近期，我院教师范伟（一作，通讯）等的研究成果“Semi-Supervised Probabilistic Predictable Feature Analysis for Concurrent Process-Quality Monitoring of a Thermal Power Plant”在中科院Top期刊《IEEE Transactions on Instrumentation and Measurement》(IF=5.6)上发表
论文简介如下：
With the advent of statistical concurrent process-quality monitoring methods, quality-relevant faults and quality-irrelevant faults can be simultaneously discovered. However, in the presence of incomplete and asynchronous measurements of quality variables, corresponding dynamic process monitoring has become a challenging task. To address such issue, we propose a probabilistic method, termed semi-supervised probabilistic predictable feature analysis (SSPPFA), for online quality-relevant process monitoring. When modeling, both easy-measured auxiliary variables and available quality variables are leveraged to form a concurrent state-space framework. Correspondingly, the improved expectation-maximization (EM) algorithm is designed to solve the parameter estimation problem, in which the Kalman smoothing method can be flexibly adaptable to the missing issues and multi-rate sampling problems of quality variables. Further, four statistical indices, namely T2x,y , squared prediction errors SPEx and SPEy, and the dynamic index DIx,y, are designed to realize the abnormal condition detection. The advantage of the proposed SSPPFA-based monitoring framework is demonstrated through two practical faults of a medium-speed coal in a thermal power plant. Index Terms-quality-relevant process monitoring, semi-supervised probabilistic predictable feature analysis, dynamic process modeling, thermal power plant.
