In response to the difficulties and poor timeliness in detecting feeding metallic foreign objects during high-yield continuous crushing operations in coal mines, this paper proposes a new method for detecting metallic foreign objects, combining pulsed eddy current testing with the Truncated Region Eigenfunction Expansion (TREE) method. This method is suitable for the harsh working conditions in coal mine crushing stations, which include high dust, strong vibration, strong electromagnetic interference, and low temperatures in winter. A model of the eddy current field of feeding metallic foreign objects in the truncated region is established using a coaxial excitation and receiving coil with a Hall sensor. The full-cycle time-domain analytical solution for the induced voltage and magnetic induction intensity of the reflective field under practical square wave signals is obtained. Simulation and experimental results show that the effective time range, peak value, and time to peak of the received voltage and magnetic induction signals can be used to classify and identify the size, thickness, conductivity, and magnetic permeability of feeding metallic foreign objects. Experimental results meet the actual needs for removing feeding metallic foreign objects in coal mine sites. This provides core technical support for the establishment of a predictive fault diagnosis system for crushing equipment.

During the high-yield continuous crushing operations in coal mines, the detection of feeding metallic foreign objects (MFOs) faces challenges such as complex noise interference and varying lift-off distances. This paper proposes a pulsed eddy current (PEC) data-driven interpretable neural network (PECDINet) based on optimized variational mode decomposition and improved wavelet threshold denoising for multi-feature classification and identification of MFOs. First, for industrial dataset collected at the crushing station, the variational mode decomposition is optimized using snow geese algorithm and envelope entropy function that incorporates environmental noise. Improved adaptive threshold function is then applied for wavelet threshold denoising of the intrinsic mode functions, effectively filtering out complex noise while preserving key characteristics of PEC signals corresponding to the time-domain solutions of PEC reflection fields in truncated regions of MFOs. Then, PECDINet with subtraction, square difference, and average layers is developed to extract start and end times, and time to peak of experimental signals, enabling characteristics identification of conductivity, thickness, and vertical width of MFOs. Finally, experiments demonstrate the proposed method’s advantages in accuracy and efficiency, addressing the challenges of multi-feature classification and identification of MFOs in coal mine field environments.