The invention provides an electromechanical
product system or equipment real-time fault diagnosis method based on width learning, which provides a plurality of methods including
loss function optimization, namely cost-sensitive learning, random inactivation Dropout and integrated learning on the basis of a traditional width learning
system to obtain an improved width learning
system. Aiming at the problems of different data types, serious imbalance of the types and the like of actual
monitoring data, on the premise of guaranteeing the training and optimization efficiency, the cost weight, the inactivation probability and the like are set as adjustable parameters, integrated learning voting is carried out by imitating a bagging
algorithm, and a final result is predicted. The ubiquitous problems of uncertain influence and
class imbalance in fault diagnosis are solved, based on the improved width learning
system, training is fast, prediction is accurate, stability and robustness are high, the method is applied to real-time monitoring of the health state of a
complex system or equipment, faults can be prevented, and maintenance suggestions can be provided.