Troubleshooting of an Industrial Batch Process Using Multivariate Methods

Abstract

Multivariate statistical methods are used to analyze data from an industrial batch drying process. The objective of the study was to uncover possible reasons for major problems occurring in the quality of the product produced in the process. Partial least-squares (PLS) methods were able to isolate which group of variables in the chemistry, in the timing of the various stages of the batch, and in the shape of the time-varying trajectories of the process variables were related to a poor-quality product. The industrial study illustrates the approach and the power of these multivariate methods for troubleshooting problems occurring in complex batch processes. Several new variations in the multivariate PLS methodology for the analysis of batch data are also implemented. In particular, an application utilizing a novel approach to the time warping of the trajectories for batches, and the subsequent use of the time-warping information, is presented. The use of the time history of the PLS weights of the process variable trajectories to diagnose problems in the dynamic operation of the batches is also clearly illustrated, as is the use of contribution plots for finding features which distinguished between the operating histories of good and bad batches.

Pseudo-sample based contribution plots: innovative tools for fault diagnosis in kernel-based batch process monitoring

“…Normal batch lengths are between 398 and 460 minutes and, therefore, before modeling the data, variable trajectory synchronization was performed by applying Dynamic Time Warping [30] stage-wise to guarantee all the process runs had the same evolution pace. The third dataset includes eight landmark features extracted from the variable trajectories of 49 batches (26 NOC and 23 off-specification) of a pharmaceutical batch drying process, originally described in [31].…”

“…The pharmaceutical batch drying process dataset [31] represents a specific case-study, in which LFE is used for monitoring batch processes. Here, the in-control models were built on 17 process runs, which evolved under NOC, while the remaining 9 were used to assess the adjustment of the limits of the resulting SPE and D-statistic control charts according to the T I risk values, as done when TCS-BWU was applied previously in this article.…”

Using warping information for batch process monitoring and fault classification

“…Note that these thresholds notably differ from each other, mainly due to the different patterns found in the warping profiles of faulty-1 and faulty-2 batches. As can be appreciated in Figure 4(a), most of the points corresponding to the warping profile of the selected faulty-2 batch fall inside the limits in the faulty-1 WICC, except in the time intervals [8,14], [45,50], [86,98] and [110,117]. The rest of the faulty-2 batches also showed the same pattern in the faulty-1 WICC (results not shown).…”

“…Hence, a study of the set of warping profiles obtained from the off-line and on-line synchronization is highly desired. Some authors have emphasized the importance of not discarding the information derived from the synchronization [10] and others have used this warping information as an extra variable in the multivariate analysis [11,14]. Nevertheless, there is no sound study on the use of the warping information for: i) unsupervised (i.e.…”

Real-time synchronization of batch trajectories for on-line multivariate statistical process control using Dynamic Time Warping

“…Westerhuis et al [25] used the cumulative warp information as an extra variable in the U-PCA model in order to ease the fault detection. García-Muñoz et al [8] used the cumulative time deviation from the average time as a new variable to reach a better fault detection, after synchronizing the batch data by the Indicator Variable approach.…”