Application of multivariate batch data analysis for troubleshooting of end-point product quality

A product is produced in a production train that consists of two batch reactors in series. The quality specifications require that a particular critical impurity (CI) in the product should be below 0.4%. The plant was experiencing series of poor quality batches with much larger critical impurity amounts and that was a cause of significant trouble. Multivariate batch data analysis discovered that possible mixing issues in the second batch reactor (R2) were the cause of the problem. Therefore the recipe was changed in a way that improved the mixing. Before the proposed solution was implemented, 18 out of specification batches were produced, out of total of 20 batches which is 90% out of specification. After the solution was implemented, only 3 out of specification batches were produced, out of total of 98 batches, which is 3% out of specification. The results achieved based on this work allowed the plant to produce the product within specifications and maintain shipping schedule.