Research on verification of the MLP network-based current sensor fault classifier for various vector-controlled AC motor drives.

The research focused on classifying fault types in stator current sensors in a field-oriented control (FOC) structure for drive systems with permanent magnet motors (PMSM) and induction motors. A unique method for extracting fault symptoms from the stator current signal of AC motors (induction and PMSM) was developed, allowing the use of a small multilayer perceptron network architecture as a classifier. The proposed solution was validated through simulations in the MATLAB/Simulink environment and on laboratory benches.Uploaded data contain respectively:Data_1The file "sim_data.m" contains the results of stator current sensor fault classifier simulation tests conducted for a PMSM motor drive. The following fault classes were considered:                             OC - complete loss of signal (open circuit),                             G - change of gain,                             OFF - appearance of a constant component (offset),                             SAT - saturation of the signal.                            The file "sim_data.m" contains an array of "s" structures. Each structure in this array corresponds to a different value of load torque. This value can be checked in the field named "load_torque". The other fields are as follows:              t - simulation time              CSFCA, CSFCB - results of fault classification of currents in phase A and B              isA, isB - measured currents              isA_VCS, isB_VCS - estimated currents              zero_cross_timesA, zero_cross_timesB - times at which the passage of current through zero occurs              wm_ref - speed set in the control structure                            File "confusion_matrix.m" calculates confusion matrix for data from file "sim_data.m" Data_2Experimental results of fault classification of induction motor current sensors. The following classes of faults were considered: OC - open circuit,                     G - gain, OFF - offset, SAT - saturation.              The signals in the files are as follows:                             CSFCA, CSFCB - results of fault classification of currents in phase A and B                             CSFCA_without_NF, CSFCB_without_NF - results of fault classification of currents in phase A and B with NF class omitted                             CSFCA_plot, CSFCB_plot - classification data used for graphs                             isA, isB - measured currents                             isA_VCS, isB_VCS - estimated currents                             zero_cross_timesA, zero_cross_timesB - moments at which the passage of current through zero occurs