판별분석을 이용한 축산폐수처리공정 이상진단

초록

This study was performed to develop a diagnosis method for operating state of a swine wastewater treatment plant, for which discriminant analysis was applied. Using ORP(Oxidation-Reduction Potential) and DO(Dissolved Oxygen) profiles which were extracted from sub-cycle, semi-real scale SBR(Sequencing batch reactor) status divided into normal case and abnormal 6 cases such as controller malfunction, influent disturbance and instrumental trouble. With the preprocessing of ORP and DO profiles using Dynamic Time Warping(DTW) algorithm, the process drift problem at each batch process could be dissolved. After DTW process, several statistical values such as maximum value (Max), minimum value(Min), Average value(Ave), standard deviation value(stdev) and quartile values(Q1, Q2, Q3) were extracted and applied to establish the discriminant function. The accumulative percentage of the third discriminant function eigen value of ORP and DO was 92.9% and 99.6%, respectively, and showed dimension reduction using principal component analysis(PCA) could be applied effectively. The discriminant analysis makes possible to classify the 7 cases with the percentage of 100% and 92.7% in ORP and DO profiles, respectively. Consequently, it was concluded that DTW algorithm and discriminant analysis can be applied to stable SBR operation for the swine wastewater treatment.

키워드