Tuning and Control Loop Performance, Fourth Edition

Gregory K. McMillan

• Publisher: Momentum Press
• Year: 2014
• ISBN:
  9781606501719
• Edition: 4
• Topic: Tecnología e ingeniería
• Subtopic: Automatización en ingeniería

Tuning and Control Loop Performance, Fourth Edition provides the knowledge to eliminate the misunderstandings, realize the difference between theoretical and industrial application of PID control, address practical difficulties, improve field automation system design, use the latest PID features, and ultimately get the best tuning settings that enables the PID to achieve its full potential. The proportional-integral-derivative (PID) controller is the heart of every control system in the process industry. Given the proper setup and tuning, the PID has proven to have the capability and flexibility needed to meet nearly all of industry's basic control requirements.

However, the information to support the best use of these features has fallen behind the progress of improved functionality. Additionally, there is considerable disagreement on the tuning rules that largely stems from a misunderstanding of how tuning rules have evolved and the lack of recognition of the effect of automation system dynamics and the incredible spectrum of process responses, disturbances, and performance objectives.

CHAPTER 1

FUNDAMENTALS

1.1INTRODUCTION

The proportional-integral-derivative (PID) as well as the proportional-integral (PI) controller is the workhorse of the process industry. More than 99.9 percent of basic control systems rely on this controller. Even when model predictive control (MPC) is widely employed, a PI controller is normally used for flow control to deal with the nonlinearities of the installed flow characteristic and to enable flow ratio control.

1.1.1PERSPECTIVE

For unmeasured disturbances, the PID has been proven to provide essentially optimal control (Bohl and McAvoy 1976). When fast and aggressive control is needed to prevent activation of a safety instrumented system (SIS) or relief device, a PID controller is necessary. A correctly tuned PID controller with a gain greater than 10 and a derivative time greater than one minute are sure signs that the immediate action of the proportional and derivative modes of the PID are needed.

In contrast, the MPC response to an unmeasured disturbance is more like the integral mode in terms of the prolonged gradual correction. The MPC output is a scheduled series of moves (changes in the manipulated variable) based on the difference between the setpoint and the MPC predicted profile that is computed from past moves with no knowledge of unmeasured disturbances. A fraction of the error between the predicted profile and the actual profile caused by the unmeasured disturbance is used to bias the predicted profile, to provide feedback control. The more gradual action and the inherent multivariable capability of MPC is particularly advantageous to interacting systems and gas or plug flow volumes. In these applications, a large primary process time constant or a small integrating process gain that would smooth out control actions and provide a separation of dynamics to minimize the consequences of interaction does not exist. MPC also provides exact dynamic compensation of measured disturbances.

This book will show that a high PID gain and rate time setting is indicative of a process time constant or integrating process gain that is extremely slow compared to the dead time. Often these dynamics are in a single loop or cascade loop, responsible for quality control. Overdrive of the output (output driven past the balance point) is needed to correct for a disturbance or to achieve a new setpoint in these near-integrating (large process time constant), true integrating, and runaway processes (highly exothermic reactors). Integral action provides the overdrive too late. Integral action has no sense of direction or anticipation. Excessive integral action can be unsafe in integrating and runaway processes. The most frequent problem observed for these processes is a reset time too small.

A PID module execution time that is less than one second is indicative of a loop with a process dead time and time constant less than one second. Since the fastest execution time of an industrial MPC is one second and the MPC latency is generally much greater than PID latency due to the complexity of the calculations, the MPC will be the largest source of dead time for these loops. The ultimate limit to the integrated error for an unmeasured disturbance on the process input will be shown in Chapter 3 as being proportional to the total loop dead time squared.

The PID controller has not been effectively used, primarily because PID performance depends almost entirely upon tuning and the correct application of PID options assuming the control strategy and configuration is correct. MPC is increasingly displacing PID because the MPC tuning is minimal and implementation is more automated. MPC is also better understood by chemical and petroleum engineers since the focus is on the steady-state response. In contrast, the expertise required for tuning and use of PID features for diverse applications is considerable and largely undocumented. The PID time frame is much shorter with the delay and initial rate of change of the response being more important than the final value of the process variable (PV) for the more important loops (e.g., concentration and temperature). Proper tuning and utilization of the PID depends upon the knowledge of the dynamics of the process excursion whereas the MPC performance is much more dependent upon the accuracy of the steady-state gains.

The tuning rules are largely a subject of continual argument. Over 100 tuning rules have been published, and all the authors are convinced their rules are the best. This strange situation is the result of an incredibly diverse range of dynamics and objectives in industrial process applications and the lack of understanding of the relationship between tuning and plant performance objectives. Gamesmanship is also at play. Everyone wants to win and show their methods are the best. Because there is no consensus on tuning rules and how performance depends upon tuning, a person can prove almost any point by how the PID is tuned for the case presented.

The PID has incredible flexibility. However, the powerful features responsible for this extensive capability are not extensively utilized because of the lack of understanding and guidance. These features can prevent oscillations from deadband (e.g., backlash), split range discontinuities (e.g., transition from steam to cooling water valve), resolution or threshold sensitivity limits (e.g., stick-slip), slow secondary loops, at-line and offline analyzers, nonlinearities, recycle streams, and interactions.

Chapter 1 starts with an introduction to the basic functionality of PID control and the advantages and disadvantages of each mode. The chapter moves on to detail the major tuning rules and to discuss the relative merits. The chapter concludes with a general solution to reduce down to two sets of tuning rules to be used in conjunction with software that can identify the process dynamics. The proper tuning gained from Chapter 1 sets the stage for Chapter 2 to concisely detail a unified methodology to get the most out of the PID for industrial process applications.

1.1.2OVERVIEW

Most of the tuning rules were designed with one or two particular types of dynamic responses in mind when in fact there are four major types of the dynamic process response. The significance of the size of the actual or equivalent primary process time constant is a commonly misunderstood key characteristic of process responses. The lack of recognition of the types of responses, nonlinearities, and process objectives has led to major differences in tuning rules. This chapter discusses the relative capabilities of methods for various processes in preparation for reaching a general solution. Without completely giving away the answer we itemize the functional recommendations.

1.1.3RECOMMENDATIONS

1.Carefully select the control action (reverse or direct) based on the process action (reverse or direct). Note that for a fail open valve if the valve action (increase-close) is not set in the PID, analog output (AO) block, current to pneumatic Transducer (I/P), or positioner reversing the direction of the output signal, the control action must be changed to the opposite direction to account for the valve action.

2.Determine whether output scale is in engineering units or percent and set the output limits to match the operating limits.

3.If separate anti-reset windup (ARW) limits exist (e.g., PRoVOX and DeltaV PID), set these ARW limits equal to the output limits using proper units. An exception to this would be when recovery from an output limit needs to be fast, set ...