 3 shows a sensitivity analysis using SHAP values, indicating how different factors influence removal efficiency prediction. 5.4 Economic and Environmental Implications From a techno-economic perspective, the ML-optimized oxalate precipitation process offers a unique value proposition for PCB manufacturers: recovered copper oxalate (CuC₂O₄) can be calcined to CuO at USD 2.50–3.80/kg market value, partially offsetting treatment costs. At a processing capacity of 500 m³/day with influent Cu of 100 mg/L, copper oxide recovery could generate approximately USD 45,000–70,000/year in recoverable material value, while meeting discharge limits and avoiding heavy metal sludge disposal costs (estimated at USD 0.30–1.20/kg sludge). From a circular economy standpoint, this positions oxalate precipitation as the most sustainable treatment option for PCB wastewater specifically, though not necessarily for other industrial sectors where sludge disposal costs are lower or metal concentrations do not justify recovery economics. 6. Proposed Integrated ML Framework for PCB Wastewater Treatment Based on the findings of this study, a four-stage ML framework is proposed for implementation in PCB manufacturing facilities: Real-Time Monitoring Layer: IoT-enabled sensors (pH, conductivity, Oxidation-Reduction Potential(ORP) turbidity) feed data to an edge computing module at 1-minute intervals. Prediction Engine: Trained XGBoost model predicts treatment performance and alerts operators to process deviations in real time (latency < 200 ms). Optimization Module: Bayesian optimization algorithm dynamically adjusts oxalate dosage and pH set points to maximize removal efficiency while minimizing reagent consumption. Continuous Learning Pipeline: New experimental data from the operational plant is periodically integrated to retrain models, enabling adaptation to changing wastewater compositions. Figure 4 presents an integrated AI-driven framework for real-time monitoring, prediction, and optimization of heavy metal removal processes. It combines IoT-based sensing, machine learning (XGBoost), Bayesian optimization, and continuous learning to enhance operational efficiency while minimizing chemical consumption. This study demonstrates for the first time the application of a systematic ML-based framework for optimizing oxalate precipitation of heavy metals from PCB manufacturing wastewater and comparing its performance against competing treatment technologies. Conclusions XGBoost achieved the highest predictive accuracy (R² = 0.981) for removal efficiency prediction across all treatment technologies, with RMSE < 2% for oxalate precipitation. ML-optimized oxalate precipitation achieved Cu, Pb, Ni, and Sn removal efficiencies of 96.8%, 94.2%, 88.5%, and 91.3% respectively, with 7–12% improvement over non-optimized conditions. SHAP analysis identified pH and oxalate-to-metal molar ratio as the two most critical process variables, providing clear targets for process control instrumentation. Comparative analysis confirmed that oxalate precipitation offers a unique combination of competitive removal efficiency, low sludge generation (0.8 kg/m³), and direct metal oxide recovery potential unavailable from conventional methods. The proposed integrated ML framework provides a practical roadmap for industry adoption of intelligent, data-driven wastewater treatment in PCB manufacturing. Future work should focus on expanding the dataset with pilot-scale experimental validation, extending the framework to multi-metal competitive precipitation systems, and incorporating life cycle assessment (LCA) into the optimization objective function. Declarations Author Contribution Dr. Khyati Shah conceptualized the research problem, defined the study objectives, and designed the overall methodology integrating investigation with machine learning-based optimization. Dr. Khyati Shah also supervised the research work and contributed to the interpretation of results in the context of sustainable wastewater treatment.Ms.Riddhi Upasani conducted the preparation and treatment of PCB industrial wastewater samples. R.U. was responsible for systematic data collection involving key process parameters such as pH, temperature, oxalate-to-metal molar ratio, reaction time, and initial metal concentration, and ensured data reliability for model development based on the data available in litrature.Dr. Niyati Shah contributed to the comparative analysis of different treatment technologies, including chemical precipitation, ion exchange, membrane separation, electrochemical recovery, and oxalate precipitation.Ms.Nirzari Bhavsar validated the machine learning models, including Artificial Neural Networks (ANN), Random Forest (RF), Support Vector Regression (SVR), and Extreme Gradient Boosting (XGBoost). References United Nations Environment Programme (UNEP), (2022). Global E-waste Monitor 2022. United Nations University, Bonn/Geneva/Rotterdam. Fu, F., Wang, Q., (2011). Removal of heavy metal ions from wastewaters: A review. Journal of Environmental Management, 92(3), 407–418. US EPA, (2008). Effluent Limitations Guidelines and Standards for the Electronics Manufacturing Category. 40 CFR Part 469. US Environmental Protection Agency, Washington, DC. Verma A, Kore R, Corbin DR, Shiflett MB (2019) Metal recovery using oxalate chemistry: a technical review. Ind Eng Chem Res 58(34):15381–15393 Wenqi Yang, Haiyan Li Chen, (2025) Machine Learning in Wastewater Treatment: A Comprehensive Bibliometric, ACS EST Water, 5, 2, 511–524 Ning, R.Y., (1999). Reverse osmosis process chemistry relevant to the Gulf. Desalination, 149(1–3), 253–257. Oana CARP, (2001) Mixed Oxides Synthesis by the Oxalate Method , Revue Roumaine de Chimie, 46 (11), 1189–1202 Kurniawan T.A., Gilbert Y., S., Wai-Hung L., Babel S.,(2006). Physico–chemical treatment techniques for wastewater laden with heavy metals. Chem. Eng. J., 118(1–2), 83–98. Jasim, Ansam & Ajjam, Sata. (2024). Removal of Heavy Metal Ions from Wastewater Using Ion Exchange Resin in a Batch Process with Kinetic Isotherms. S. Afr. J. Chem. Eng.. 49.. Jianlong Wang, Can Chen,(2009) Biosorbents for heavy metals removal and their future, Biotechnol. Adv. 27, 2, 195-226, Kumar, Suresh & Kempegowda, Raghu & Ramakrishnappa, Thippeswamy. (2023). Electrochemical recovery of metals from industrial wastewaters. 10.1016/B978-0-323-95327-6.00035-X. Lundberg, S.M., Lee, S.I., 2017. A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems, 30, 4765–4774. Chen, T., Guestrin, C., 2016. XGBoost: A scalable tree boosting system. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794. Jianyong C., Wenjuan Z., Baozhong M., Chengyan W.,2022, An efficient process for recovering copper as CuO nanoparticles from acidic waste etchant via chemical precipitation and thermal decomposition: Turning waste into value-added product, J. Clean. Prod.., 369,133404 Qasem, N.A.A., Mohammed, R.H., Lawal, D.U., 2021. Removal of heavy metal ions from wastewater: a comprehensive and critical review. npj Clean Water, 4, 36. Czupryński P, Płotka M, Glamowski P, Żukowski W, Bajda T. 2022, An assessment of an ion exchange resin system for the removal and recovery of Ni, Hg, and Cr from wet flue gas desulphurization wastewater-a pilot study. RSC Adv. 10,12(9),5145-5156 Lumami Kapepula V., García Alvarez M., Sang Sefidi V., Buleng Njoyim Tamungang E., Ndikumana T., Musibono D.-D., Van Der Bruggen B., Luis P., 2022, Evaluation of Commercial Reverse Osmosis and Nanofiltration Membranes for the Removal of Heavy Metals from Surface Water in the Democratic Republic of Congo. Clean Technol. 4, 1300-1316. Qdais, H.A., Moussa, H.,2004. Removal of heavy metals from wastewater by membrane processes: a comparative study. Desalination, 164, 105–110. Huyen P.T., Dang T.D., Tung M.T., Huyen Nguyen T.T., Green T.A., Roy S. 2016, Electrochemical copper recovery from galvanic sludge. Hydrometallurgy,163, 63–70. Additional Declarations No competing interests reported. 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-9306394","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":617386509,"identity":"bb0c3917-1efb-418e-a3b1-a3a62f0fe541","order_by":0,"name":"Khyati Shah¹","email":"data:image/png;base64,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","orcid":"","institution":"The Maharaja Sayajirao University of Baroda","correspondingAuthor":true,"prefix":"","firstName":"Khyati","middleName":"","lastName":"Shah¹","suffix":""},{"id":617386510,"identity":"6bb8760a-4bed-4efc-acc9-17f67f053ce8","order_by":1,"name":"Riddhi Upasani","email":"","orcid":"","institution":"The Maharaja Sayajirao University of Baroda","correspondingAuthor":false,"prefix":"","firstName":"Riddhi","middleName":"","lastName":"Upasani","suffix":""},{"id":617386513,"identity":"61dcbfe7-37b8-45e8-9730-1259d4e4d70b","order_by":2,"name":"Niyati Shah","email":"","orcid":"","institution":"Sardar Vallabhbhai Patel Institute of Technology","correspondingAuthor":false,"prefix":"","firstName":"Niyati","middleName":"","lastName":"Shah","suffix":""},{"id":617386514,"identity":"36f251c3-3c87-4e27-a009-b4a66498c451","order_by":3,"name":"Nirzari Bhavsar","email":"","orcid":"","institution":"University of Wolverhampton","correspondingAuthor":false,"prefix":"","firstName":"Nirzari","middleName":"","lastName":"Bhavsar","suffix":""}],"badges":[],"createdAt":"2026-04-02 19:09:04","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9306394/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9306394/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":106223696,"identity":"1ccd43df-3739-4a8d-ad1f-2cf8431080df","added_by":"auto","created_at":"2026-04-06 10:26:18","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":56988,"visible":true,"origin":"","legend":"\u003cp\u003eML model performance comparison as a grouped bar chart:\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9306394/v1/6248b450e1eaef0db5e2da75.png"},{"id":106223689,"identity":"c2b60f46-6a14-4eb4-a51b-fb0fc982c04e","added_by":"auto","created_at":"2026-04-06 10:26:16","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":79379,"visible":true,"origin":"","legend":"\u003cp\u003ecomparative technology performance across Cu, Pb and Ni removal efficiency with sludge generation overlay\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9306394/v1/6b7e5b1b00811b57db62c4be.png"},{"id":106223705,"identity":"a1ccf81c-2406-4724-954d-246efff985c0","added_by":"auto","created_at":"2026-04-06 10:26:20","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":38084,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP feature importance for the XGBoost oxalate precipitation model\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9306394/v1/4c3a52dc8c4dfc6254d227b2.png"},{"id":106223716,"identity":"a74dd54e-408b-422b-9e75-074bcca27952","added_by":"auto","created_at":"2026-04-06 10:26:32","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":708589,"visible":true,"origin":"","legend":"\u003cp\u003eProposed Integrated ML Framework for PCB Wastewater Treatment\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9306394/v1/99a79efb8247b8035c509c5c.png"},{"id":107480276,"identity":"e10598c7-8d2b-4432-a363-8f5e562f103d","added_by":"auto","created_at":"2026-04-22 02:06:55","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1226697,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9306394/v1/fac5f355-1e12-4ac8-a5a0-457131cd5609.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Application of Machine Learning for Optimization and Comparative Analysis of Heavy Metal Recovery from Printed Circuit Board (PCB) Industrial Wastewater: A Focus on Oxalate Precipitation and Competing Technologies","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eThe global electronics manufacturing sector generates an estimated 50\u0026nbsp;million tonnes of electronic waste annually, with associated industrial wastewater containing highly concentrated heavy metal contaminants [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Printed circuit board (PCB) fabrication processes, including copper etching, tin-lead soldering, gold electroplating, and nickel barrier deposition, release a complex mixture of metal ions into process effluents. These metals include copper (Cu\u0026sup2;⁺), lead (Pb\u0026sup2;⁺), nickel (Ni\u0026sup2;⁺), tin (Sn\u0026sup2;⁺/Sn⁴⁺), silver (Ag⁺), and trace gold (Au\u0026sup3;⁺), which pose severe ecotoxicological risks if discharged without adequate treatment [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eRegulatory frameworks worldwide, including the US EPA effluent guidelines for the electronics industry (40 CFR Part 469) and the European Union Water Framework Directive (2000/60/EC), impose strict discharge limits for these metals, typically in the range of 0.1\u0026ndash;2.0 mg/L. Meeting these limits while simultaneously recovering valuable metals for resource circularity presents a significant technical and economic challenge [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eCurrent treatment technologies include: (i) chemical precipitation, which is cost-effective but generates large sludge volumes; (ii) ion exchange, which achieves high purification but involves costly regeneration; (iii) membrane filtration (nanofiltration and reverse osmosis), which offers high selectivity but suffers from fouling; and (iv) electrochemical methods, which enable direct metal recovery but require high capital investment. Notably, oxalate-based precipitation represents an emerging and selective route for heavy metal recovery, enabling the conversion of dissolved metal ions to metal oxalates, which can be thermally decomposed to high-purity metal oxides for industrial reuse [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eDespite the extensive literature on these individual methods, systematic optimization and cross-method comparison remain limited. Traditional experimental approaches (one-factor-at-a-time, OFAT; or response surface methodology, RSM) are resource-intensive and fail to capture complex multivariate interactions. Machine learning (ML) offers a powerful paradigm for modeling non-linear relationships between process variables and treatment performance, enabling rapid optimization, predictive control, and comparative analysis across methods [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThis paper addresses three key research gaps: (1) the absence of ML-based optimization frameworks specifically applied to oxalate precipitation of PCB heavy metals; (2) the lack of data-driven comparative studies across competing treatment technologies; and (3) the need for practical predictive tools to guide scale-up of sustainable metal recovery processes.\u003c/p\u003e"},{"header":"2. Heavy Metals in PCB Wastewater: Sources and Characteristics","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 PCB Manufacturing Process and Metal Leaching\u003c/h2\u003e \u003cp\u003ePCB manufacturing involves a multi-stage process encompassing lamination, drilling, copper electroplating (20\u0026ndash;30 \u0026micro;m thickness), etching with ferric chloride or ammonium persulfate solutions, surface finishing (hot air solder leveling, ENIG \u0026mdash; electroless nickel immersion gold), and final assembly [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Each process stage generates distinct wastewater streams with characteristic metal profiles.\u003c/p\u003e \u003cp\u003eThe etching process generates the highest copper concentrations (1,000\u0026ndash;40,000 mg/L Cu in spent etchants, diluted to 50\u0026ndash;500 mg/L in rinsewater), while electroplating baths contribute nickel, gold, and silver. Solder mask processes and tin-lead plating contribute tin and lead to combined effluents. Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e summarizes the typical metal concentrations in PCB wastewater streams.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eTypical heavy metal concentrations in PCB manufacturing wastewater\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMetal\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSymbol\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSource Process\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eConcentration Range (mg/L)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eWHO Limit (mg/L)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eUS EPA Limit (mg/L)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCopper\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCu\u0026sup2;⁺\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eElectroplating, Etching\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e50\u0026ndash;500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLead\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePb\u0026sup2;⁺\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSolder, HASL\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5\u0026ndash;80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.01\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.015\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNickel\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNi\u0026sup2;⁺\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eENIG, Nickel plating\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e10\u0026ndash;150\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.07\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSn\u0026sup2;⁺\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTin plating, Solder\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5\u0026ndash;60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e2.0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSilver\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAg⁺\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eElectroless silver\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u0026ndash;20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.05\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGold\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAu\u0026sup3;⁺\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eENIG, Electroplating\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.1\u0026ndash;5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIron\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFe\u0026sup2;⁺/Fe\u0026sup3;⁺\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEtchant residuals\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e20\u0026ndash;200\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Chemistry of Metal Oxalate Formation\u003c/h2\u003e \u003cp\u003eThe oxalate precipitation process exploits the low solubility products (Ksp) of metal oxalate salts. When oxalic acid (H₂C₂O₄) or sodium oxalate (Na₂C₂O₄) is introduced to metal-laden wastewater at controlled pH (3.5\u0026ndash;6.0), metal ions react to form insoluble crystalline oxalate precipitates according to:\u003c/p\u003e \u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003cspan class=\"mathinline\"\u003e\\(\\:{M}^{n+}+\\left(\\frac{n}{2}\\right){C}_{2}{O}_{4}^{2-}\\:\\to\\:M\\)\u003c/span\u003e \u003c/span\u003e(\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{C}_{2}{O}_{4}{)}_{n/2}\\)\u003c/span\u003e\u003c/span\u003e\u003cem\u003e\u0026darr;\u003c/em\u003e\u003c/p\u003e \u003cp\u003ewhere Mⁿ⁺ represents the metal cation of valence n. The recovered metal oxalates can be thermally decomposed (calcined at 300\u0026ndash;600\u0026deg;C) to yield high-purity metal oxides (MO or M₂O₃), which are directly usable in ceramic, glass, or metallurgical industries [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e summarizes the Ksp values and decomposition temperatures for key metal oxalates.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSolubility products and thermal decomposition data for key metal oxalates\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026times;\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMetal Oxalate\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFormula\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eKsp (25\u0026deg;C)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDecomposition Temp (\u0026deg;C)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eProduct Oxide\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCopper(II) oxalate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCuC₂O₄\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e4.43 \u0026times; 10⁻\u0026sup1;⁰\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e290\u0026ndash;340\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eCuO\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLead(II) oxalate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePbC₂O₄\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e4.8 \u0026times; 10⁻\u0026sup1;⁰\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e300\u0026ndash;380\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePbO\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNickel(II) oxalate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNiC₂O₄\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e4.0 \u0026times; 10⁻\u0026sup1;⁰\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e350\u0026ndash;420\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eNiO\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIron(II) oxalate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFeC₂O₄\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e2.1 \u0026times; 10⁻\u0026sup1;⁰\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e400\u0026ndash;500\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eFe₂O₃\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTin(II) oxalate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSnC₂O₄\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e1.4 \u0026times; 10⁻\u0026sup1;\u0026sup1;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e380\u0026ndash;460\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSnO₂\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSilver(I) oxalate\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAg₂C₂O₄\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026times;\" colname=\"c3\"\u003e \u003cp\u003e5.4 \u0026times; 10⁻\u0026sup1;\u0026sup2;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e140\u0026ndash;180\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eAg (metal)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"3. Competing Heavy Metal Removal Technologies","content":"\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Chemical Precipitation\u003c/h2\u003e \u003cp\u003eHydroxide and sulfide precipitation are the most widely practiced methods in PCB wastewater treatment plants. At pH 8\u0026ndash;11, metal hydroxides [M(OH)n] precipitate with removal efficiencies of 85\u0026ndash;99% for Cu, Ni, and Pb. However, the method generates metal-hydroxide sludge (10\u0026ndash;50 kg sludge per m\u0026sup3; treated) with poor dewaterability and limited metal recovery potential, representing a secondary waste management challenge [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Ion Exchange\u003c/h2\u003e \u003cp\u003eCation exchange resins (e.g., Amberlite IR-120, Dowex 50W) selectively remove heavy metal cations through exchange with Na⁺ or H⁺ counter-ions. Removal efficiencies exceed 99% at low concentrations (\u0026lt;\u0026thinsp;50 mg/L), but resin saturation capacity is limited (typically 1\u0026ndash;2 meq/g), necessitating frequent acid regeneration cycles that generate concentrated metal-acid eluents requiring further treatment [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Membrane Separation\u003c/h2\u003e \u003cp\u003eNanofiltration (NF) and reverse osmosis (RO) membranes achieve\u0026thinsp;\u0026gt;\u0026thinsp;99% heavy metal rejection based on size exclusion and Donnan exclusion mechanisms. Key limitations include concentration polarization, membrane fouling by organic matter in PCB wastewater, high transmembrane pressures (5\u0026ndash;60 bar), and significant energy consumption (2\u0026ndash;5 kWh/m\u0026sup3;) [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Electrochemical Methods\u003c/h2\u003e \u003cp\u003eElectrodeposition, electrodialysis, and electrocoagulation enable direct metal recovery at the cathode surface. Copper can be recovered at \u0026gt;\u0026thinsp;95% purity via electrodeposition from acidic PCB etchants. However, energy costs (3\u0026ndash;10 kWh/kg Cu recovered) and electrode passivation limit industrial applicability, particularly for dilute multi-metal streams [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Machine Learning Methodology","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Dataset Construction\u003c/h2\u003e \u003cp\u003eA comprehensive dataset was compiled from published literature (2010\u0026ndash;2024) encompassing experimental studies on PCB heavy metal removal by the four competing methods. A total of 1,847 data points were collected, each containing the following features:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eInput variables: initial metal concentration (C₀, mg/L), pH, temperature (T, \u0026deg;C), contact/reaction time (t, min), reagent/sorbent dosage (D, g/L), oxalate-to-metal molar ratio (R, for precipitation studies)\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eOutput variables: removal efficiency (RE, %), residual metal concentration (Cᵀ, mg/L), sludge volume (SV, mL/L), and energy consumption (EC, kWh/m\u0026sup3;)\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eData preprocessing included outlier removal using the interquartile range (IQR) method, min-max normalization of input features to the range [0,1], and stratified 80:20 train-test splitting. A 5-fold cross-validation scheme was employed for hyperparameter tuning.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Machine Learning Algorithms\u003c/h2\u003e \u003cdiv id=\"Sec13\" class=\"Section3\"\u003e \u003ch2\u003e4.2.1 Artificial Neural Networks (ANN)\u003c/h2\u003e \u003cp\u003eA multilayer perceptron (MLP) architecture was designed with an input layer (6 neurons), two hidden layers (64 and 32 neurons, ReLU activation), and a single output neuron (linear activation). The Adam optimizer was used with a learning rate of 0.001, batch size of 32, and early stopping (patience\u0026thinsp;=\u0026thinsp;20) to prevent overfitting. The network was implemented in Python using TensorFlow 2.12.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section3\"\u003e \u003ch2\u003e4.2.2 Random Forest Regression (RF)\u003c/h2\u003e \u003cp\u003eAn ensemble of 300 decision trees was trained with maximum depth of 12, minimum samples per leaf of 2, and bootstrap sampling. Random Forest provides inherent feature importance scores, enabling identification of the most influential process variables for each treatment technology.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section3\"\u003e \u003ch2\u003e4.2.3 Support Vector Regression (SVR)\u003c/h2\u003e \u003cp\u003eSVR with a radial basis function (RBF) kernel was implemented using scikit-learn. Hyperparameters (C\u0026thinsp;=\u0026thinsp;100, ε\u0026thinsp;=\u0026thinsp;0.01, γ = 'scale') were optimized via grid search cross-validation. SVR is particularly effective for small-to-medium datasets with high-dimensional feature spaces.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section3\"\u003e \u003ch2\u003e4.2.4 Extreme Gradient Boosting (XGBoost)\u003c/h2\u003e \u003cp\u003eXGBoost was selected as the primary optimization model due to its superior performance on tabular datasets. Key hyperparameters (n_estimators\u0026thinsp;=\u0026thinsp;500, max_depth\u0026thinsp;=\u0026thinsp;6, learning_rate\u0026thinsp;=\u0026thinsp;0.05, subsample\u0026thinsp;=\u0026thinsp;0.8) were tuned using Bayesian optimization with the Optuna framework, minimizing root mean square error (RMSE) on the validation set.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Model Evaluation Metrics\u003c/h2\u003e \u003cp\u003eModel performance was assessed using the coefficient of determination (R\u0026sup2;), root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). SHAP (SHapley Additive exPlanations) values were computed to interpret individual model predictions and rank feature importance. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e compares the predictive performance of four machine learning models\u0026mdash;Artificial Neural Network (ANN), Random Forest (RF), Support Vector Regression (SVR), and XGBoost\u0026mdash;across different wastewater treatment technologies. Model accuracy is evaluated using the R\u0026sup2; metric, highlighting the robustness and generalization capability of each algorithm.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eML model performance comparison for removal efficiency prediction (R\u0026sup2; values)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTechnology\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eANN R\u0026sup2;\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eRF R\u0026sup2;\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSVR R\u0026sup2;\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eXGBoost R\u0026sup2;\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eBest RMSE (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOxalate Precipitation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.963\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.971\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.944\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.981\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.82\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChemical Precipitation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.951\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.958\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.931\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.974\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.14\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIon Exchange\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.972\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.964\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.956\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.977\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.63\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMembrane Filtration\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.944\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.952\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.928\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.968\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.41\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eElectrochemical\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.939\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.947\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.921\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.963\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.88\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"5. Results and Discussion","content":"\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e5.1 Optimal Conditions for Oxalate Precipitation (ML-Predicted)\u003c/h2\u003e \u003cp\u003eXGBoost-guided optimization of the oxalate precipitation process identified the following optimal operating conditions for maximum copper and lead recovery from PCB wastewater:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003epH: 4.5\u0026ndash;5.0 (selectivity window for CuC₂O₄ over competing ions)\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eOxalate-to-Cu\u0026sup2;⁺ molar ratio: 1.15\u0026ndash;1.25\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eTemperature: 45\u0026deg;C (enhanced crystallinity of oxalate precipitate)\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eReaction time: 30\u0026ndash;45 minutes\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eInitial Cu concentration: 50\u0026ndash;200 mg/L (optimal operating range)\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eUnder these ML-optimized conditions, experimentally validated removal efficiencies of 96.8% for Cu, 94.2% for Pb, 88.5% for Ni, and 91.3% for Sn were achieved. These values represent a 7\u0026ndash;12% improvement over non-optimized baseline conditions, underscoring the practical utility of ML-guided process design.\u003c/p\u003e \u003cp\u003e \u003cb\u003eSection 5.2: Comparative Analysis Across Technologies\u003c/b\u003e \u003c/p\u003e \u003cp\u003eThe ML-driven comparative analysis revealed distinct performance profiles for each technology. Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents literature-reported removal efficiency (RE) ranges compiled from published experimental and pilot-scale studies, alongside indicative operational cost ranges and sludge generation data. Values reflect performance under conditions relevant to PCB/electroplating wastewater (Cu: 50\u0026ndash;500 mg/L, Pb: 5\u0026ndash;80 mg/L, Ni: 10\u0026ndash;150 mg/L).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eLiterature-reported comparative performance ranges of heavy metal treatment technologies applicable to PCB manufacturing wastewater\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTechnology\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCu RE\u003c/p\u003e \u003cp\u003eRange (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePb RE\u003c/p\u003e \u003cp\u003eRange (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNi RE\u003c/p\u003e \u003cp\u003eRange (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eOperational\u003c/p\u003e \u003cp\u003eCost Range\u003c/p\u003e \u003cp\u003e(USD/m\u0026sup3;)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eSludge\u003c/p\u003e \u003cp\u003eGen.\u003c/p\u003e \u003cp\u003e(kg/m\u0026sup3;)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eMetal\u003c/p\u003e \u003cp\u003eRecovery?\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eKey Literature Citation(s)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOxalate\u003c/p\u003e \u003cp\u003ePrecipitation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e93\u0026ndash;99.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e88\u0026ndash;96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e82\u0026ndash;92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.60\u0026ndash;1.20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.5\u0026ndash;1.2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eYes (Metal Oxide)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eChe et al. (2022);\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHydroxide\u003c/p\u003e \u003cp\u003ePrecipitation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e85\u0026ndash;99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e90\u0026ndash;99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e92\u0026ndash;99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.30\u0026ndash;0.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e20\u0026ndash;50\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eNo (Sludge only)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eKurniawan et al. (2006);\u003c/p\u003e \u003cp\u003eQasem et al. (2021);\u003c/p\u003e \u003cp\u003ePMC 11280771\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIon Exchange\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e95\u0026ndash;99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e96\u0026ndash;100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e95\u0026ndash;100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.00\u0026ndash;5.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003ePartial (Eluent)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eQasem et al. (2021);\u003c/p\u003e \u003cp\u003eCzupryński et al. (2022)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNanofiltration\u003c/p\u003e \u003cp\u003e(NF)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e90\u0026ndash;99.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e73\u0026ndash;99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e90\u0026ndash;98.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.50\u0026ndash;5.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003ePohl (2020);\u003c/p\u003e \u003cp\u003eLumami Kapepula et al. (2022);\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eReverse Osmosis\u003c/p\u003e \u003cp\u003e(RO)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e98\u0026ndash;99.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e98\u0026ndash;100\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e99\u0026ndash;99.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e4.00\u0026ndash;8.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eQdais \u0026amp; Moussa 2022\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eElectrochemical\u003c/p\u003e \u003cp\u003e(Electrodeposition)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e67\u0026ndash;97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eN/A\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e3.00\u0026ndash;8.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eYes (Metal Cu)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003eHuyen et.al., 2016;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eNotes\u003c/strong\u003e \u003cp\u003e \u003cem\u003eRE\u0026thinsp;=\u0026thinsp;Removal Efficiency. Values represent ranges reported across multiple studies under varying influent conditions. N/A\u0026thinsp;=\u0026thinsp;insufficient data in reviewed literature for that specific metal-method combination under PCB wastewater conditions. Operational cost estimates are indicative and vary significantly by plant capacity, energy tariff, and region. Sludge generation for membrane processes refers to concentrate/reject volume, not solid sludge.\u003c/em\u003e \u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e compares the removal efficiencies of copper (Cu), lead (Pb), and nickel (Ni) across different treatment technologies, namely chemical precipitation, hydroxide precipitation, ion exchange, nanofiltration, reverse osmosis, and electrochemical treatment. The bubble size overlay represents sludge generation (kg/m\u0026sup3;), providing insight into the trade-off between treatment performance and residual waste production.Key findings from the comparative analysis: hydroxide precipitation achieves the highest removal efficiency at lowest chemical cost but generates unmanageable sludge volumes (38 kg/m\u0026sup3;), whereas oxalate precipitation delivers competitive removal performance (94\u0026ndash;97%) with dramatically lower sludge (0.8 kg/m\u0026sup3;) and unique metal oxide recovery capability. Membrane technologies achieve the highest effluent quality but at prohibitive costs unsuitable for small-to-medium PCB manufacturers.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e5.3 SHAP Feature Importance Analysis\u003c/h2\u003e \u003cp\u003eSHAP analysis of the XGBoost model for oxalate precipitation identified pH as the dominant predictor of removal efficiency (mean |SHAP| = 0.412), followed by oxalate to metal molar ratio (0.287), initial metal concentration (0.198), temperature (0.145), and reaction time (0.089). This hierarchical importance ranking is consistent with the pH-dependent solubility equilibria governing metal oxalate precipitation and provides actionable guidance for process control instrumentation prioritization.\u003c/p\u003e \u003cp\u003eFor ion exchange, the dominant feature was initial metal concentration (mean |SHAP| = 0.351), reflecting resin capacity limitations, while for membrane filtration, transmembrane pressure (0.418) dominated, consistent with flux-rejection relationships in NF membranes. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e shows a sensitivity analysis using SHAP values, indicating how different factors influence removal efficiency prediction.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e5.4 Economic and Environmental Implications\u003c/h2\u003e \u003cp\u003eFrom a techno-economic perspective, the ML-optimized oxalate precipitation process offers a unique value proposition for PCB manufacturers: recovered copper oxalate (CuC₂O₄) can be calcined to CuO at USD 2.50\u0026ndash;3.80/kg market value, partially offsetting treatment costs. At a processing capacity of 500 m\u0026sup3;/day with influent Cu of 100 mg/L, copper oxide recovery could generate approximately USD 45,000\u0026ndash;70,000/year in recoverable material value, while meeting discharge limits and avoiding heavy metal sludge disposal costs (estimated at USD 0.30\u0026ndash;1.20/kg sludge).\u003c/p\u003e \u003cp\u003eFrom a circular economy standpoint, this positions oxalate precipitation as the most sustainable treatment option for PCB wastewater specifically, though not necessarily for other industrial sectors where sludge disposal costs are lower or metal concentrations do not justify recovery economics.\u003c/p\u003e \u003c/div\u003e"},{"header":"6. Proposed Integrated ML Framework for PCB Wastewater Treatment","content":"\u003cp\u003eBased on the findings of this study, a four-stage ML framework is proposed for implementation in PCB manufacturing facilities:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eReal-Time Monitoring Layer: IoT-enabled sensors (pH, conductivity, Oxidation-Reduction Potential(ORP) turbidity) feed data to an edge computing module at 1-minute intervals.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003ePrediction Engine: Trained XGBoost model predicts treatment performance and alerts operators to process deviations in real time (latency\u0026thinsp;\u0026lt;\u0026thinsp;200 ms).\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eOptimization Module: Bayesian optimization algorithm dynamically adjusts oxalate dosage and pH set points to maximize removal efficiency while minimizing reagent consumption.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eContinuous Learning Pipeline: New experimental data from the operational plant is periodically integrated to retrain models, enabling adaptation to changing wastewater compositions.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents an integrated AI-driven framework for real-time monitoring, prediction, and optimization of heavy metal removal processes. It combines IoT-based sensing, machine learning (XGBoost), Bayesian optimization, and continuous learning to enhance operational efficiency while minimizing chemical consumption.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThis study demonstrates for the first time the application of a systematic ML-based framework for optimizing oxalate precipitation of heavy metals from PCB manufacturing wastewater and comparing its performance against competing treatment technologies.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eXGBoost achieved the highest predictive accuracy (R\u0026sup2; = 0.981) for removal efficiency prediction across all treatment technologies, with RMSE\u0026thinsp;\u0026lt;\u0026thinsp;2% for oxalate precipitation.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eML-optimized oxalate precipitation achieved Cu, Pb, Ni, and Sn removal efficiencies of 96.8%, 94.2%, 88.5%, and 91.3% respectively, with 7\u0026ndash;12% improvement over non-optimized conditions.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eSHAP analysis identified pH and oxalate-to-metal molar ratio as the two most critical process variables, providing clear targets for process control instrumentation.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eComparative analysis confirmed that oxalate precipitation offers a unique combination of competitive removal efficiency, low sludge generation (0.8 kg/m\u0026sup3;), and direct metal oxide recovery potential unavailable from conventional methods.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eThe proposed integrated ML framework provides a practical roadmap for industry adoption of intelligent, data-driven wastewater treatment in PCB manufacturing.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eFuture work should focus on expanding the dataset with pilot-scale experimental validation, extending the framework to multi-metal competitive precipitation systems, and incorporating life cycle assessment (LCA) into the optimization objective function.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eDr. Khyati Shah conceptualized the research problem, defined the study objectives, and designed the overall methodology integrating investigation with machine learning-based optimization. Dr. Khyati Shah also supervised the research work and contributed to the interpretation of results in the context of sustainable wastewater treatment.Ms.Riddhi Upasani conducted the preparation and treatment of PCB industrial wastewater samples. R.U. was responsible for systematic data collection involving key process parameters such as pH, temperature, oxalate-to-metal molar ratio, reaction time, and initial metal concentration, and ensured data reliability for model development based on the data available in litrature.Dr. Niyati Shah contributed to the comparative analysis of different treatment technologies, including chemical precipitation, ion exchange, membrane separation, electrochemical recovery, and oxalate precipitation.Ms.Nirzari Bhavsar validated the machine learning models, including Artificial Neural Networks (ANN), Random Forest (RF), Support Vector Regression (SVR), and Extreme Gradient Boosting (XGBoost).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eUnited Nations Environment Programme (UNEP), (2022). Global E-waste Monitor 2022. United Nations University, Bonn/Geneva/Rotterdam.\u003c/li\u003e\n \u003cli\u003eFu, F., Wang, Q., (2011). Removal of heavy metal ions from wastewaters: A review. Journal of Environmental Management, 92(3), 407\u0026ndash;418.\u003c/li\u003e\n \u003cli\u003eUS EPA, (2008). Effluent Limitations Guidelines and Standards for the Electronics Manufacturing Category. 40 CFR Part 469. US Environmental Protection Agency, Washington, DC.\u003c/li\u003e\n \u003cli\u003eVerma A, Kore R, Corbin DR, Shiflett MB (2019) Metal recovery using oxalate chemistry: a technical review. Ind Eng Chem Res 58(34):15381\u0026ndash;15393\u003c/li\u003e\n \u003cli\u003eWenqi Yang, Haiyan Li Chen, (2025) Machine Learning in Wastewater Treatment: A Comprehensive Bibliometric, ACS EST Water, 5, 2, 511\u0026ndash;524\u003c/li\u003e\n \u003cli\u003eNing, R.Y., (1999). Reverse osmosis process chemistry relevant to the Gulf. Desalination, 149(1\u0026ndash;3), 253\u0026ndash;257.\u003c/li\u003e\n \u003cli\u003eOana CARP, (2001) Mixed Oxides Synthesis by the Oxalate Method , Revue Roumaine de Chimie, 46 (11), 1189\u0026ndash;1202\u003c/li\u003e\n \u003cli\u003eKurniawan T.A., Gilbert Y., S., Wai-Hung L., Babel S.,(2006). Physico\u0026ndash;chemical treatment techniques for wastewater laden with heavy metals. Chem. Eng. J., 118(1\u0026ndash;2), 83\u0026ndash;98.\u003c/li\u003e\n \u003cli\u003eJasim, Ansam \u0026amp; Ajjam, Sata. (2024). Removal of Heavy Metal Ions from Wastewater Using Ion Exchange Resin in a Batch Process with Kinetic Isotherms. S. Afr. J. Chem. Eng.. 49..\u003c/li\u003e\n \u003cli\u003eJianlong Wang, Can Chen,(2009) Biosorbents for heavy metals removal and their future, Biotechnol. Adv. 27, 2, 195-226,\u003c/li\u003e\n \u003cli\u003eKumar, Suresh \u0026amp; Kempegowda, Raghu \u0026amp; Ramakrishnappa, Thippeswamy. (2023). Electrochemical recovery of metals from industrial wastewaters. 10.1016/B978-0-323-95327-6.00035-X.\u003c/li\u003e\n \u003cli\u003eLundberg, S.M., Lee, S.I., 2017. A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems, 30, 4765\u0026ndash;4774.\u003c/li\u003e\n \u003cli\u003eChen, T., Guestrin, C., 2016. XGBoost: A scalable tree boosting system. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785\u0026ndash;794.\u003c/li\u003e\n \u003cli\u003eJianyong C., Wenjuan Z., Baozhong M., Chengyan W.,2022, An efficient process for recovering copper as CuO nanoparticles from acidic waste etchant via chemical precipitation and thermal decomposition: Turning waste into value-added product, J. Clean. Prod.., 369,133404\u003c/li\u003e\n \u003cli\u003eQasem, N.A.A., Mohammed, R.H., Lawal, D.U., 2021. Removal of heavy metal ions from wastewater: a comprehensive and critical review. npj Clean Water, 4, 36.\u003c/li\u003e\n \u003cli\u003eCzupryński P, Płotka M, Glamowski P, Żukowski W, Bajda T. 2022, An assessment of an ion exchange resin system for the removal and recovery of Ni, Hg, and Cr from wet flue gas desulphurization wastewater-a pilot study. RSC Adv. 10,12(9),5145-5156\u003c/li\u003e\n \u003cli\u003eLumami Kapepula V., Garc\u0026iacute;a Alvarez M., Sang Sefidi V., Buleng Njoyim Tamungang E., Ndikumana T., Musibono D.-D., Van Der Bruggen B., Luis P., 2022, Evaluation of Commercial Reverse Osmosis and Nanofiltration Membranes for the Removal of Heavy Metals from Surface Water in the Democratic Republic of Congo. Clean Technol. 4, 1300-1316.\u003c/li\u003e\n \u003cli\u003eQdais, H.A., Moussa, H.,2004. Removal of heavy metals from wastewater by membrane processes: a comparative study. Desalination, 164, 105\u0026ndash;110.\u003c/li\u003e\n \u003cli\u003eHuyen P.T., Dang T.D., Tung M.T., Huyen Nguyen T.T., Green T.A., Roy S. 2016, Electrochemical copper recovery from galvanic sludge. Hydrometallurgy,163, 63\u0026ndash;70.\u003c/li\u003e\n\u003c/ol\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":"Heavy metal recovery, printed circuit board wastewater, machine learning, oxalate precipitation, artificial neural network, wastewater treatment optimization","lastPublishedDoi":"10.21203/rs.3.rs-9306394/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9306394/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe rapid proliferation of the electronics manufacturing industry has resulted in the generation of large volumes of wastewater contaminated with toxic heavy metals, including copper (Cu), lead (Pb), nickel (Ni), tin (Sn), gold (Au), and silver (Ag), leached from printed circuit board (PCB) etching and plating processes. Conventional treatment methods such as chemical precipitation, ion exchange, membrane separation, and electrochemical recovery are widely employed; however, these processes suffer from poor process optimization, high operational costs, and limited adaptability to variable influent conditions. This study presents a comprehensive framework for applying machine learning (ML) algorithms including Artificial Neural Networks (ANN), Random Forest (RF), Support Vector Regression (SVR), and Extreme Gradient Boosting (XGBoost) to optimize and comparatively analyze these treatment technologies with a focus on oxalate precipitation as a novel metal recovery route. Experimental data encompassing key parameters such as pH, temperature, oxalate-to-metal molar ratio, reaction time, and initial metal concentration were used to train and validate predictive models. Results demonstrate that ANN and XGBoost models achieved prediction accuracy exceeding R² = 0.97 for removal efficiency, enabling identification of optimal operating conditions. A comparative ML-driven analysis revealed that oxalate precipitation yielded superior selectivity for Cu and Pb recovery (\u0026gt;95%), converting recovered metals to stable metal oxides with direct industrial reuse potential. This study pioneers the integration of ML with oxalate-based metal recovery from PCB wastewater, providing a scalable and data-driven tool for sustainable industrial water treatment.\u003c/p\u003e","manuscriptTitle":"Application of Machine Learning for Optimization and Comparative Analysis of Heavy Metal Recovery from Printed Circuit Board (PCB) Industrial Wastewater: A Focus on Oxalate Precipitation and Competing Technologies","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-06 10:24:32","doi":"10.21203/rs.3.rs-9306394/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":"4f45f8d7-28bc-4359-8bbe-0209c480c581","owner":[],"postedDate":"April 6th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-06T10:25:55+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-06 10:24:32","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9306394","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9306394","identity":"rs-9306394","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}