A Novel Incipient Fault Detection and Diagnosis Scheme Based on Kernel Density Weighting Support Vector Data Description: Application on the DAMADICS Benchmark Process

Support vector data description (SVDD) is a classical process monitoring skill and usually uses Euclidean distance to evaluate the status of process. It should be noted that proposed evaluation method restricts detection performance for some faults, when overall fault has structural deviation compared with normal data. To address this problem, novel incipient diagnosis scheme based on kernel density weighting SVDD (KDWSVDD) proposed. Firstly, multidimensional estimation function threshold are obtained by training Next, adaptive weight given test sample through measuring probability difference between samples. Then, statistic in reconstructed complete weighted Finally, contribution graph extended diagnose abnormal variable fault. KDWSVDD can increase scale giving samples, so as effectively monitor The experimental results two numerical cases DAMADICS benchwork show SVDD, better