A diagnostic model for gear complex faults with deficient data

As an essential unit of Large Scale NC Machine Tools, the gear faults diagnosis is especially critical. However, due to the deficient data collected from monitoring system, these complex diagnostic issues usually exposing in modern industrial is still intractable. Therefore, a diagnostic model for gear complex faults with deficient data is developed in this study, where a size-variable re-sampling measure is designed to handle the mission and challenges with scarce fault monitoring data. Secondly, a gradual anti-noise recovery scheme is formed to solve the recognition troubles with missing surveillance data. Thirdly, an improved sparse auto-encoder based diagnostic model is built to identify the running state of gear. Finally, the effectiveness and superiority of the proposed diagnostic model are verified by gears complex faults tests.