Modified convolutional neural network with global average pooling for intelligent fault diagnosis of industrial gearbox

develop the condition monitoring and fault diagnostic techniques for the gearboxes. Most of the modern gearbox fault diagnostic methods utilize vibration analysis to extract the fault features, and then make decision according to sophisticated signal processing techniques or expert knowledge of diagnosticians [1, 2, 6, 9, 18, 33]. For instance, Feng et al. [10] successfully introduced the Vold-Kalman filter into timefrequency analysis to extract fault features of the planetary gearbox under unstable operation conditions. Tang et al. [28] firstly presented a novel fault detection method to identify the categories of gearbox