Automatic baseline-sample-selection scheme for baseline predictive maintenance

A virtual-metrology-based (VM-based) baseline-predictive-maintenance (BPM) scheme was proposed by the authors recently. By applying the BPM scheme, fault diagnosis and prognosis can be accomplished and the requirement of massive historical failure data can also be released. The accuracy of the BPM scheme highly depends on the correctness of the baseline models in the BPM scheme. The samples of creating the target-device (TD) baseline model consist of the concise and healthy (C&H) historical samples and the fresh samples just after maintenance. Originally, each one of the C&H samples was checked manually to ensure that the sample was generated under healthy status and its data quality is good. However, this health-&-quality check process is so tedious and may also neglect deleting contradictory samples, which may deteriorate the BPM results and prohibit the usage of the BPM scheme. The purpose of this paper is to develop an automatic baseline-sample-selection (ABSS) scheme for selecting the C&H samples and deleting the contradictory samples automatically.