Troubleshooting an Industrial Batch Process for the Manufacturing of Specialty Chemicals using Data Analytics

The troubleshooting of an industrial, multi-unit, multi-phase batch process for the manufacturing of specialty chemicals is considered in this study. The investigated problem is inconsistency in the final product quality, leading to the need of applying “corrections” to some batches with consequent significant increase of the processing time. Product quality information is scarce and available only for the last unit in the process flow diagram. It is shown that, by coupling the use of multivariate statistical methods to engineering understanding, one can step back in the process flow diagram to identify the unit wherein the problem originates, and to single-out the root-cause of the fault. For the process under investigation, fault isolation led to reduction of the cycle times and increase of productivity.