In thermal power generating plants the boiler tube failures causes for major downtime and outages of plant. The steam or water leakage on boiler due to tube puncture decreases the efficiency of the plant. The leakage inside the boiler reflects through the control parameters. It is necessary to take action before the problem become major defects. Data mining is the used to extort valuable knowledge from huge amount of data in which Data outliers highlights more significant information. Boiler control objects are non-linear control system. In this work operation abnormal condition has been examined using thermal power unit parameter history data, and then the data mining methods are applied to identify the fault occurrence. The multivariate data is reduced to single variate data using MAHALANOBIS DISTANCE and the results are tested with the statistical methods, GRUBB’S Test, HAMPEL’s Test and DIXON test. From the statistical experimental results, HAMPEL’s test is efficient in terms of outlier detection is proved.
Volume 11 | 07-Special Issue
Pages: 972-982