PCA-Based approach for fault detection in cement rotary kiln

Abstract

Principal component analysis (PCA) is a well-known data dimensionality technique that is widely used in industrial processes detection fault. Dynamic PCA is acknowledged for its capability to cope with autocorrelation in time-series. False indications impose one of the greatest problems in the monitoring of many processes. A study is carried in the first part of this work in order to compare the performances of Static and Dynamic PCA approaches in cement rotary kiln. The issue of false indications in fault detection systems is investigated in the second part. A monitoring approach based on constant false alarms rate (CFAR) is proposed to reduce false detections. A piecewise constant threshold is developed for ..2 and .. statistics to limit the rate of false alarms to a given percentile at each time instant. A control chart constructed using the rate of false alarms per window is used to monitor the process. Finally, the proposed monitoring technique is tested on SPCA to confirm the ability of the proposed approach to result in zero false detections without imposing high delays or misdetections.