In automation implementation, control engineers need to master best practices for increasing loop tuning performance and reliability, including improving efficiency, controller gain, and understanding loop interactions.
Control engineers often face questions and concerns about loop tuning. There are usually no ready-made answers unless there is sufficient time to conduct step-by-step testing or study the loop's past debugging records.
Using the following guidelines, control engineers can determine whether loop tuning is appropriate, or if it is the root cause of performance problems, then appropriate changes should be developed or recommended.
01. Don't waste time on the flow loop. Set the adjustment to 0.25/0.25/0.0 and then proceed to the next step. Spending more time usually doesn't make much sense, as the process gain typically varies with the valve position.
The relatively small gain (the total ideal flow controller gain is 1.0) is reflected in operation at smaller valve openings, where the actual flow response is typically greatest. At larger valve openings, the difference can be compensated by a shorter integral time, typically increasing the flow to the setpoint within minutes, which is usually much faster than the rest of the regulation process.
This is usually the most important factor to consider when it is necessary to adjust any loop in a process environment in a timely manner while minimizing control instability. Large, slow valves may require longer reset times; occasionally, some loops may require lower gain, especially if they are operating with very small valve openings.
02
Don't waste time on level loops. Set the adjustment to 1.0/response time/0.0 and proceed to the next step. In a level control environment, response time is the average time it takes for a disturbance to occur without any control response, roughly the time required to complete filling (or emptying) from the setpoint (typically 50%). For example, typical response times for refining processes are: 1 to 3 minutes for small tanks, 5 to 15 minutes for drums, and 30 to 60 minutes (or longer) for large tanks. If a severe disturbance, overflow, or incomplete filling could have serious consequences, a larger gain should be used instead of a smaller integral time (excessive integral action on level loops is the most common cause of process oscillations). Level loop gains rarely exceed 2.0. This level control guideline applies regardless of whether the level flow is cascaded.
Level is an integrator, which is typically nonlinear, so proportional control provides the most reliable long-term overall performance. Under model-based control, several common alternative strategies exist, such as bias gain, surge control regulation, and dynamic optimization of tank inventory. However, none of these methods add value or improve reliability.
03
Don't waste time on temperature, pressure, components, etc., unless they are already cascaded with flow (and sometimes pressure). If cascading is not available, implementing cascading is the next best step. Cascading will linearize the process response, thus achieving optimal and reliable regulation across the entire operating range. In the absence of a flow meter, the controller output can be characterized based on valve characteristics; however, this technique introduces some uncertainties and loopholes, and is a suboptimal option.
04
Time is invested in pressure, temperature, and component loops, cascading them with flow (and sometimes pressure). In addition to linearizing the process response, the cascaded structure allows for precise determination of loop gain from process data at two or more operating points.
05
Because process conditions can vary, setting the gain to 1/2 to 3/4 of the observed gain value is recommended to maintain long-term stable and reliable performance, a situation that can be anticipated in most loops for various reasons. The integral is set to be equal to the process response time, which is determined using step-by-step testing, process experience, or historical data deviations. A slightly smaller gain and a longer integration time offer significant long-term performance advantages compared to traditional error minimization or 1/4 amplitude attenuation criteria (e.g., Zeigler-Nichols), up to the point where process time constraints are exceeded.
06
Do not use differentiation. Differentiation is a method of reducing the total error by using a larger gain and reset, and then relying on differentiation to brake. This is like accelerating and then braking near a stop sign to reach the destination faster. In most industrial process control applications, where process speed limits, minimizing overshoot and oscillations, and always protecting process stability need to be considered, differentiation is inappropriate.
Many modern control software packages often recommend setting the derivative to non-zero, which is an inexperienced approach. Model-based multivariable control algorithms do the same thing in achieving "error minimization" or "profit maximization," but as in single-loop control, it is often necessary to forgo aggressive regulation to provide long-term reliable performance. This is also a concern in addition to the traditional problem of large derivative abrupt changes due to transmitter noise or instability.
07
Controller gain is highly dependent on range. For example, a pressure controller with a range of 0 to 1000 pSIG per square inch (using a modern smart transmitter) requires a gain 10 times greater than the same controller with a range of 900 to 1000 pSIG (in past designs, similar loops were often used with higher accuracy in the actual operating range) to provide the same control response for a given error. Without understanding this difference, one might be unwilling to accept a larger gain value for the same function. Generally, for large-range temperature and pressure controllers, a gain of 1.0 is typically found for every 100 to 200 units of range.
08
Understanding loop interactions. When the action of one loop strongly affects other loops, the user needs to decide which loops should be tuned normally and which need to have their gain reduced and reset time increased. Doing so technically requires methods such as "decoupling" and necessitates altering the reality of process gain, which would disrupt any such loop. Therefore, there is often no practically reliable option to tune all loops in a set of interacting loops to achieve a timely response.
The speed requirements for interactive loops are based on similar principles to traditional cascading rules. For example, secondary loops should be detuned to be at least 3 to 5 times slower than the primary loop (high priority). This fundamental rule has been consistently ignored in multivariable control. This often leads to performance instability and ultimately performance degradation, which is practically unrealistic unless all involved models remain highly accurate.
09
When tight control is required, boldly increase the gain and carefully reset. There's a view that too much gain can cause oscillations, and since resets are time-based, shorter settings might result in faster control. In reality, proportional control actions are instantaneous, and excessive resets, especially when combined with too little gain, are the most common cause of process oscillations. Therefore, for situations requiring tighter control, use a larger gain (up to the limit of the actual average process gain) and tune the integral more precisely (being careful not to make it less than the actual process response time).
01 Understand when to use and not to use feedforward. Feedforward can be beneficial when there is a full understanding of the main disturbances, when the model (gain, response time, dead time) fails to change significantly in time for any reason, and it is necessary to ensure that hard process limits are avoided or to gain more benefits or avoid huge losses.
If these criteria are not met, especially if the model dynamics (response time and dead time) cannot be accurately known, feedforward should be avoided. Every feedforward model requires engineering implementation, reliability, and maintenance costs, a point already emphasized when using feedforward in batches in model-based multivariable control. Successful loop tuning means minimizing rework, detuning, readjustment, and remodeling. Understanding traditional single-loop tuning tools and methods (such as Ziegler-Nichols) also relies on understanding process operational performance criteria. When the process operational perspective is ignored, loop tuning is often prioritized during rework cycles (see Figure 1) rather than implemented at the stage where it is actually needed.
Figure 1: On-site efforts often prioritize loop tuning and allocate significant financial and human resources to remedy the impacts of disturbances or downtime, such as manual looping, frequent application of loop tuning to handle disturbances or downtime, and "redesigning" degraded multivariable controllers using the latest and most promising tools and technologies. Unfortunately, from a process operation perspective, important factors are often overlooked. Image source: ACP Performance LLC
Another important aspect is that model-based control tuning requires consideration of factors similar to traditional loop tuning, such as the effects of variable process gain, the effects of interactions, feedforward discretion, and leveraging robust tuning to maximize process reliability.
Initially, model-based control was thought to be a panacea for most problems, but experience with performance degradation, model maintenance, and short lifecycles has revealed how these principles still apply. Using these loop tuning techniques can help reduce rework cycles, increase success rates, and deliver years of reliable, maintenance-free process control performance for most loops.
Disclaimer: This article is a reprint. If it involves copyright issues, please contact us promptly for deletion (QQ: 2737591964). We apologize for any inconvenience.
