Remaining Useful Life Prediction of a steam generator based on Control Chart X-Rm, Area-Threshold and Markov models

Abstract

The condition-based maintenance (CBM) uses fault diagnosis and prognosis methods to characterize system future behavior and to advise some efficient maintenance strategies. The purpose of this work was modelling the Remaining Useful Life (RUL) of a steam generator used in a Thermoelectric Power Plant. The method consists of identifying early failure symptoms through the control chart X-RM; and the automatic recognition of non-random variation patterns, to generate the area-threshold (ATh) as a health indicator (HI). The transition between degradation states was simulated using Continuous-Time Markov Chain (CTMC) models. The sampling was taken from on a real-time data acquisition system. As a result, the RUL was simulated with adequate predictive performance, the maximum deviation from active time was 22 h. Predicted versus actual plot for the superheated steam temperature achieved slope 0,81, intercept 41, 20, R² 99,64 % and Mean Squared Error (MSE) 39,56; on the other hand, for the condenser pressure resulted m = 0,878, n = 45,149, R² 99,83 % and MSE = 30,98. Once the RUL is simulated, effective actions must be taken to avoid emergency shutdowns.