Key points are not available for this paper at this time.
Due to the requirements of the system safety and reliability, the correct diagnosis or prognosis of abnormal condition plays an important role in the maintenance of industrial systems. In the last several decades, based on the welldeveloped physical model constructing techniques, numerous model-based diagnosis and prognosis approaches have been proposed, and many of them find successful applications in industry. On the other hand, with the wide application of sensors, the process data reflecting the system operation status can be easily collected. Based on these process data, the data-driven diagnosis and prognosis approaches study using data-mining and machine learning techniques for the purpose of process monitoring of industrial systems. Due to its potentials to boost efficiency and cut costs of industry, the diagnosis and prognosis under the data-driven framework have been an attractive research topic, and lots of related research results have been reported.
Yin et al. (Thu,) studied this question.