Mechatronic systems are widely used in modern manufacturing. The key machinery of a manufacturing system should be reliable, flexible, intelligent, less complex, and cost effective, which indeed are distinguishing features of a mechatronic system. To achieve these goals, continuous or on-demand design improvements should be incorporated rapidly and effectively, which will address new design requirements or resolve existing weaknesses of the original design.

With the advances in sensor technologies, wireless communication, data storage, and data mining, machine health monitoring (MHM) has achieved significant capabilities to monitor the performance of an operating machine. The extensive data from the MHM system can be employed in design improvement of the monitored system. In that context, the present dissertation addresses several challenges in applying MHM in design optimization of a mechatronic system.

First, this dissertation develops a systematic framework for continuous design evolution of a mechatronic system with MHM. Possible design weaknesses of the monitored system are detected using the information from MHM. The proposed method incorporates an index to identify a possible design weakness by evaluating the performance, detecting failures and estimating the health status of the system. Second, improved approaches of intelligent machine fault diagnosis (IMFD) that can be applied to more general machinery and faults, are presented. This dissertation develops an IMFD approach based on deep neural networks (DNN).

It uses the massive unlabeled MHM data to learn representative features. Using very few items of labeled data, this approach can achieve superior diagnosis performance. The dissertation presents another IMFD approach, which uses the convolutional neural networks (CNN) and sensor fusion and has increased diagnosis accuracy and reliability. The end-to-end learning capability of the two approaches enables diagnosis of fault types or machines for which limited prior knowledge is available.

Third, a hierarchical DNN-based method of remaining useful life (RUL) prediction is developed. It achieves high accuracy of RUL prediction by modeling the system degradation on different health stages. This method generates a better estimate of the system RUL, which provides accurate information for the evaluation of system design.