Geometric Process Control in Batch Process Industries

Batch processes have always been difficult to analyse and control because there are so many variables changing at once and so many quite different sets of activities taking place in sequence. Users have not been able to directly compare the many tens or hundreds of process variables and quality results across batches in a meaningful way, let alone the reasons for the differences in performance of two or more apparently identical reactors.

Process control has been limited to forcing the trajectory of a single variable, such as reactor temperature, to follow the same trajectory as that in a ‘golden’ batch.

That’s now at an end. Over the last four years working with some of our leading customers we have extended both the CVE and CPM products to analyse and control Multi-Stage and Multi-Phase batch processes.

• C Visual Explorer (CVE) can analyse multi-stage processes allowing you to link from end-of-batch product qualities, for many batches simultaneously, back to what was happening in an individual phase in the first stage of a batch.
• C Process Modeller (CPM) follows a trajectory tube composed of many variables and of constantlychanging size depending on the interactions between the variables and the present position of the process within the tube.

CPM uses the well-proven "no maths" Operating Envelope technology of Geometric Process Control to model each phase of a batch given a selection of good batches. Operating advice is given to the operator in open-loop but since the same software is used for both batch and continuous processes, closed-loop control is already available for use if and when required.

But how many batches are needed to build a model? We recently answered this question in a project with a pharmaceutical customer by rebuilding the model with successively fewer batches and running the same test data through it watching for the alarm rate to rise sharply. The answer was 6 batches if the manufacturing conditions were chosen using Design of Experiments methods... and 16 if they were chosen arbitrarily.

Best of all, there is no maths, so the models are easy to build and even easier to maintain. And because maths and programming skills are not required, the costs are an order of magnitude less than some other methods, meaning you can afford to apply Multi-Phase Batch Process Control to both major and minor products and processes.