Since the 1980’s the field of process control has become increasingly important in chemical and petrochemical plants, oil refineries and other manufacturing units. The widely used and still very powerful tool in process control domain is the PID controller. In order to get optimal performance of any PID controller and to extract the full economic and safety benefits of it, the PID tuning is a crucial step. This paper examines and compares industrial mostly-used old-fashion PID controller tuning methods with the brand new and most powerful PITOPS technology. Old-fashion PID controller tuning methods use Trail-and-Error approach or Empirical sets of rules, whereas the PITOPS technology uses very powerful mathematical NC-GRG (Nonlinear Constrained General Reduced Gradient) optimization approach developed by International American automation and process control company PiControl Solutions. The main goal of the paper is to highlight the benefits of PITOPS over the mostly-used old-fashion methods for industrial PID controller tuning, over several typical examples.
Need for automation and process control was felt back in early 1900’s, old factory mills, distilleries, breweries, mechanical transmission, and in many other examples. Since the 1960’s the field of process control has generated numerous money-saving ideas. Therefore, has become increasingly important in chemical and petrochemical plants, oil refineries and in other manufacturing, units in order to improve and keep stable their entire operation and production. Process control continues to be one of the most fascinating and growing areas with tremendous future prospects related to economy, safety and process stability. Optimal process control strategy can help by:
The primary control layer is the backbone in the process control hierarchy and it supports all advanced control and complex optimization applications. Study of the chemical and other manufacturing plants reveals that often the modern, complex advanced control applications receive primary focus and attention and often the underlying bottom-layer, primary control, issues are somewhat neglected because of lesser emphasis.
The heart and soul of a primary control layer is an PID controller. The PID control algorithm is the oldest, yet most popular and widely used control method. Amazingly, the PID control algorithm is unclear and misunderstood by many. If understood clearly, the PID control algorithm can provide, tremendous benefits, control improvements in a simple and robust manner.
A manufacturing plants can have anywhere from few tens to several thousands of PID controllers. No matter how large or small is the plant, PID tuning is a prerequisite for achieving any benefits. Also, changes in process conditions, market or economic conditions, hardware and equipment changes result in a need for changes in the PID control, which are rarely made, often due to lack of available tools or the required skill set. On the other hand, many control loops are not fully optimized due to lack of awareness of the potential and the resulting benefits.
Often the old-age Trial-and-Error, Ziegler Nichols (ZN) and similar empirical PID tuning methods are used with little chance of correctness and weak control quality performance. Guess work on PID controller tuning results in poor control quality, often oscillations or sluggish control followed by the control room operator turning off the advanced process control schemes or putting the PID controllers in manual mode. A plant then could run in this mode with many PID controllers in manual (inactive) mode for years and even decades. Therefore, proper process control PID tuning is increasingly important today in many chemical plants.
The basic PID control algorithm stands for Proportional-Integral-Derivative controller. Each of these terms in most of the times depends on the error (e). Error is the calculation value between the desired process trajectory i.e. setpoint (SP), set by the operator or some advanced process control logic and the actual measurement signal i.e. process value (PV), coming from the field measurement sensor. According to the present and past error values the controller output (OP), which moves final control element (valves, motors, etc.) has been calculated, in order to minimize the error. In the PID control algorithm PID parameters play a key role, where the most effective control action depends on their optimal values, as shown in Figure 1.
The PID tuning parameters determine the speed and stability of the PID control action. Proportional gain (P) represents the level of aggressiveness in the PID controller. The P parameter is like a gain multiplier. High P value means more aggressive PID controller action, called proportional contribution. Integral tuning parameter (I) represents the level of impatience in the PID controller. Most of the times the “I” parameter is in the denominator. Derivative parameter (D) shows the level of anticipation based on past history of PV movement (PV trajectory). When dead time is long, adding derivative is like being able to increase the proportional gain without causing oscillations. When derivative action is used it is allowed also to use higher proportional action as well. Derivative action is lead response of the PID controller which compensates the process lag.
For the best performance of the PID controller the values of P, I and D parameters have to be at their optimum. If PID tuning parameters are not optimal, control action will be sluggish or oscillatory. Bad PID control action can reduce product quality, make products sub-prime, off-spec, prevent production rate maximization and distract the operators/engineers focusing on the other important tasks in the plant.
The procedure opted to find the optimum values of PID tuning parameters is known as controller PID tuning or optimization. There are plenty of methods, tools & theories which are available for tuning of PID controllers, however finding the best parameters for the dedicated PID controller is still a tricky task. Mostly used industrial PID tuning methods are old fashion. In 90% of the cases they are: Trial-and-Error, Ziegler Nichols, Cohen Coon (CC), IMC (Internal Model Control) and Lambda methods. Against all of them, this paper introduces brand new and more powerful PITOPS PID tuning technology. It uses very powerful and sophisticated mathematical optimization algorithm NC-GRG (Nonlinear Constrained General Reduced Gradient) developed by International American automation and process control company PiControl Solutions and it will be compared with previously mentioned old-fashion methods.
As mentioned above still mostly-used industrial old-fashion PID tuning methods are Trial-and-Error, Ziegler Nichols, Cohen Coon, IMC and Lambda methods. These methods mostly rely of trial and error steps of questing PID tuning parameters or on some old-fashion empirical rules and equations set back around 1960’s when processes were not so heavily interactive, production grades an rates were not so changing and strict and tight product and environmental specifications were not so tight as today. Nowadays still, many processes suffer from typical problems such as:
All these old-fashion PID tuning methods rely on time consuming and large step-test changes by forcing the PID controller to manual mode and producing either too aggressive or sluggish PID control action. Many times, their performance and success rely on the experience of a process control engineer whose knowledge is based on many years of PID tuning and process understanding. Therefore, all these methods will never give optimal PID tuning parameter values required for tight and crisp PID control in auto nor in cascade mode when typical and frequent SP changes and the effect of process disturbances are present.
A typical PID tuning session begins when a plant operator requests help from a process control engineer to tackle with an oscillatory or sluggish response of PID controller to achieve its SP after a condition change or a load upset. According to gained past experience he starts to work on the PID tuning parameters using one of several old-fashion PID tuning methods. To tune a PID controller often process control engineer first needs to put the PID controller in manual mode. Often, due to process sensitivity and plant interactions this is a not a desirable situation. In manual mode he will do several step-tests in order to identify process dynamics needful for PID tuning parameter calculations. Many times, due to presence of process disturbances and extensive signal noise these step-tests need to be large enough for a human eye to distinguish the real process dynamics from typical process upsets. There can be cases of a frequent presence of disturbances and then the process control engineer is required to repeat already undesirable step-tests. Some of the methods require to make deliberately PID controller oscillatory and others require to wait PV to settles in order to identify desired process dynamics. But, many times long term step-tests or oscillatory behavior of a PID controller is a tricky and not a desirable situation in the plant. Approximate process dynamics helps to find some operable (but not always) PID tuning parameters. These operable PID tuning parameters need to be tested by changing online SP of a PID controller. If new PID tuning parameters are not performing well the job of process control engineer is to fine tune them. Fine tuning method is purely based on his experience and trial-and-error procedure. As seen from all above steps tuning of a PID controller by old-fashion approach require an experience process control engineer, uses a lot of time, produces high uncertainty of results and upsets the plant seriously due to needful step-tests.
A large number of manufacturing plants still use these methods. The reasons which account for usage of these methods are absence of engineering knowledge and understanding, unavailability of robust process control software tools for closed-loop system identification, PID tuning and optimization, and PID tuning parameters simulation and testing without conducting serious plant step-tests due to fear of causing shutdowns and plant problems.
PITOPS is basically short form of “Process Identification, Tuning and Optimization Software”. The reason why PITOPS is ahead of other market competitor’s, especially the ones which still apply old-fashion methods, is due to its ability to identify multivariable process models based on the closed or open-loop history or live data. It operates purely in time domain and mathematically optimizes and simulates the behavior of new vs. old PID tuning parameters based on the SP change, existing disturbances and signal noise, valve stiction or some other real-life and frequently seen process constrains. It does not require new, time consuming and large plant step-tests.
A typical PID tuning session begins when a plant operator requests help from a process control engineer to tackle with an oscillatory or sluggish response of PID controller to achieve its SP after a condition change or a load upset. Process control engineer will gather in excel past closed-loop or open-loop process data from the DCS or a PLC containing just PV and PID OP. Gathered data will be imported to PITOPS and based on them true process dynamics will be identified, without forcing PID controller in manual, producing oscillatory control action or making damaging step-tests. Complete process dynamics identification procedure takes just 3-5 minutes, whereas for old-fashion methods this step would require at least one to several hours, depending on the speed of the process dynamics. The next step would be to use identified process dynamics and based on it to tune the PID controller. In PITOPS the user can choose one of the following goals for PID tuning:
Once the PID tuning and optimization has been finished, the produced new PID tuning parameters can be tested using robustness algorithm. When the process control engineer is satisfied with new PID tuning parameters he just needs to enter them back to DCS or a PLC system and they will work properly. Additional SP tests and fine tuning are not required. The complete PID tuning procedure in PITOPS takes again 3-5 minutes, while the time required for old-fashion methods to work would be at least 30 minutes to 1 hour, again depending on the speed of the process dynamics. Therefore, it can be concluded that while one PID controller is tuned by any old-fashion method at least 5-10 PID controllers could be tuned using PITOPS, without any plant tests, upsets and mode changes. Table 1 shows complete list of practical process control functionalities which can be performed by old-fashion and PITOPS technologies.