Process data are the most important information in all aspects of plant monitoring and control applications. These data, stemming from instruments, carry the necessary information that assists plant operations. One of the common problems of process instrument readings is their deviation from true values due to instrument bias or systematic error. Detection of change points in process data is the first step for a more insightful analysis of hidden factors affecting the process. In this paper, both Bayesian and Expectation and Maximization (EM) methods are considered for change point detection problem of multivariate data with both single and multiple changes. The performance of EM is compared with the Bayesian approach. Simulation results show superiority of EM in the case of improper selection of priors while the Bayesian approach has less computation demand. The proposed algorithms are evaluated through several examples, two from simulated random data and one from a CSTR problem. It is also verified through an experimental study of a hybrid tank system. (C) 42013 Elsevier Ltd. All rights reserved.