Prof. Long Wen
Abstract: With the development of artificial intelligence and smart manufacturing, the data-driven fault diagnosis method has received increasing attentions from both academic and engineering field. To learn from the history mechanical data and establish the data models for fault prediction is the key characteristic of data-driven fault diagnosis. Deep learning is one of the most popular methods applied in data-driven fault diagnosis as the advanced data modeling technique. However, even though deep learning based fault diagnosis has achieved significant improvements, tuning the hyper-parameter for deep learning is still challenge. Many hyper-parameters have great effects on the performances of deep learning. To find the optimal hyper-parameters highly depends on the expert’s experience, and it is also a time consuming and labor intensive process. In this research, the automated machine learning (AutoML) method is applied in the deep learning based fault diagnosis to make the decisions for the hyper-parameters in a data-driven, objective and automated way. The AutoML method can also provide the state-off-the-art performance of deep learning method, but avoid the traditional tedious trial and error turning processing, showing a great potential for the usage AutoML methods.