By implementing advanced batch control, they were able to completely eliminate the bad batches.
A batch process at Mitsubishi Chemical Corporation was producing an emulsion polymer for which there was a very stringent specification on the average diameter of the final latex particles. The particle nucleation process is very susceptible to parts per million concentrations of undetectable impurities.
This leads to the nucleation of too many or too few particles, which in turn results in the particles at the end of the batch being too small or too large. This resulted in 10% of the batches having an unacceptable particle size over the years. The goal of this work was to reduce or eliminate these bad batches.
The results of this work are impressive. Bad batches were completely eliminated by implementing mid-course control.
The ProSensus Approach
Models were built using batch-wise unfolding of the available trajectory data to provide a real-time prediction of the end-of-batch quality as the batch progresses. This feedforward prediction is made at a key decision point midway through the batch.
Optimized control action is taken only on the batches where the predicted average particle size of the finished product is too large or too small.
As mentioned, the implementation of this advanced batch control technology completely eliminated the bad batches as well as reducing the quality standard deviation significantly. The figure below shows the dramatic increase in quality, which resulted in significant savings to the company.
- Y. Yabuki, T. Nagasawa and J.F. MacGregor, “An Industrial Experience with Product Quality Control in Semi-Batch Processes”, Computers & Chem. Eng., 24, 585-590, 2000.