Optimal PCA-based modeling and fault diagnosis for uneven-length batch processes

Principal component analysis (PCA) has been widely studied and applied in continuous process monitoring and fault diagnosis. However, PCA can´t be applied directly in batch processes due to the common multi-dimensionality of data matrix, uneven-length duration. Since the changes in the correlation may be used to indicate changes in the process operation stages, an optimal sub-stage PCA modeling method based on A-unfolding for uneven-length batch process is proposed, in which on the basis of analyzing the characteristics of sub-stage PCA modeling, the optimal model is established and the genetic algorithm is adopted to obtain the solution of optimal model. It is effective for batch processes with limited-runs modeling data and can improve the model precision. Simulation results to an injection molding process shows that the proposed method can partition the sub-stage accurately and it has better ability of process monitoring.