Integration of Batch-to-Batch and Within Batch Control Techniques: Application to a Simulated Nylon-6,6Process

Using a simulated nylon-6,6 batch process, this work presents three batch control schemes, 1) within batch, 2) batch-to-batch, and 3) integrated batch-to-batch and within batch, as improvements over fixed-recipe operation alone for disturbance rejection. The control schemes were developed using process understanding gained through analysis of a historical database of easily measured batch profiles. Various concerns regarding development and implementation of each strategy were discussed. The strengths and weaknesses of each controller's performance were discussed as well.

The analysis method used focused on separating batch measurement variability into time-axis and magnitude-axis components. Partitioning the data in this way generated time and magnitude "scale parameters" that described the normal variability in the process. These scale parameters provided improved process understanding and formed the basis for the improved control schemes developed in this work.

The within batch controller was a feedforward strategy that made mid-course recipe adjustments based on predicted deviation from target quality. The batch-to-batch controller utilized quality measurements to provide feedback adjustments to subsequent batches. The integrated control scheme utilized the predictive feedforward performance of the within batch controller tempered by the off-line feedback of the batch-to-batch controller in a cascade arrangement.

The three control schemes were compared to fixed-recipe operation. All three provided significant improvement in quality control. The within batch controller resulted in a 91% reduction in mean squared target error (MSE) over fixed recipe operation. The batch-to-batch controller provided an 87% reduction in MSE. The integrated control scheme was found to be the most effective providing a 99% reduction in MSE over fixed-recipe operation.