(72a) Sustain and Improve APC Performance without Breaking the Bank

Since most Advanced Process Control (APC) applications are model-based, they will require on-going maintenance to sustain their performance and benefits. Poor APC performance is often due to a difference (mismatch) between the APC model and the actual process behavior. Model mismatch may be the result of normal changes in process behavior over time (e.g. fouling), minor process improvement projects (e.g. replacing a heat exchanger or valve), major debottleneck or turnaround projects, or other operational changes. An effective APC maintenance plan requires close coordination with plant operations, maintenance, planning/scheduling, project engineering, etc.

Making the changes required to improve the performance of a complex APC application can be very intimidating, In some extreme cases, there may even be a reluctance to make process improvements that might impact a critical APC application. APC should never be an obstacle for process improvement projects (or vice versa). Maintaining an APC application is often thought of as a major project that may involve significant costs and/or disruption to operations. But with preliminary planning and analysis, maintaining an APC application isn’t as costly or disruptive as many people believe.

This presentation will review general tools and techniques that can be used to identify the source and scope of APC performance issues and several different options if the model needs to be updated. In many cases, standard trending and data analysis tools can identify specific instrumentation, regulatory control and/or APC issues. There are several approaches to consider if the model needs to be updated: manual model adjustments (stop-gap solution), focused testing (when only a small part of the APC is affected), closed-loop testing (when the model mismatch is relatively small). In some cases, dynamic and/or steady-state simulation can be used to obtain or validate a model, with an appropriate degree of caution and verification.