Explaining tuning rules poses a dilemma because understanding what the tuning can do requires understanding the significance of lags in the process. However, to understand the importance of lags, you must first understand what the tuning can do with lags present, similar to the complexity of the control loop. To address this dilemma, I’ll begin by discussing the tuning rules, but newcomers to the subject may need to switch back and forth between the tuning rules and the lags.

The fundamental principle of tuning involves setting the time and amount parameters of the controller to match the time and amount parameters, also known as dynamics, of the process. Tuning procedures teach the necessary dynamic characteristics of the process, typically by upsetting the process. While computer-based methods exist to learn the process dynamics, two simple, traditional ways include the automatic (closed-loop) and manual (open-loop) methods. It’s essential to understand both approaches and use them in combination as needed. Although the open-loop method is more challenging to use, it provides fundamental time and amount information about the process. It’s crucial that those involved feel confident performing the required procedures. Both approaches include a section on what to do and another on how to do it (procedures and techniques). The key is to disrupt the process sufficiently to obtain the necessary information without causing issues.

Tuning settings can be calculated before or after a loop is established, but this writing addresses tuning in the field (although tuning for level loops could be calculated at any time).

When it comes to the tuning rules, they assume that the controller algorithm is of the interacting type. The importance of this distinction is addressed in another section. However, you can safely assume that equipment from a major manufacturer is of the interacting type. In the case of digital systems, there may be a choice available, and it’s recommended to choose the interacting type.

Preparation

Before delving into the tuning rules, it’s important to prepare properly. Just like you wouldn’t drive a car without first learning the rules of the road, it’s crucial to take the necessary steps to ensure a successful tuning process. This may seem like overkill, but it’s vital to avoid any potential dangers or political fallout. Even if you’re experienced in tuning, it’s wise to use the following checklist:

Understand what you’re trying to accomplish and what to expect.
Know how fast and how far the process will respond to controller output and be confident in your ability to restore stability if needed.
Agree on the amount of change allowed in the process and controller output.
Ensure the person adjusting the controller settings is knowledgeable and won’t be confused in case of urgency.
Determine who will switch between manual and automatic and make setpoint changes if necessary, with confidence in their ability to take prompt action.
Be aware of safety interlocks and any other safety concerns for the process.
Choose a suitable time to tune, and be available for some time afterwards in case of any problems.
Record the existing settings and controller output for future reference.
Communicate any changes made to all operating personnel, preferably in a log book.

Closed-Loop Tuning: What to Do

The Ziegler and Nichols method consists of these steps:

Turn the integral time to as high a number as possible (often called “off”). With digital controllers this is usually to set the integral time to zero, which is not zero time but a convention to indicate no integral action. Turn the derivative time (if used) to as small a number as possible (often called “off”).
Increase the gain in steps until the loop cycles (there is technique to this). It is a fundamental truth that any control loop will cycle if the controller gain is made high enough.
Observe the period of cycling, Pn, to be called the natural period. Implant this securely in your mind. Nothing so simple is more important. Note also the gain at which it cycled, KCU, to be called the ultimate gain.
Set the controller settings to:

$K_{C}=K_{CU}/2$ (aggressive)

$K_{C}=K_{CU}/4$ (conservative)

$T_{i}=1.2P_{n}$

$T_{d}=P_{n}/8$

$T_{f}<=P_{n}/8$

where:

$K_{C}$ = Controller Gain, $\%$ output / $\%$ input

$K_{CU}$ = Controller gain that produced the sustained cycle, $\%$ output / $\%$ input

$T_{i}$ = Integral time, minutes

$T_{d}$ = Derivative time, minutes

$P_{n}$ = Natural period, minutes (from step 3, discussed below).

$T_{f}$ = Filter time

The optimal setting for the filter time is zero, as it is primarily used to minimize valve activity due to noise. However, in real-world scenarios, you may increase the filter time until the valve motion is acceptable, but it is advised not to go beyond the recommended value mentioned earlier. Additionally, if the recommended filter time is used in conjunction with derivative action, the two actions may offset each other.

According to most references, such as Ziegler and Nichols, the recommended tuning rules may vary slightly depending on the type of controller (P, PI, PD, or PID). For instance, the absence of integral action would typically call for a slight increase (about 10%) in gain, and the use of derivative action would allow for a slight increase in gain (between 10 and 20%) as well as a reduction in integral time (by about 30%). However, these adjustments are relatively minor and fall within the normal tolerances for setting the controller adjustments. Therefore, this book will not place special emphasis on whether the controller is P, PI, PD, or PID. Nevertheless, it’s worth noting the directions in which these settings may be changed, and as you become more familiar with the concept, you may understand why.

Closed-Loop Tuning | Approach

The fundamental concept behind the closed-loop tuning method is to allow the loop to cycle without encountering any issues, monitor the natural period and ultimate gain at that moment, and then retreat. Before conducting any tuning, it’s crucial to follow the safety precautions mentioned earlier. After completing the safety preparations, the next step is to disable the integral and derivative functions, or reduce them to the lowest level possible, in order to maximize the integral time and minimize the derivative time.

If you are unsure of the stability with the current gain setting, the next step is to make a small setpoint change in the direction that is deemed safest and observe the response for signs of cycling. It is beneficial to have an approximate idea of what the cycle period is likely to be at this point. If you suspect that the current gain setting is too low, you can save time by not making a setpoint change and instead increase the gain by a factor of two. Making small changes in gain, such as 10% or 20%, is generally not useful during this initial effort to determine how close the loop is to cycling.

To start with, if the first setpoint change does not lead to the beginning of a cycle, the gain should be increased by a factor of two, and another small setpoint change should be made in the opposite direction to ensure safety. The goal is to identify any indications of a cycle in the process. This process of increasing the gain and testing for stability should be continued until a suggestion of a cycle is observed. At this point, the gain should be changed by less than a factor of two, for example, by 50%, and the procedure should continue as before. As the loop becomes more oscillatory, smaller changes in gain should be made. If the controller design allows, it is often possible to safely adjust the gain while the controlled variable is already in motion. This can save the need to make a new setpoint change to disrupt the process, since the process is already being disturbed, and that is the intended purpose of the initial step.

Be vigilant for any variations that appear to signal the start of a cycle. When searching for indications of cycling, it is helpful to monitor both the controlled variable and the controller output. The first signs of cycling are often visible in the controller output. If the controller output becomes saturated, that is, reaches a high or low limit, exercise caution. The results may not be valid for use in this process. Try reducing the magnitude of the setpoint change or inducing a disturbance to the process, rather than using a setpoint change. Nevertheless, remain cautious if the controller output becomes saturated. If it becomes saturated during the test, it may also become saturated during normal operation, and this could potentially cause stability issues.

If the amplitude of the cycle begins to increase, be prepared to revert to the original gain setting, or even to switch to manual mode if deemed the safest course of action. The switch should be timed to coincide with the output approaching its original value. It is important to consider any potential characteristics of the controller, such as some controllers that cannot be switched to manual, or others that bump the output when the gain is changed. These characteristics should be taken into account when adjusting the gain. If the controlled variable is already in motion, it may not be necessary to disturb the system with a setpoint change, as the sole purpose of such a change is to observe some activity.

Figure 1.1 [With a proportional-only controller the error (offset) is reduced as gain is increased, but at the price of an increased tendency to cycle.]

The response curves expected from this procedure are displayed in Figure 1.1. When the gain is significantly lower than the ultimate gain, the process will respond without any indications of cycling, and the process variable will only change a small fraction of the amount requested by the setpoint change. It should be noted that the gains and offsets (errors) depicted in Figure 1.1 are merely examples and do not necessarily represent the numerical values you will encounter, but rather they illustrate trends. The figure is not meant to suggest that the loop being tuned will be stably controlled at a gain of 1 and become unstable at a gain of 6. Rather, it demonstrates that increasing the gain will result in two outcomes: a reduction in offset (error) and an increase in oscillatory (unstable) response.

Determining the exact gain that produces a steady cycle is generally of little value. The difference between this gain and one that produces a slightly decaying cycle is small, and the period of cycling observed with a slightly decaying cycle is close enough to the true natural period to determine the controller settings. The potential reward for determining the exact gain that sustains a steady cycle is typically not worth the increased risk that the cycle will continue to increase in amplitude and cause safety or quality problems.

Once the gain that will sustain, or almost sustain, a continuous cycle is determined, observe the period of the oscillation. Then, promptly cut the gain in half to achieve stability. This period is known as the natural period, which some authors refer to as the ultimate period, and the gain that caused it as the ultimate gain. These values are then input into the formulas provided to establish the controller settings. Note that the gain can be set within a range of values. Once the settings have been established, verify their acceptability through a small setpoint change or by observing the behavior under operating conditions.

Observe the period of the oscillation.

Figure 1.2 illustrates the impact of adding integral (reset) action to the process used in Figure 1.1. The numerical values presented should be used to comprehend trends and approximate relationships only. The addition of integral action reduces the error to zero over time and increases the likelihood of cycling. The period of this cycle will be longer compared to when only proportional action was employed.

Figure 1.2 [Integral action in a PI controller eliminates the offset. Decreasing the integral time increases the tendency to cycle, and at a longer period than for proportional-only.]

If the performance of the selected settings is deemed unsatisfactory, it is generally not necessary or advisable to repeat the entire testing procedure, including turning off the integral and derivative functions. The natural period, which these settings are based on, has already been established. It is important to remember that the tuning rules are meant for typical loops. To gain experience and confidence in tuning, you can modify the settings to achieve the desired performance using the concepts presented in Figures 2.1 and 2.2 to adjust the response. Stability can be increased by reducing gain, and increasing integral time can also increase stability if the dampened period is considerably longer than the natural period. However, if the derivative time is too long, it will increase the tendency to cycle, but the period will be shorter than the natural period. More will be explained about this later in the section on open-loop testing.

There are digital systems that offer programmed tools for tuning, although I cannot provide specific information on their effectiveness. It is possible that a certain level of tuning knowledge is required in order to use these tools effectively.

To ensure that others can benefit from your tuning efforts, it is important to record the ultimate gain and natural period values. It may be useful to keep your own records as well, and to note any relevant operating conditions such as production rate, setpoint, and controller output.

Finally, it is important to inform anyone who may be affected by the tuning changes of the original settings and the new settings. This helps to ensure that everyone is aware of the changes and can adjust their processes or procedures accordingly.

Open-Loop Tuning | Introduction

Ziegler and Nichols proposed the first-ever procedure for determining controller settings from an open-loop test back in 1942, and since then, other authors have refined the method. Ziegler’s procedure involves placing the controller in manual mode, and then making a step change in the controller output when the process is stable enough. The process response is then observed and is expected to resemble Figure 1.3. This curve is now known as the process step response curve, and if the process levels out, it is considered self-regulating, while if it doesn’t, it is non-self-regulating or integrating, similar to the integrating function of a controller. The open-loop tuning rules generally apply to both self-regulating and non-self-regulating processes, though modifications to the rules are needed if the process is self-regulating, and the time it takes to reach equilibrium is short relative to L.

From the step response of Figure 1.3 the tuning settings are determined as follows:

$K_{C}=1/RL$ (aggressive)

$K_{C}=1/2RL$ (conservative)

$T_{i}=5L$

$T_{d}=L/2$

$T_{f}<=L/2$

$R$ = Rate. It is the change per minute in the process variable, (expressed as a % of the transmitter span), divided by the step change (expressed as a % of the controller output span).
$L$ = The apparent dead time.

Previous comments have addressed the use of the filter and the type of controller used (P, PI, PD, or PID). However, there is a particular response that requires special attention: when the lag after the apparent dead time is relatively short compared to the apparent dead time, indicating that the process is self-regulating. To discuss this situation, Figure 1.4 is provided.

Figure 1.3 [The open-loop step response yields parameters R and L, from which controller setting may be determined.]

Several additional parameters are introduced, including the step size (B) and the final change in the process (A). The 95% response time is also defined, as shown in the figure. Although the ideal response time is 63%, this may be difficult to determine beforehand, especially if the final change (A) is unknown. Even if the response is being recorded, an exact determination of the 63% response time is not essential.

You may “eyeball” the 95% response time and then divide by 3 to estimate the 63% response time, $T_{63}$ . If your system is responding according to typically assumed ideal responses this relationship of $T_{95}$ to $T_{63}$ is essentially exact. Since most processes are not that cooperative, I often use the 90% response time. If you are judging it by eye it is hard to tell the difference between 90% and 95%. This special situation that I am discussing is worth separating from the rest of the field only if $T_{63} = L$ is 3 or less (perhaps even 2 or less). If it is, then the rules for tuning are:

$K_{C}=1/3RL$

$T_{i}=T_{63}$

These modifications to the rules have you setting the gain lower than you would with the closed loop rules and the integral time much shorter than with either the closed- or open-loop rules. This means, of course, that if you choose to use these modifications, you must use them together, you cannot use them to set just the gain or just the integral time. Essentially you are going toward integral-only control as $T_{63} = L$ becomes small, an option that was rarely exercised before digital control. The closed-loop rules will still give you stability and reasonably good performance, but they will give you different settings.

On rare occasions the slope, $R$ , will continue to increase. This situation is often called open-loop unstable or conditionally stable or runaway. On other occasions the response may go in the other direction first, called inverse response.

The open-loop rules for tuning do not apply in either of these cases, and indeed, the closed-loop procedure is subject to pitfalls then too.

Figure 1.4 [The step response of a self-regulating process can yield parameters helpful to guide tuning if the response time is short relative to the apparent dead time.]

Open-Loop Tuning | Approach

The open-loop method for tuning controllers is a more detailed approach that provides fundamental insights. However, it requires additional attention compared to the closed-loop method since the operator may not be confident enough to manually control the process or may be unsure of the appropriate step size to apply. The fundamental concept of an open-loop test is to identify key time and amount characteristics of the process by introducing a step change in the controller’s output.

To ensure the usefulness of the open-loop test, it is crucial that the process is running smoothly beforehand. Otherwise, it might be difficult to distinguish between the effects of the step input and other disturbances. Therefore, it is important to achieve a reasonable level of process stability before switching the controller to manual mode. Most modern controllers have a bump less transfer function for switching between automatic and manual modes, but if your controller does not have this feature, do the best you can. It is essential to avoid any interference with the planned step change in the controller output, such as from bumps during the transfer or stabilization process. In such cases, it is best to let any resulting effects settle before attempting the step test.

To prepare for observing the results of the open-loop test, there are a few options available. One could mentally note the apparent dead time, or record the results on a test recorder. If a DCS with a SCADA layer display is available, it’s recommended to set it up to show the process variable and controller output. However, in this case, the scales must be selected, which is a guess-and-test process. It’s important to select a scale that displays the step response well, without exceeding the set scales or being too small to see. It’s recommended to use a time span of about 10 times the apparent dead time to observe the results effectively. In case someone has to record the resets manually, it’s important to occasionally have them call out and record the data.

Before performing the step change, it’s important to determine the appropriate size based on safety and quality considerations. It’s possible that a conservative value may be chosen. With modern controllers, it’s typically easy to precisely make the desired change, but with older equipment, it may be challenging to make small adjustments accurately. It’s important not to continuously tweak the controller output to achieve the exact desired change as it can result in an upset that is not a step. This is especially problematic if it happens slowly relative to the apparent dead time. If this occurs too slowly, the results may not be accurately interpretable as intended.

To perform an open-loop test you must be confident you can control the process in manual. It is very desirable to have some idea of what $R$ and $L$ , and especially $L$ , will be.

Figure 1.5 [A useful technique for impressing a step change in the controller output is to impose a balanced disturbance, as shown above.]

Perform the step change in the controller output and observe the resulting changes in the process. Once the results are observed, reverse the step change direction and repeat the process, returning the output to its initial value. If the process does not return to the expected value, manually adjust it to the desired value. This situation is likely to occur in integrating processes or when equilibrium cannot be reached. In such cases, doubling the step size in the opposite direction, as illustrated in Figure 1.5, provides a balanced disturbance to the process. The goal is to return the process to its initial state. This method is also useful when the appropriate step size and duration are unknown. Begin with a conservative step size and duration and gradually increase both until the desired results are obtained.

The essence of the open loop test is to get a useful response curve without exceeding safety or quality limits.

Performing the open-loop test multiple times is highly recommended, especially if there are other factors affecting the process that may interfere with the results. In some cases, these factors can be identified and eliminated by putting another controller in manual. If not, increasing the step size may help to highlight the effects of the step change. It is also useful to create a hard copy of the results if possible, especially if monitoring the process on a screen. This can help to accurately determine the values for the apparent dead time, $L$ , and the rate, $R$ . If these values vary significantly, it is recommended to use the largest $R$ and longest $L$ when calculating the controller settings, to err on the side of caution.

If the response of the process is self-regulating and has a short response time compared to the apparent dead time, then it is recommended to use the modified rules described earlier. This means using a shorter integral time and a slightly smaller gain compared to the general rules.

It’s also a good practice, if possible, to perform step changes of different sizes and directions. If the response curves are not proportional to the step size, it indicates the presence of nonlinearities, which could cause issues.

To check for dead band, one can make small changes in the same direction and then reverse them with the same two small changes. If the process doesn’t repeat itself for the same outputs, it’s an indication that significant dead band is present.

After completing the open-loop testing, it’s time to calculate and implement the settings, and switch the controller back to automatic mode. It’s advisable to perform small setpoint changes and observe the response to ensure it aligns with expectations. If the response is sluggish, try increasing the gain by a factor of 1.5 or 2. If it’s oscillatory and the period is much longer than the natural period, increase the integral time by 50%. These suggested adjustments are a starting point, but avoid making small changes of 10% or 20%. The goal is to achieve a response that’s complete within three to five dead times, not 20 to 50.

It’s essential to monitor the performance over time to confirm there are no issues. If the response is significantly different from what’s expected, it’s possible that there’s an abnormality in the loop, or the loop is continuously impacted by disturbances. If the tuning changes don’t produce the anticipated results, and no disturbance could have caused the behavior, seek the advice of a more experienced person.

This, of course, assumes that the issue justifies further examination, and other efforts to address the observed behavior have been unsuccessful. After tuning, don’t forget to perform responsibility tasks, such as staying available and keeping people informed of the changes made.

When to Use Open-Loop Tuning

The controller settings obtained from open-loop and closed-loop testing methods may not be identical, but they should be similar. Open-loop testing is often preferred when a monitor is available to display the controller output and controlled variable on an expanded scale, as it allows for a more detailed examination of the response. This enhanced level of scrutiny has greatly increased the use of open-loop tuning. Additionally, the open-loop test is a valuable tool for diagnosing issues, as deviations from expected step responses can provide insights into the underlying problem.

The open-loop test is very useful for troubleshooting.

If the process is self-regulating, the open-loop response can provide more valuable insights. Assuming a step change of size $B$ results in a final steady state change of A, as shown in Figure 1.4, the process gain $K_p$ is equal to $A / B$ , with units of % process / % controller output. Setting the controller gain $K_C$ to a value less than $1/K_p$ (less than $B/A$ ) ensures stability in the loop. Understanding this can be particularly useful when it’s difficult to establish a slope $R$ for the open-loop response curve, for example, when the response is too fast to observe on available equipment. If $A/B$ can be calculated, the gain can be set to guarantee stability.

When the time constant L is long, like minutes instead of seconds, the open-loop approach to tuning may be preferred before using the closed-loop approach. This may be the case for composition and temperature loops compared to pressure, level, and flow loops. For pressure, level, and flow systems, the values of L and R are typically too fast for readily available recording equipment to establish them accurately, making the closed-loop approach the preferred tuning method.

Many complex situations can be understood by using the open-loop concept in reverse.

John Ziegler’s contribution to the open-loop tuning method is particularly notable, given that in the late 1930s, there was very little understanding of control principles and the mathematics of control in the process industries. There were only a handful of experts, who couldn’t even agree on how to tune a controller, and Ziegler wasn’t considered one of them. However, in 1940, when Taylor Instrument Co. introduced the first PID controller all in one case, they had to help their salesmen and customers learn how to tune it, and Ziegler took on the task.

Through discussions with industry experts, Ziegler discovered that “lags were bad and capacitance was good,” with capacitance referring to factors that contribute to $R$ and lags referring to factors that contribute to L. Based on this understanding, Ziegler developed the graphical open-loop method for determining controller settings. This approach involves considering the open-loop step response and how changes in $R$ and L affect the system’s stability and desired settings. The open-loop tuning rules can also be used in reverse to understand the effects of interactions and nonlinearities, which are common causes of control loop misbehavior.

If you’re new to the field of control, it’s helpful to keep the open-loop concept in mind, as it can be particularly useful for understanding complex situations and the effects of changes to $R$ and L on the system’s stability.

When Not to Tune by the Rules

There are two common times when going through the tuning procedure by-the-numbers is either not desirable or not necessary. These are level loops and flow loops.

Figure 1.6 [Averaging level control can smooth out ow fluctuations, keeping them from being passed on from one part of the process to another.]

Level Loops

Compared to other control loops such as flows, temperatures, pressures, and compositions, level control is different because it typically does not play a significant role in quality control. In fact, it is often preferred to have loose level control in order to prevent upsets in one part of the process from propagating to other parts. This approach, known as averaging level control, relies on the storage capacity of the vessel to absorb fluctuations in the inflow, as shown in Figure 1.6. By allowing the tank to smooth out these fluctuations, the impact on downstream processes is reduced.

To determine the appropriate settings for averaging level control, specific formulas are required. The first step is to determine the allowable level variance. For a surge tank, a variance of 80% may be suitable, but other situations may require different values. For example, in a distillation column, the level must be carefully controlled to avoid adverse effects on heat transfer in the reboiler or entrainment of liquid up the column.

First determine the following parameters:

$UL$ % = Allowed upper limit, % of scale

$LL$ % = Allowed lower limit, % of scale

$SP$ % = Setpoint, \% of scale

$V$ = Volume between $UL$ % and $LL$ %, $ft^{3}$

$Q$ = Maximum flow rate, $ft^{3}$ / min

Then calculate the controller settings:

$K_{C1}=100\% / (UL\%-SP\%)$

$K_{C2}=100\% / (SP\%-LL\%)$

$K_{C}$ = Smaller of $K_{C1}$ and $K_{C2}$

$T_{i}=4V / (QK_{C})$

In most cases, the rules mentioned earlier are sufficient for controlling level loops. The need to use formal tuning procedures for level control is rare. If the valve’s maximum to minimum flows span does not require the valve to travel its full range, a lower gain can be used. This allows for more surge averaging and requires a recalculation of the integral time.

These recommendations are suitable if the valve and level are closely coupled, and the dead time between valve movement and level response is short (approximately one to 10 seconds). If the dead time is long, the considerations become more complex, and it is advisable to apply the open-loop concepts to the tuning task.

It is possible to argue against using integral action for level control altogether. With digital control being more prevalent, this is now a more viable option than in the days of predominantly analogue control. For standardisation purposes, few proportional-only controllers were specified, resulting in PI controllers being used for level control. However, the integrating nature of level processes, coupled with controllers with integral action, may lead to a runaway system that could cause stability problems. This may result in a long cycle period relative to the natural period.

In rare cases, level control may fall into the category of quality-affecting factors, requiring tight control. An example of this is a polymerization vessel where a shoreline builds up a deposit of degraded polymer that should not be disturbed to prevent contamination of the product. Other examples may exist.

In summary, tuning rules are applicable for most level loops, and formal tuning procedures are seldom necessary. However, if the dead time is long or the process is integrating, open-loop concepts may need to be applied, and the use of integral action for level control may require careful consideration.

Flow

Now that you have learned about tuning loops, you understand that the tuning process is dependent on the dynamic response or time of the loop. Flow loops are generally designed in a similar way and have a similar dynamic response, which means that they can often be tuned using the same settings. As a result, the typical tuning procedure for a loop can often be bypassed for flow loops. A gain setting of 0.3 to 0.7 is typically suitable for most flow loops, and the natural periods of flow loops generally fall within the one to 10 second range, with most falling within the two to five second range. Because flow loops are typically fast compared to other processes, they do not need to be tuned to the highest degree of precision. Therefore, an integral time of 0.1 to 0.3 minutes is usually sufficient for most flow loops, and these settings will provide a good starting point.

If more precise tuning is necessary, a closed-loop approach can be used to make small setpoint changes and test for stability. However, it is important to be aware of one factor when tuning flow loops: the open-loop gain, which is how much the flow moves (as a percentage of scale) in response to a change in the controller output (as a percentage of scale). If the open-loop gain is higher at high flows, tuning that was done at low flows may become unstable at high flows.

In some cases, a flow may be controlled indirectly, meaning that the valve and flow meter are not in the same line. In these instances, the typical dynamic response and tuning settings of flow loops do not apply, and the loop should be tuned like any other loop, using either the open or closed-loop methods depending on the circumstances.

While level and flow loops are examples of loops that do not usually require tight tuning, many other individual loops only need to be stable and perform reasonably well. These loops do not need to achieve peak performance. However, if peak performance is desired, it is important to keep in mind that the speed at which one can run the mile is limited and can vary depending on individual factors.

Tuning Rules Overview

Take a moment to consider the following: processes come in an almost infinite variety, yet controllers only have two or three adjustments to accommodate them. Admittedly, the adjustments are Proportional, Integral, and Derivative (PID), but the integral and derivative adjustments are determined by the same process parameter. In essence, the goal of tuning is to match the time and amount parameters of the controller to the time and amount parameters of the process. It’s akin to fitting shoes (or other clothing) to a person, where only two measurements are used, such as length and width. Surprisingly, it often works out quite well in both cases.

The task of tuning is concerned with setting the time and amount parameters of the controller to the time and amount parameters of the process.

Computers offer the ability to create controllers that are custom-designed to fit a particular process, similar to how a tailor would create bespoke clothing for an individual. However, if the process or individual changes, the result of the customization could be worse than if a more general approach had been taken.

The time settings of a controller are tied to a time parameter of the process.

The reset, derivative, and filter times of a controller are all linked to the time parameter of the process. However, the controller gain is not uniquely tied to the amount parameter of the process, as it depends on both time and amount parameters. The controller cares only about the time and amount characteristics of the process and is indifferent to the physical variable being controlled, as well as its units. Nonetheless, processes with similar names usually have similar time and amount characteristics.

Tuning rules are designed to provide tight control, but they are based on several assumptions. Firstly, there should be no troublesome noise, which refers to undesirable variations in the measurement that are either not meaningful or too fast for the controller to respond to. Secondly, there should be no troublesome nonlinearities, dead band, velocity limiting, or other issues that may affect the control. Thirdly, there should be no troublesome interactions with other loops, and finally, the loop should have typical lags and not exhibit an inverse response or be open-loop unstable.

It is essential to note that these rules are only approximate and aim to get the controller settings in the ball park for tight control. Additionally, the calibration of analogue controllers can often be poor, with inaccurate dial markings and a considerable gap between them. Therefore, it is advisable to use the tuning rules for tight control and to distrust the dial calibrations if the performance is not up to expectations and if one knows what they are doing.

PID.02 / Tuning Rules and Procedures +