Voices: Magazine Exclusives

Software that Reduces Maintenance of Multivariable Predictive Controllers

The Software Is Easy to Use. It Takes About 30 Minutes to Create

By Charles and James Cutler

The degradation of the models used by Multivariable Predictive Controllers (MPC) is primarily due to changes in the PID controller's tuning or configuration.  The physical plants vessels, piping, and inventories which set the basic dynamics do not change significantly with time.  As the models degrade, the performance of the controller eventually reaches such a low point that the controller is either turned off or the process unit is retested to upgrade the model.  Turning the controller off is not desirable, since the full value of the controller is lost.   Retesting is usually a high cost event because it requires engineering support from a contractor or some other part of the organization.  Many times the poor performance of the controller is due to equipment failure such as a valve becoming stuck or losing its plug.  The loss of one key valve that requires a shutdown to fix, forces the controller to be put on manual which can destroy the entire MPC's model and cause the controller to be turned off.

Also Read: Model Predictive Control - Past, Present and Future, Part 1

Cutler Technology's Universal Process Identification (UPID) software will be used to demonstrate how the problems with PID configuration and tuning can be reduced to a minimum.  A depropanizer model will be used as the process example.  The independent manipulated variables in the model are the set points for the feed rate, the top pressure on the column, the tray 5 temperature (5 trays from top), and the heat medium flow to the reboiler.  The tray 5 temperature is controlled by manipulating the reflux to the column.  The top pressure controller by passes the overhead condensing system and sends the hot vapors directly to the overhead accumulator to increase the pressure.  The MPC on the column controls the top temperature and bottom temperature.  The differential pressure across the column and the valve position on the pressure controller are the main constraints.  These controlled variables are dependent variables in the controller's model.  The other dependent variables in the controller model in Display 1 are there to facilitate the rearrangement of the control matrix when it becomes prudent to put one or more of the PID controllers in manual in the MPC model.

In Display 1, it should be noted that the PV for a specific manipulated variables comes to 1.00, i.e. the PV comes to its setpoint.  The response of this PV for the other manipulated variable is zero or comes back to zero at steady state for the other manipulated variables.  This response is true because the manipulated variables are independent from one another.  The valve positions of the manipulated variables change to make this independence possible.

In Display 2 the Top Pressure PID controller has been switched to manual in the model.  The Top Pressure valve position becomes the manipulated variable.  The setpoint for the pressure replaces the valve position as a dependent variable and tracks the pressure PV.   During the summer when the cooling water temperature goes up, it becomes desirable to switch the pressure controller to manual, to tell the MPC that it needs to use the other manipulated variables to control the pressure on the column.  As can be seen in Display 2, all of the other manipulated variables have a significant effect on the pressure.  The MPC controller with the pressure PID controller in manual can completely close the valve on the bypass around the condensers.

Closing the valve permits the maximum cooling for the column.

The Tray 5 Temperature PID controller is switched to manual in Display 3   The PID controller on tray 5 temperature is a cascaded loop that moves the setpoint on the reflux to the column.  The MPC controller model in Display 3 has the reflux valve position as one of its independent manipulated variables.  
In Display 4, both the temperature PID controller and pressure PID controller are in manual.   The MPC model is significantly changed from the original model when all the setpoints were functioning.  There are many reasons PID controllers are switched to manual during the course of a run.  A run may last a number of years, so every effort should be made to keep MPC on control.  A controller may be switched to manual when a valve is sticking so frequently that it is putting cycles into the entire process unit.  The process piping may not have adequate by passes to permit the valves to be blocked in and removed for repaired.

In Display 5, the models for the key constraint CVs are compared for the four set point case and the two set point case.  The valve position responses are obviously different from the set point responses.  Major differences in the gains also exist for the setpoint versus setpoint responses.  When the configurations of the PID controllers within a multivariable controller are changed, it affects the entire controller model.

Changes in the process mode of operation may cause one or more controllers to become unstable.   The PID controllers must be tuned to gain stability.   The new tuning changes the MPC model.  In Display 6 the PID controllers have become unstable.  The software allows the PIDs controllers in the MPC model to be tuned to match the plants tuning as illustrated in Display 7.  The ability to change the PID controller's tuning and configuration in the MPC model to match the current conditions on the process unit can significantly increase the on stream factor, the quality of the MPC control, and the profitability of the controller.

The software is easy to use.  The six models that follow, took about 30 minutes to create.  The controller has fourteen dependent variables.  Seven of the dependent variables were dropped out of the displays to make them readable.

More from this voice

Title

Do You Know Who Made Your Valves?

Counterfeiting is a slippery business. Here’s how to minimize the risk and danger of fake process controls.

11/06/2007

Cyber Threat to Control Systems: Are Companies Expecting Too Much Info?

The Industry Uses the General Term "Threat Information," but During More Detailed Discussions, It Seems That The Information Companies Seek Is More Like the Traditional Military Concept of "Tactical Information"

01/25/2011

Controlling the Whole Column

In the next installment of this series, CONTROL Columnist Béla Lipták, PE, discusses the overall control strategies for optimizing the distillation column to maximize production and minimize operating costs.

06/08/2007

Control Systems Cyber Security—The Current Status of Cyber Security of Critical Infrastructures

Testimony of Joseph M. Weiss Control Systems Cyber Security Expert before the Committee on Commerce, Science, and Transportation U.S. Senate

03/23/2009

Considerations for Selecting a Controller- or Server-Based Batch Sequencer

To Identify the Right Solution, Let's Consider the Three Types of Batch and Sequencing Solutions: Hard-Coded, Controller-Based and Server-Based

09/30/2013

Charlie Cutler's Latest Ideas for Multivariable Controllers

Process Variable Models and Adaptive Transforms

01/05/2012

Caro's SP100 Report

The ISA SP-100 standards committee plans to hold a demonstration of ISA100 at the Conference and Expo to be held in Houston on October 2-4. This, to stress the readiness of the networking technology and the cooperation among the potential suppliers.

08/30/2007

Cardinal Ethanol - Plant Sustainability With a Plant-Wide Optimization Control System

Achieving the Highest Levels of Process Efficiencies, Yields to Successfully Competing in the Industry

06/13/2011

Capital Preservation, Businesses’ Life Blood

If They Were Called ‘Instrumented Capital Preservation Systems’ Instead of Safety Instrumented Systems, Every CEO Would Insist the Company Invest in Them

02/05/2008

Brewery in a Box?

Bringing the Benefits of Distributed Control to Non-Traditional Industries Can Be Both Tricky and Lucrative

04/06/2010

Biodiesel Production Can Be Safe and Efficient

See How Greenleaf Biofuels Built Its Biodiesel Production Plant Within the Fuel Storage Tank Farm of New Haven Harbor in Conn. to Provide Easy Access to Truck, Rail, Barge and Deep-Water Vessel Delivery Points

02/21/2013

Bidirectional Flow Measurement

The Right Flowmeter Is a Balance Between Technical Needs and Cost-Efficiency

05/11/2012

Beach Blanket Control

“PlantBerries” Won’t Control Processes Any Time Soon, but They Do Break the Link Between Many Workers and Their Control Rooms

07/16/2008

Automation Project Management - Org Chart Omissions

Project Management Is as Much of an Art as It Is a Science. It Is the Art and Science of Getting Things Done

06/29/2009

Asset Planning vs Activity Planning

Dick Morley Becomes Resigned to Air travel, Messes Up a PowerPoint Presentation and Remembers a Valuable Lesson

03/28/2008

ASM Alarm Management Guidelines and ISA-18.2: How Do They Stack Up?

The ISA Standard and the ASM Consortium Guidelines Both Offer Important Information About Alarm Management. Although Their Approaches Differ, They Complement One Another

10/13/2009

A Return to Single-Loop-Integrity

Control in the field is becoming mainstream once again – just as the ISA SP50 fieldbus committee predicted it would be. CONTROL's regular contributor Dave Harrold looks at the integrity of the control loop.

03/19/2007