Designing a Hybrid Equipment-Failure Diagnosis Mechanism under Mixed-Type Data with Limited Failure Samples

The rarity of equipment failures results in a high level imbalance between failure data and normal operation data, which makes the effective classification prediction such difficult. Furthermore, many are dominated by mixed model unable to adapt this type problem. Second, replacement cycle production increases difficulty collecting data. In paper, an diagnosis method is proposed solve problem poor accuracy due limited method, synthetic minority oversampling technique combined with conditional tabular generative adversarial network. can be used predict mixture numerical categorical Experimental indicate that improve 6.45% compared other similar methods when account for less than 1% total