Historical databases are usually filled with information about plant operation during normal as well as faulty situations. This wealth of information acquired over time, if analyzed properly, can be beneficial in two ways: i) identifying current plant operation status and ii) abnormal situation management if such abnormality had occurred earlier. Here, a new data driven, unsupervised pattern matching algorithm is presented. Effectiveness of the proposed pattern-matching algorithm stems from the proposed similarity factor that is based on correspondence analysis. Correspondence analysis is a multivariate statistical analysis and it has been shown to possess better diagnostic abilities compared to principal component analysis. An efficient pattern-matching algorithm should be able to discriminate between normal modes and fault modes of plant operation. Here the proposed algorithm is shown to have better discriminatory ability compared to PCA based similarity factor. A simulation case study involving the benchmark Tennessee Eastman Challenge problem is presented here to validate the efficacy of the proposed approach. © 2013 AACC American Automatic Control Council.