Bearing fault diagnosis based on DNN using multi-scale feature fusion

Deep Neural Networks (DNN) are widely used in the field of fault diagnosis since it can extract more abstract feature involved in data layer by layer. Existed DNN based fault diagnosis does not transfer all the features to the classifier since only the most abstract feature is fed into the classifier. This will lead to an inaccurate fault diagnosis result. In order to overcome this problem, a deep multi-scale feature fusion mechanism is developed by designing an additional multi-scale feature fusion network to fuse feature extracted at different scales. Thus more comprehensive multi-scale feature can be fed into the fault classifier to achieve more accurate diagnostic results. Experiment test for rolling bearing fault diagnosis shows the efficiency of the proposed method.