Skip to Main Content
A new method is proposed for fault detection and identification in complex systems. The technique of Fuzzy Wavelet Analysis is introduced which uses a fuzzified wavelet transform to analyze wide bandwidth fault features. A special attribute of this tool is its ability to employ localized time/frequency analysis of fuzzy data for fault detection and identification purposes. Performance measures of detectability and identifiability are defined to assist in assessing the performance of the algorithm. Performance improvement is achieved through a learning mechanism based on the detectability and identifiability measures. A fuzzy similarity measure is also introduced to reduce sensitivity to noise. The algorithm uses both on-line and off-line learning for designing and updating the rulebase.