Within the health-monitoring frame, fault diagnosis includes the following steps: modelling, detection, isolation andestimation. Quantitative-based methods have been successfully used so far in diverse applications. However when dealing withgradual fault and particularly in noisy environment the diagnosis becomes more challenging to obtain good performancesmeaning low false alarm and low miss detection rates. Recent results have shown that data-driven methods based on statisticalfeatures in the time, frequency, time-frequency or time-scale domains are effective for the monitoring of incipient faults (highSignal to Noise Ratio and low Fault to Noise Ratio).
Topics of the Session:
Data-driven approaches (mono or multi-dimensional),
Fault modelling, detection, estimation
Statisticalfeature extraction, distance measures,
Parametrical and non-parametrical methods,
Signal processing techniques (mono and multivariate),
Classification, discrimination
11月05日
2017
11月08日
2017
注册截止日期
留言