Non-Invasive Closed Loop Step Testing Technology for MPC Applications

Developed by: AIChE
  • Type:
    Conference Presentation
  • Conference Type:
    AIChE Spring Meeting and Global Congress on Process Safety
  • Presentation Date:
    April 30, 2013
  • Duration:
    30 minutes
  • Skill Level:
  • PDHs:

Share This Post:

You will be able to download and print a certificate for PDH credits once the content has been viewed. If you have already viewed this content, please click here to login.

Multivariable Predictive Control (MPC) is the most widely used advanced process control technology in the process industries. For instance, there are more than 6,000 DMCplus based applications currently in service worldwide. A common and challenging problem is that MPC control performance degrades over time due to inevitable changes in the underlying process unit, due to fouling, catalyst deactivation and mechanical wear. Among all possible causes of control performance degradation, poor model accuracy is the primary factor in most cases.  To sustain good control performance, the model needs to be periodically recalibrated.

However, it is a technically challenging and resource-intensive task to pinpoint a problematic model curve and re-identify a new model curve for replacement in a MPC application.  In a large scale MPC application, over a hundred variables may be involved, and there are often more than 500 model curves in a typical MPC model matrix.  Conducting a re-test and re-identification by a conventional approach may take an experienced engineer several weeks of intensive work and cause significant interruption to the normal operation. The process industry has been looking for an automated, safe and less invasive approach to conducting plant testing for many years. Over the past ten years, some technologies and tools have emerged to address this need; technologies such as step testing in an automated way with the process safeguarded by a specially designed multivariable controller.  Further the automated test can simultaneously perturb multiple input variables instead of the traditional single variable test, greatly increasing the average signal-noise ratio, leading to superb model accuracy. Even using these tools, the plant test will inevitably have a negative impact on normal process operation: the process may have to operate in an economically unfavorable condition for one to two weeks at a time. Therefore, the available technology in this field is still considered too invasive for the process industry. 

In this paper, an innovative approach for non-invasive closed loop step testing is presented.  It can be configured and executed to generate the required test data for model quality assessment and model re-identification, while also minimizing any negative impact to the underlying process operation by using a tunable trade-off between optimal process operation and process perturbation. More importantly, this technology can tolerate high model uncertainty in that it can keep an otherwise inoperable controller running smoothly with certain trade-offs in economic optimization. The new technology provides the process industry an opportunity to run a small amplitude background step test over a relativelive long period that is suitable for model identification purposes while can still keep the controller runing.  Such a test will not move too far away from optimal control targets therefore minimizing the interruption to the normal operations.

Multiple industry applications and simulation results are used to illustrate the advantages of the new technology, including: maintaining robusteness when the model is significantly degraded, keeping the process operating inside a sub-optimal range specified by maximum allowed economic give-away, and generating step testing signals maximizing the signal-noise ratio subject to both the process constraints and operational optimization requirements.




Do you already own this?

Log In for instructions on accessing this content.


AIChE Member Credits 0.5
AIChE Members $15.00
AIChE Fuels and Petrochemicals Division Members Free
AIChE Undergraduate Student Members Free
AIChE Graduate Student Members Free
Non-Members $25.00