An expert control system, as we think of them, is an expert system, or artificial intelligence program, that has been specifically created to monitor and control some kind of processing plant in real-time. They are sometimes also referred to as a supervisory control program although supervisory control is not usually thought to utilize artificial intelligence components.

An easy way to understand where the expert control system fits into a typical plant’s control system is to understand the control pyramid. Many different pyramids have been defined over the years and the one below is one of the simpler ones.

The first level of the pyramid includes the instrumentation needed to monitor and provide basic control to start and stop the plant equipment. The second layer is all about stable operation once the plant is running and providing the ability to run equipment and processes at desired set point levels. In most cases millions of dollars are spent up to this point and often, further investments in operator training or supervisory systems are not made. The top level introduces an important final step up the pyramid. That is supervisory control of the process control system. The pivotal question once the basic stabilizing system is in place is, “at what levels do I set the set points and when do I change them.”

In our opinion the answer to this question should be found by looking at the rules and strategies that make up the plant expert system. In today’s competitive world it doesn’t make any sense to go to the expense of instrumenting a plant, providing closed loop control and then not taking the next step to expertly monitor the process and make timely set point changes to optimize all aspects of an operation or plant.

So with this brief background regarding where real-time expert control systems fit in the scheme of plant control we can define what an expert control system actually does. It monitors in real-time plant conditions and performance relative to process and management objectives and makes continuous set point changes in an attempt to improve performance as measured against the process and management objectives. One of the keys here is that the monitoring is continuous and relentless. Operators are well known to have an infinite variety of viewpoints on how to monitor and control a process. What is important to each is certainly not universally shared and is apt to change from one day to the next. Expert systems, however, are repositories of knowledge, experience, physical realities and management short term and long-term objectives. They can always explain their logic in the context of current operating conditions, recent history and their goals and objectives. It’s no wonder that the expert systems we have installed in minerals plants over the past 20 plus years always improve the average performance over that which the operators can achieve.

The best way to show what expert systems do is to look at histograms of plant performance achieved during statistically designed on-off tests run over weeks to even months.

This histogram shows an 8 percent increase in the average throughput rate and also shows a bimodal distribution indicating two different ore types.

The heart of an expert control system is the rules that comprise it. Examples of the types of rules that are used in KSX to maximize throughput rate are shown below.

So, in summary, an expert control system monitors a process in real-time and then makes set point changes continuously that a true "expert" would make to optimize the operation and performance based upon fundamentally correct rules that express correct cause and effect relationships.