You are on page 1of 5

A Introduction to Advanced Process Control (APC) for Managers Sandeep Lal

This is a introduction to advanced process control as practiced in chemical process industries. Most of the information given here is based on the authors experience in implementing and maintaining APC applications.

I.

Process control

The development of process control is based on control objectives. Some of these control objectives are listed below 1. Safety It is imperative that plants operate as safely so that the people and equipment is the plant is protected. 2. Environment regulations Process units must comply with the various environment regulations concerning the solid, liquid and gaseous discharge from the plant. 3. Plant operations A process plant needs to product that meets the customer specifications. The control objectives should be consistent with the economic objectives that are to make the product as cheaply as possible. The control system should also facilitate stable plant operation and maximize the production rate.

II.

Where does APC fit in Chemical Industries?

Most chemical businesses integrate their corporate tasks of Enterprise Resource Planning with Plant level functions of actually producing products. Even though the arrow shows information flow in one direction, it is actually a bidirectional arrow, with information flowing back to the enterprise of actual production rates that were achieved as well as what the plant can actually produce (given its equipment limitations, Maintenance activities etc). The ERP system would also contain other modules like Procurement and inventory, Financials, HR, Payroll etc. but we will focus on data flow to the plant as far as production planning is concerned. The advanced control gives its signals to the regulatory control which contains thousands of process variables that are measured and then controlled by PID loops running at a fast rate in the DCS systems. Besides these traditional control methods, sometimes advanced control methods are used for better control. These may be applied to processes that have. 1. Strong interactions between process variables 2. Strong non-linear behavior 3. Long time delays

4. Frequent disturbances that shake the process.

ERP

Supply Chain Management Production Planning & scheduling

Real time Optimizer


Plant Control network

Advanced Process Control

Regulatory Control
DCS

Process

Valves Sensors Analyzers

Figure 1 Hierarchy of information flow

Most of the software inside the Plant boundaries is controlling the actual process. Sometimes the production planning and scheduling is done at the plant level, but if there are multiple plants and multiple products being made and sold by the company, this could be at the enterprise level. So lets make a business case for APC.

III.

APC Success Story

APC makes money by doing the following: 1. 2. 3. 4. Constraint Management Throughput Maximization Reduce Energy Costs Stability

APC can be implemented in quite a few ways. The most basic way, was that used years ago by implementing advanced regulatory control (ARC) by Feed forward, constraint control, cascade control etc Recently Model Predictive Control (MPC) has become very popular for implementation as a APC solution in chemical process industries. MPC differs from traditional controller in 2 ways

1. MPC uses a dynamic Model to calculate the relationships between the Manipulated, feed forward and controlled variables. 2. It contains a optimization algorithm (usually a LP) that calculates the most efficient method of moving the process to and keeping at the desired conditions.

IV.

The players
1. Controlled variables (CV) are those variables that will be in the objective of the controller to maintain either at a limit or within limits, e.g. Product purity in a distillation column. 2. Manipulated variables (MV) are variables that the APC application will manipulate to achieve the objectives set by the controlled variables, e.g. Reflux flow in a distillation column. 3. Feed forward variables (FF) are those variables that influence the process performance, but cannot be manipulated by the controller, e.g. ambient temperature.

V.

Control calculations

As mentioned a MPC controller uses a model of the plant to calculate the plant performance in parallel to the plant. The controller uses the current plant measurements to 1. Calculate prediction of future plant behavior 2. Computes a optimum operating point for the plant 3. Computes appropriate control action required to drive the plant as close as possible to the desired value. The controller output is implemented in real time and the procedure is repeated every sampling interval with actual plant data.

Controlled Variables

Manipulated Variales

Feed Forward Variables

Control Move Plan


Figure 2 - Typical control calculation

VI.

Reasons for failure

As good as APC may have been portrayed in this document as well as documented evidence of it being the best thing since sliced bread, there have been instances where APC has not delivered promised benefits or not even been ON control to be able to control the process. I will try to highlight some of the reasons that I think could be responsible for this sorry state of affairs. 1. 2. 3. 4. 5. 6. Insufficient trained manpower to maintain the application. Changed economics needing retuning Bad regulatory control loops Changed plant after turnaround, which changed dynamic responses needing remodeling. Sub optimal initial design of the application Operator/operations/management acceptance - Bad communication

Most of these could be rectified by having trained manpower to maintain the applications. Many a applications were initially deployed with Consultants who came and implemented a application and everyone hoped that they would run themselves. Just like when you car makes some noises, we go to a Trained mechanic who looks at the problem and fixes it, APC is also a hi-tech application needing some TLC (Tender Love and Care).

VII.

Maintenance of APC applications

Maintaining APC applications is very important after the initial implementation. Figure 3 shows some of the long term status of APC applications with different levels of support.

With no support or monitoring the benefits of the application soon drop down and may even become zero. With poor support and no monitoring the life may be extended but the benefits soon drop to zero. With good support and no monitoring the application can stay alive but the benefits can drop down significantly. With good support and on-line monitoring of the application there is ownership and if there is operations buy-in the application can be improved over time and the benefits can be improved. When the application is first developed there is a set of constraints that are active in the plant, after the implementation of the APC there could be a different set of constraints that become active. Besides that the units have a turn-around cycle (typically 5 years). During the turnaround there could be de-bottle

necking projects that influence the application in 2 ways. One, by changing the set of constraints and two by changing the dynamic behavior of the unit. E.g. a pinch exchanger can now change the % feed vaporization to the tower and thus the flooding characteristics of the tower. Some of the factors that influence the APC lifecycle are equipment modifications, debottlenecking, deterioration of the regulatory control layer, feed quality changes, fouling, changes in operating philosophy, etc On a positive note, software vendors providing APC technology have realized the opportunities for revamping of APC applications and are offering tools for automated testing. These tools generate moves in the manipulated variables that are based on the seed models. The moves are small enough not to disturb the unit as well as large enough to get a information rich model by using a high Signal to noise ratio, also ensuring that there is not much correlation between moves. Since the testing ensures that the unit remains with constraints and abandons the test if the process violates the constraints programmed and makes aggressive moves to bring the unit back to normal operations these are safe to implement.

You might also like