Process monitoring with global probability boundary-based on Gaussian mixture model

Considering that the operation data hardly follow a uniform Gaussian distribution in complex industrial process, the Gaussian mixture model (GMM) is utilized as the tool of process monitoring in this article. The classic expectation maximization (EM) algorithm is adopted to estimate the model parameters, which often results in the model structural redundancy. Thus the consolidation operator is proposed and introduced to Figueiredo-Jain algorithm that is an advanced method of EM. The new approach can automatically optimize the number of Gaussian components on one hand, and solve the poor convergence problem of F-J method when Gaussian components overlapping too much during initialization on the other. With the obtained model, a criterion based on global probability is exploited for the real-time process monitoring. The validity and effectiveness of the proposed approach are illustrated through the coal-water slurry gasification control system.