DC‐link capacitor is one of the most vulnerable passive components in the drive system of electric vehicle, so the condition monitoring of DC‐link capacitors can significantly improve the reliability of drive system and even driving safety. The existing monitoring methods are subject to the low accuracy, added hardware and the irreversible impact on system. Therefore, based on Back Propagation (BP) neural network with Improved Gray Wolf Optimization (IGWO), a parameter identification method for the DC‐link capacitor in electric vehicle inverter is proposed. In this method, the capacitance (C) is taken as health parameter. The A‐phase current and DC‐link capacitor voltage in electric drive system are taken as inputs, and the capacitance (C) is taken as output for condition monitoring. IGWO algorithm can be applied to obtain the optimal weights and thresholds. Ultimately, under four actual working conditions in electric drive system, the condition monitoring test is carried out. The results are compared and analyzed, which show that the monitoring using IGWO‐BP neural network has better performance. © 2021 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.

Neural network has made remarkable achievements in the field of image classification, but they are threatened by adversarial examples in the process of application, making the robustness of neural network classifiers face danger. Programs or software based on neural network image classifiers need to undergo rigorous robustness testing before use and promotion, in order to effectively reduce losses and security risks. To comprehensively test the robustness of neural network image classifiers and standardize the test process, starting from the two aspects of generated content and interference intensity, a variety of robustness test sets are constructed, and a robustness testing framework suitable for neural network classifiers is proposed. And the feasibility and effectiveness of the test framework and method are verified by testing LENET-5 and the model reinforced by the adversavial training.