A Sensor-Based Magnetite Ore Sorting System Integrating Empirical Mode Decomposition and Convolutional Neural Network

To address the challenge of poor separation performance exhibited by conventional magnetic separation equipment when processing coarse-grained, low-grade magnetite ore, this paper proposes a novel ore recognition method that integrates empirical mode decomposition (EMD) with a convolutional neural network (CNN). First, the normalized magnetic induction intensity signals are decomposed using EMD to yield a series of intrinsic mode functions (IMFs). IMFs containing prominent characteristic information are then selected and fused based on their dominant frequency and kurtosis values, resulting in a reconstructed signal with significantly reduced noise. Subsequently, the reconstructed signals undergo inversion, re-normalization, and dimensional transformation into a two-dimensional matrix format to construct the training and testing sample datasets. A convolutional neural network is then designed and optimized to automatically extract discriminative features from these preprocessed samples, enabling accurate classification of magnetite ore grades. Experimental results demonstrate that the proposed EMD-CNN framework achieves effective and stable classification performance across different ore grades. In particular, the application of EMD for noise component removal substantially enhances the CNN’s recognition accuracy for waste rock and medium-grade ore, which are traditionally the most difficult categories to distinguish.