Advanced Process Control go way beyond traditional PID loops and automated controls.

What is Advanced Process Control (APC)?

Advanced process control (APC) refers to several proven advanced control techniques, such as feedforward, decoupling, and inferential control. APC is implemented on computer hardware and in the simplest of applications on control devices like PLCs and DCS.

For complex applications where the multivariable inputs are significant and simple feedforward algorithms are unsuitable, Model Predictive Control (MPC) is applied.

Multivariable Model Predictive Control (MPC) leverages optimization algorithms to control multiple variables simultaneously. The greatest amount of value can be derived on complex & non-linear process control applications. Most often, MPC is hosted on computer servers with smaller localised applications embedded in devices. MPC is uncommon in PLCs.

 

Advance Process Control Technologies

What are the benefits of Model Predictive Control?

Model Predictive Control is a proven technology that reduces process variability and inefficiency, improves product consistency, increases throughput by allowing operations to push constraints to the limits and achieve higher return on assets.

MPC has decisively demonstrated its value as a best practice by increasing throughput and improving yield, energy usage, raw material usage, product quality, safety, and responsiveness.

The application of MPC supports the path to autonomous plants unable to be controlled by classical control strategies or where operators are required to constantly make changes to operational parameters.

Where is MPC used?

MPC is used in heavy industry and large process applications. Significant benefits of MPC can be realised for major asset operators, mining, refining, petrochemical, and large manufacturers.

It is surprising that APC take up has been soft given its pedigree is over two decades old, however this in no way diminishes the value of the outcomes. What this possibly means is there are few specialists or capable implementers and perhaps, limited vendors with a variety of case studies to share.

 

How do I implement Model Predictive Control?

At the very basic level, MPC applications start with the end in mind. Being with identifying what process outcome you need to create stability for, or what saving do you need to make. It might be you can yield a number of benefits, but one benefit may trump another, so being very clear about the number one target and what the hard constraints are, is the starting point. It is unlikely you can have everything but usually a number of benefits will make the solution viable.

Implement Model Predictive Control

Not all processes are economically viable for MPC. An initial study of the process, the instrumentation on the plant, the controls available and the target yield, will determine the viability of the application.

The quickest method to conduct a preimplantation study is to use historical operational data. If the data is of reasonable quality, then instead of using real time data captured over a number of months, models can be built and tested against your specific application. The model and historical data provides an indication that the yield is achievable.

Even though the model is built, the job is not done. The algorithm is still in its infancy. All predictable disturbances (beyond the basic model developed in the initial study) must be introduced to expand the algorithm’s strength in managing the control process. The preparatory study gives confidence that the strategy will produce tangible benefits.

During this implementation stage, adjustments polish off the solution in terms of control. New anomaly detection and analytics can be applied to deviations from the algorithm. This annunciation of performance deviation is a stark improvement over basic alarm management applied to most applications the world over. During implementation, the system is set to capture real time data and analyse performance against the model and then the systems control of the plant. Some further adjustments will be made to optimise the solution and move to fully automated. Sometimes there are unexpected surprises including forming some basic conclusions about pre-existing issues with elements of the plant.

How long does it take to implement Model Predictive Control?

Expected ROIs for MPC applications typically range from 1-2 years. This is a fantastic outcome given the high cost of production for complex processes, and in particular especially for high energy consumption equipment.

A desktop study may take 3-6 months per unique control process. In simple applications the solution will be online in 6 months.

 

If you need advice on the viability of your situation, contact one of our team for a chat.

New ways to solve old problems, with Parasyn.Model Predictive Control emulates plant operation

 

YOU MIGHT ALSO BE INTERESTED IN:

Asset Performance Management

Asset performance management is a mature discipline for the explicit purpose of improving the reliability and availability of physical assets. What is Asset Performance Management (APM)? Is APM software reliable? What does APM software do? What are the objectives of Asset performance management? What are the components of an APM system? How does APM help with cost avoidance?  Read more

Digital Transformation Platform Reporting Systems

What is the best reporting system for my process historian or my SCADA system How important is "dimensioning" data? How expensive are Business Reports? How important is Trusted Data? How do we ensure process data is correct? Does it really matter? Do I need new technology for Digital Transformation? From time to time we are asked, “what is the best reporting system for my o  Read more