Project Description

A leading snack food producer was experiencing occasional out-of-specification product and significant variation in the product quality from their batch fryers.

Project Details

Client:PepsiCo

The Challenge

ProBatch Control Technology Case Study

The goal of this implementation was to improve product quality and plant productivity by implementing batch closed-loop model predictive control.

By implementing ProSensus’ ProBatch Control technology our client reduced the standard deviation on their quality variation by 50% and also significantly increased process capacity, thus maximizing the return on the capital invested in process equipment.

The Results

The advanced batch control implementation results were impressive, providing major improvements in product quality and increases to productivity and profitability. The standard deviation of the product quality was reduced by 50%. This reduction allowed for a shift in a key quality target to a more economical level without exceeding the upper control limit.

Additionally, the production capacity was increased by 20%. With a greater than 99% uptime in 24/7 production and over 500,000 batches processed, this system has provided a very favourable return on investment!

Advanced Batch Control Implementation

The ProSensus Approach

The ProSensus ProBatch Control system calculates and implements mid-batch corrections to ensure that product quality targets will be met. This is accomplished through two main steps:

Batch Control Case Study1) By working with plant engineers, multivariate models (in this case PLS) are built from historical and designed experimental (DOE) data. Raw material amounts /quality information as well as process measurements and any fundamental models are incorporated into the models. Batch alignment is performed in this step when required.

These multivariate models will be used for process monitoring and control, but they also provide a deeper understanding of the process itself by highlighting key relationships in the data.

2) A few key points in the batch progression (decision points) are selected. The multivariate models are used to predict the expected final quality attributes using all of the information available up to each decision point. Numerical optimization methods are then employed to calculate control actions to achieve a desired quality while making sure any and all constraints are satisfied.

Monitoring metrics are also calculated at each decision point to provide insight into how well the batch is progressing compared to historical batches and the degree of confidence on the final quality prediction.

The controllers are implemented using the state of the art ProSensus On-Line Calculation Server (PSOCS). The PSOCS provides a live connection to any PLC/DCS system through a standard OPC-DA protocol adapter, and also simultaneously to relational databases via a JDBC connector. Once the ProBatch Control system is connected to live data via PSOCS, it can be set to run at triggered process events (in this case the decision points), or at desired time intervals.

References

  1. White paper available from ProSensus on request.