Regulatory Control is the Foundationfor Advanced Process Control
Everyone knows you can't build anything lasting on a poor foundation; that's especially true when adding advanced process control over basic regulatory control.
W ant top-notch advanced process control? Get the basic regulatory control right, first.
A dictionary definition of optimization might read: 'A strategy giving the best result obtainable under a given set of conditions.' To the practicing control engineer, optimization usually means a highly theoretical exercise, which is not really relevant in the real world, where pipes leak, sensors plug, pumps cavitate, and valves stick.
Optimization in an operating plant requires integrating process know-how to maximize productivity, including paying attention to field devices, control strategy suitability, controller tuning, and wise use of basic regulatory control (BRC)-cascade, feedback, feedforward, ratio, lead/lag, etc.-at the lowest possible level of implementation. It is upon this foundation that effective advanced process control (APC) can be constructed.
Maximizing return on investment (ROI) is always an important goal of plant management. However, purchasing control equipment, transmitters, and a digital control system loaded with the latest advanced control software does not guarantee good control. If not managed, understood, and optimized to match control objectives for which these purchases were made, the ROI is poor.
Function of control systems
Process control systems convert raw materials and energy into usable products through an intricate series of processing steps and control techniques. By themselves, plant processes do not produce levels, flows, or temperatures and the introduction of such variables represent-operating constraints designed to ensure productivity, efficiency, and/or quality product is produced.
For a plant to produce high quality product at the least costs requires a structure of people, systems, and facilities integrated together and built on a foundation of management commitment.
Among the most significant management challenges are cost containment, product quality, conforming to government regulations, and achievement of production schedule targets. Managers who recognize and embrace the value robust process control contributes toward overall operating results gain competitive advantage for their company. However, many managers have been 'burned' by past promises of improved process performance from control system investments that never materialized. Too frequently, the result has been underutilized control systems.
Gaining maximum advantage from the available capability of currently installed control systems requires an organizational philosophy that empowers, supports, and organizes in a way that achieves the desired results. Simply purchasing advanced hardware and software does not ensure success; without management's commitment any system optimization program is doomed to failure. Success requires effective management of the people, facilities, and systems working together to achieve established goals. (See CE , Jan. '01 for articles about integrating people, facilities, and systems.)
People and systems
For any optimization process to be successful, an infrastructure of systems must be in place that provide the training and tools so qualified people can successfully carry out activities that lead to meeting defined goals and objectives. The introduction of high technology equipment-digital control systems, graphic monitors, smart transmitters, and artificial intelligence-into the work place do not reduce, but increase the requirement for training personnel. (See CE , Aug. 2000, ' Developing Intellectual Capital .')
Effectively optimizing regulatory functions of a process control system requires knowledge of the static and dynamic factors that exist between, and among, loops.
Control theory, as presented in university courses, is typically abstract mathematics. Once out of school, control engineers quickly forget what was learned and process control is practiced on an ad hoc basis without reference to the theory that governs the behavior of dynamic systems.
During the past sixteen years, Techmation (Scottsdale, Ariz.) consultants have worked in over 2,700 operating plants, testing, analyzing, and improving the dynamic operational characteristics of tens of thousands of control loops. During this time procedures and data analysis techniques have been developed and refined into a knowledge base that bridges the gap between theory and real-world practice.
Because process operations are dynamic, it's important that after consultants complete and leave, customer employees can maintain the control system at the level of optimization achieved.
To ensure a satisfactory level of customer employee proficiency, consultants mix classroom theory with practical experience to form a process audit and improvement team.
Following classroom instruction on how to efficiently and effectively use a personal computer to test and analyze process data, the audit team, under the supervision of the consultant, apply a methodology of identifying, auditing, testing, analyzing, and optimizing to establish a solid foundation for APC. Techmation's methodology follows.
Indentification consists of designing and conducting tests that identify critical measurements and control loops.
Auditing uses test data to determine the accuracy and capability of all measurements and loops. The team corrects problems discovered during the audit, such as calibration and valve errors. Critical loops are tested to determine installed characteristic. When installed characteristic are found to be highly non-linear, especially in the normal operating range, the problem is corrected.
Where required, controller algorithms are re-programmed and re-tuned to meet control requirements. If audit testing indicates installed control strategies do not meet objectives, the audit team designs, implements, tests, and documents new strategies that will meet defined control objectives. When completed, the audit report provides a 'capability signature' of the process personnel can use to maintain system integrity
Analyzing and optimizing consists of using new data, collected following the audit phase, to identify where changes or new control strategies could improve unit control. During this phase, the emphasis is on identifying unit optimization strategies and system configuration implementation methods.
By following an identifying, auditing, testing, analyzing, and optimizing methodology and mixing formal classroom training with practical experience, under the guidance of a consultant, users can optimize regulatory control systems and ensure they stay optimized.
Process control equipment, strategies
A poorly designed control strategy, inaccurate measurements, and poorly functioning control valves will result in less than optimum control.
Control system testing typically documents a number of equipment related problems that need to be fixed before the regulatory control systems dynamic operation can be optimized. Attempting to mask equipment problems by adjusting the PID (proportional/integral/derivative) and filter settings in controllers is often the response where issues of training, time constraints, and management commitment are manifest. In fact, as many as 50% of loops need some maintenance or configuration changes before the loop can be tuned to provide minimum variance regulatory control.
Regulatory control consists of two types of control functions that can be combined in an almost infinite number of configurations. The two types of regulatory control are: feedback and feedforward.
Feedback control can be configured for cascade, selective, ratio, and any number of other types of control schemes. All feedback control implementations have one thing in common. The controlled process variable measurement is compared to a reference, called the setpoint, and the deviation results in a corrective action by the controller. In short, feedback control only allows upsets to be corrected after they are detected. Tuning regulatory controls requires knowledge of the transfer function that describes the dynamic relationship between a change in the controller output and the response of the variable being controlled. Simplified first order models, rule of thumb procedures, and simplified rule based self-tuning controllers can, not in many cases, provide the best regulatory control solution for many of the dynamically complex processes found in control applications.
Feedforward control measures a load disturbance and introduces a dynamically compensated corrective action before the load disturbance affects the controlled variable. Tuning of feedforward control loops requires knowledge of the process transfer function in addition to the load upset transfer function. The only way to accurately obtain the necessary knowledge of the system dynamics is to test the installed system under operating conditions to obtain time domain data that accurately represents the process response. Time domain data must be accurately transformed into the frequency domain to obtain the best control solutions.
The function of the regulatory portion of the control system is to reduce variability in the face of changing conditions. Without an effective regulatory control system, each successive unit operation can introduce variation that can accumulate throughout the process and is reflected in the final product quality and overall cost of production. To produce a uniform product that consistently meets customer demands at the lowest cost, a regulatory control system must be in place to minimize variance throughout the processing cycle.
A modern process plant may have thousands of control loops. Advanced control systems cannot be operated as designed without a majority of these loops being in automatic control. The regulatory control loops provide four functions:
Allow the process to operate at a chosen target;
Minimize effects of load disturbances;
Reduce the effect of raw material variability; and
Provide for safe and efficient startup, operation, and shutdown of the process.
Hence, the regulatory control system's function is to maximize product uniformity under dynamic conditions.
PID is the most common feedback controller, has been in use for more than 60 years, yet is not covered by any implementation standard. In fact, no two digital control systems implement PID control in the same manner, understandably complicating tuning methods and calculations. For example, digital controller manufactures write software for PID controllers using the ideal, series, or parallel algorithm configurations. Because of PID implementation algorithm differences, loop-tuning parameters can be significantly different to accomplish the same task depending on the algorithm type implemented. Adding to the confusion, controller-tuning units can be expressed in different units and time domains such as:
Proportional setting being in gain, percent proportional band, or throttling range;
Integral setting in seconds per repeat, repeats per second, minutes per repeat, repeats per minute, or-in some cases-scan rate per repeat and repeats per scan rate; and
Derivative terms expressed in seconds, minutes, or scan rate.
Muddying the waters still further, some manufactures allow the end-user to select the controller algorithm and tuning units. As if the preceding was not enough to create a sufficient tuning maze, add derivative filter constant settings, positional or velocity digital implementation of the PID, configurations for PID, PI-D, I-PD setpoint response, and PV filtering options, and it's easy to understand why 'simple' tuning can be confusing.
Conversely, over the years, numerous special linear and non-linear versions of the PID controller have been developed that provide better control of particular processes. For example, a non-linear PID controller algorithm is available that eliminates the stick-slip cycling found in as many as 30% of fast control loops.
Other special controller algorithms are used for averaging control in level systems, eliminating hysteresis cycling in integrating processes, preventing overshoot when filtering is used, and the conditional integral configuration for batch control to name just a few.
Understanding how to correctly implement these feedback controller algorithms is important to insure minimum variance control. Indeed, the crucial consideration in regulatory controllers is to understand the complex picture that exists with a trained eye toward the application of the appropriate techniques and solutions.
Testing regulatory loops
Attempts to achieve 'total product quality,' using SPC (statistical process control) and compliance with ISO quality standards, ignore control loop details such as:
Correctly pairing of controlled and influential measurements;
Hysteresis, stick-slip, and sizing of control valves; and
Quality of the measurements, signal ailiasing problems, control algorithms, signal filtering, system configuration, and tuning of the regulatory control system.
Audit experience repeatedly indicates regulatory control systems are operational, but not providing optimum control. Findings indicate the typical regulatory control system contributes to as much as 50% of the non-uniformity of the final product. Testing of tens of thousands of unit operations with regulatory control systems applied consistently reveals the installed dynamics and loop tuning information, identifies equipment problems, installed characteristics of process loops, and relative gain of coupled loops.
Among the variety of problems identified, a few appear over and over in one of three general categories of: control valves, measurement, and control strategies.
Control valve problems
Stick-slip cycling -Tests reveal as many as 30% of rotary and high friction globe valves exhibit a tendency to produce stick-slip cycle at steady state operating conditions when the controller is tuned based on installed loop dynamics.
The stick-slip characteristics in a control valve results in cycling at steady state and can produce excessive process variability in unit operations. Stick-slip is a result of an excessive ratio of static to dynamic friction in the control valve, pneumatic stiffness in the actuator, and the performance of the valve positioner.
The steady state cycling of the controller output and the process variable being controlled is a typical stick-slip cycle. A linear PID controller measures the error between the setpoint and the process variable and ramps the controller output at a rate that is a function of the controller tuning parameters to correct the error. After a small and typically slow ramp change in the controller output, the valve position 'jumps' to a new position and the flow overshoots the setpoint. The error being on the other side of the setpoint causes the controller to again ramp its output to correct the error. The valve again 'jumps' to a new position and the cycle is repeated.
High-friction valves typically require tuning parameters that are 10 to 20 times slower than required, based on the installed loop dynamics to eliminate stick-slip cycling.
Correcting this very common problem requires a non-linear PID algorithm that sets the controller integral value to a lesser (slower) value only when the error is very small in the range where stick-slip cycling occurs. This algorithm has been installed in thousands of loops on numerous different controller brands with a net result of rejecting fast load disturbances and 'smoothing' steady state operation.
Hysteresis -Loose linkages in the actuator or positioner mountings-combined with friction in the valve-cause hysteresis or deadband in pneumatic control valves. Loop analysis testing reveals the normalized magnitude of the hysteresis in each loop.
Small amounts of hysteresis can usually be tolerated in self-regulating loops but will result in continuous cycling at steady state in integrating loops such as level, large volume pneumatic pressure, and batch temperature loops.
Techniques to eliminate hysteresis cycling in integrating loops include:
Fixing the control valve;
Placing the integrating loop in cascade where the inner loop is a self-regulating process; or
Implementing an error squared on integral control algorithm.
Installed characteristic -No control loop has a completely linear installed characteristic. In most loops this non-linear installed characteristic can be easily handled using an appropriate gain margin in the calculation of the controller tuning parameters. In some instances the installed characteristic in the loop is so non-linear the loop must be made linear before the loop can be tuned for optimum closed loop response under all system load conditions. When this is the case software, such as Techmation's Protuner, is used to record the controller output in percent and the measured variable in percent. This data can be analyzed and used to determine the cam characteristic that will result in a linear response.
The illustration shows an equal percentage installed characteristic. Contrasted with linear characteristics, when the loop in the illustration is tuned with the valve at low-end travels, the loop will become unstable at high-end travels. Conversely, tuning the loop at high-end valve travel will cause sluggish performance at low-end valve travels. Even when the loop is tuned for stability under all loads, the non-linear installed characteristic will result in a varying closed-loop time constant. To operate properly, the APC closed loop model needs to be varied as a function of load.
The compensating cam is the mirror image of the normalized installed characteristic. The illustration indicates the cam characteristics graphically and as a polynomial equation, making it easy to implement in the controller output or as a digital cam in a digital (smart) positioner. Implementing the can will linearize the process and loop tuning will be effective under all load conditions with the added benefit of constant closed loop dynamic in support of APC modeling.
Other problems -Valve calibration and valve sizing are also common problems found.
A recent paper released by a major valve manufacture revealed as many as seventy percent of installed valves tested require zero and span calibration.
Incorrect zero and span of control valves can lead to various control problems not easily solved. For example, if the valve is calibrated to operate 0% to 100% travel from 10% to 80% controller outputs there is a potential integral windup problem. Windup can result in large and unexpected overshoot and slow startup of the affected variable.
Techmation's experience reveals as many as 30% of valves are oversized and 15% are undersized for the installed application. Oversized valves typically provide poor performance due to lack of rangibility; undersized valves can result in production bottlenecks.
A generally understood first law among process control engineers is, 'You can't control what you can't measure.'
In testing of regulatory control systems, impediments to accurately measure the variables being controlled is hindered by such things as:
Lack of or improper setting of the required anti-aliasing filter constant in the transmitter;
Excessive noise on the measurement signal;
Improper use of or excessive signal filtering;
Incorrect PID controller configurations when measurement filtering is present;
Incorrect placement or mounting of measurement sensors; and
Incorrect transmitter calibration and scaling.
Unless measurement issues are addressed, optimum regulatory control is impossible and APC cannot provide the increased system performance expected.
Control strategy problems
Experience indicates as many as 17% of regulatory unit operation control strategies are incorrectly implemented and must be redesigned. The following example illustrates the results of a typical control strategy redesign to control column pressure.
The original strategy used a single controller split ranged to control both valves. Testing the system revealed the pressure control loop was very non-linear using the original strategy. Because the process is a mass balance its dynamics are integrating. As shown in the 'Typical closed loop hysteresis cycle in level loop under PI control' diagram above, any hysteresis or deadband in the valve will result in continuous cycling at steady state. To retain the original control strategy, and make it effective, would require linearizing the loop in software, and replacing the large butterfly valve because it is not a precision control valve.
The new strategy retains the existing large butterfly valve and controls the column pressure using the small precision control valve. The VPC (valve position controller) controls the position of the large valve to keep the small valve in range. The VPC control loop is a self-regulating process where the hysteresis and deadband does not result in cycling. Tuned correctly, the new control strategy provides accurate pressure control using existing valves with no cycling at steady state.
Regulatory control optimization results
Poorly implemented control strategies, faulty equipment, improper setup, bad installation, lack of anti-aliasing filters, and Murphy's Law are the conditions that require on-site system analysis testing to tune the regulatory control system.
Over the last several years, a major company's engineers have used the Protuner System Analyzer and the test procedures learned from the consultants to optimize unit operations at a number of facilities, yet they still find room for improvement. This clarifies the fundamental need for conscientious and consistent attention to the optimization of the regulatory control system to ensure economic APC ROI is achieved.
The tuning of the single and interactive loops is ascertained based on the analysis of the actual installed dynamic transfer functions of the individual processes. Tuning parameters are determined based on pole cancellation with adequate gain and phase margins, and damping factor requirements as a function of the installed linearity of each loop.
Feedforward tuning is based upon the actual dynamic transfer function models of the process and load disturbance.
The net result has been better, safe, and more efficient operating plants for this customer.
Advanced process control
Advanced control systems are implemented to control the process, not as individual levels, flows, pressures, or temperatures, but rather as each variable relates to productivity or efficiency of the process. From an advanced control system viewpoint, the control system is not single loop controls but a multi-variable envelope viewed as a polygon, with each side representing the constraints of pressure, temperature, etc. Within the envelope, the process is continuously maximizing efficiency.
There are a large number of techniques employed that come under the general category of advanced process control. The most common, yet least discussed advanced control strategy, is operator knowledge and confidence the regulatory control system works. In many cases, no matter what control strategy is implemented, operators will set the individual process variable setpoints at 'safe,' though not necessarily optimum target setpoints. Until operators are confident the regulatory control system is capable of safe and reliable operation-at or near process limits-the operator safe factor will often result in undesirable process integrity.
Another advanced-or optimization control strategy-is the proper application of regulatory control that is designed and implemented to maximize efficiency of the operation.
Examples include the use of variable speed pumps and correlated control strategies to allow the control system to better follow demand. Additional widely discussed advanced control strategies include dynamic matrix control, fuzzy logic, and multi-variable control. Each has several things in common including the obvious goal to continuously adjust regulatory control to maximize the operational efficiency of the process. Therefore, the success of an advanced control system is directly impacted by how well the regulatory control system functions.
Years of experience reinforce there is no magic panacea for the optimizing of a control system. In some cases, an advanced control package is sold as the ultimate answer to reducing variability and improving efficiency without adequate consideration given to the underlying regulatory control systems operation. Expectations of this nature are too frequently disappointing and can give process control an unjustified 'black eye.'
Most regulatory control systems are de-tuned at startup for steady state operation to mask design and equipment related problems, or because of a lack of knowledge about the processes dynamics. De-tuning the regulatory control system avoids 'troublesome oscillation,' but almost always results in the need to constantly adjust regulatory loop setpoints to overcome upsets 'caused' by the APC.
Management insight and commitment, trained and motivated people, the proper tools, and plain hard work on the regulatory portion of the control system form the necessary foundation to successfully apply APC.
The knowledge gained during testing and analyzing regulatory loops provides exactly the knowledge needed when developing, applying, and tuning the APC models and establish the foundation for continuous process improvements.