Introduction to Feedforward Control

Summary

In this educational screencast, the focus is on explaining feedforward control and its importance in control systems, particularly when feedback control falls short. The video discusses the limitations of feedback control, such as its inability to prevent disturbances before they affect the process and the challenge of dealing with slow or dead-time processes. Feedforward control is introduced as a solution, allowing for preemptive action on disturbances by using models that predict how changes affect the system. The video also highlights the need for measurable variables and reliable models for effective feedforward control, and explains how it works alongside feedback control to maintain desired outcomes. Ultimately, feedforward control is depicted as beneficial when used with feedback control but less effective on its own due to its reliance on model accuracy and lack of real-time variable measurement.

Highlights

• Feedback control acts only after disturbances affect the process, necessitating feedforward control. ⏳
• Feedforward control anticipates changes and adjusts processes accordingly, preventing disturbances. 🚀
• Reliable models and measurable variables are crucial for successful feedforward control. 📉
• Combining feedback with feedforward offers superior control versus using either alone. 👩‍🔧
• The importance of model accuracy in feedforward control cannot be overstated. 🎯
• Choosing control schemes depends on the type of disturbances and existing system capabilities. 🎛️

Key Takeaways

• Feedback control often struggles with preemptively managing disturbances, making feedforward control a useful complementary tool. ⚙️
• Feedforward control can adjust for disturbances before they affect the controlled variable by using predictive models. 🔮
• Effective feedforward control requires variable measurability and reliable predictive models. 📏
• The synergy of feedback and feedforward control can enhance process stability and accuracy. 🔄
• Feedforward control alone is less reliable without feedback due to its dependency on model accuracy. 📊
• Feedforward control isn't necessary if disturbances are minimal and manageable through existing feedback systems. 🤔

Overview

Have you ever tried to fix a problem after it happens, only to find it’s too late to avoid the disruption? That's exactly where feedback control falls short! This screencast dives into the fascinating world of feedforward control, a predictive hero stepping up where feedback alone isn’t enough. By anticipating disturbances and adjusting processes before they occur, feedforward control becomes a game changer. 🚀

Imagine a blending tank scenario where you're frantically juggling flow rates to maintain the perfect composition. With feedforward control, there's no juggling needed! This proactive approach sends signals to adjust the flow in advance, stopping disturbances in their tracks. But remember, to pull off this magic, you need variables you can actually measure and trust in your predictive models. 🔮

While feedforward control is like a crystal ball in process management, it works best side-by-side with feedback control. Feedforward alone can’t measure real-time outcomes, meaning it won't know if it's a bit off the mark. That's why blending both methods brings the balance of foresight and adjustment, making sure everything runs like a well-oiled machine! 🤖

Chapters

• 00:00 - 00:30: Introduction to Feedforward Control Concepts This chapter introduces the concept of feedforward control in the context of control systems. It begins by emphasizing the importance of understanding the limitations of feedback control systems, as most control processes rely solely on feedback mechanisms. The chapter sets the stage for discussing the scenarios where feedforward control becomes essential by first addressing the shortcomings of feedback control.
• 00:30 - 01:00: Limitations of Feedback Control In the chapter titled 'Limitations of Feedback Control,' the discussion focuses on the reasons why perfect control is unattainable using feedback mechanisms. Three primary limitations are explored, with particular emphasis on how feedback control cannot act until a disturbance affects the controlled variable, rendering preemptive correction impossible.
• 01:30 - 02:30: Feedforward Control and Blending Tank Example The chapter discusses the challenges and limitations of feedforward control in managing disturbances within a system, particularly highlighting the issue with processes that have significant dead times. It explains that because the controlled variable is only measured at the process's endpoint, predicting its future state is difficult, making it hard to effectively counteract disturbances quickly.
• 03:00 - 04:30: Importance of Variable Variation and Modeling in Feedforward Control This chapter discusses the importance of acknowledging the variation of variables and modeling in feedforward control, particularly in processes that are slow to respond. Feedforward control addresses the drawbacks of delayed reaction to changes in process conditions by anticipating and adjusting for them rather than waiting for feedback. The chapter illustrates this with a simple example of controlling the output concentration in a blending tank, where two streams are involved. It contrasts this with a feedback-only control scheme, highlighting how feedforward control can more effectively manage the process variables.
• 05:00 - 06:30: The Feedback and Feedforward Control Scheme The chapter discusses the development of a control scheme by integrating feedback and feedforward control methods. It addresses the issue of flow rate fluctuation in a process, which can disturb the outlet composition control. By incorporating both feedback and feedforward controllers, the scheme aims to detect and compensate for potential disturbances, ensuring better control over the composition despite changes in flow rates.
• 07:30 - 09:00: Role of Feedforward and Feedback Synergy in Control Schemes The chapter explains the concept of feedforward and feedback synergy in control schemes, focusing on how these methods enhance the ability to manage process flows. It highlights that a feedforward controller can proactively adjust the manipulated stream's flowrate by preemptively signaling changes to valves before they occur. This preemptive action allows for better control and management of flow changes in the system.

Introduction to Feedforward Control Transcription

• 00:00 - 00:30 In this screencast we will talk about feedforward control and its role in developing appropriate control schemes. So before we can really talk about where feedforward control can be used we have to first focus on where feedback control fails. Because a large majority of control processes are of a pure feedback only control scheme, so therefore in order to deviate from
• 00:30 - 01:00 those schemes there have to be reasons why, and there are a number of reasons why, but here we'll focus on three of them. The first is that perfect control cannot occur. A disturbance happens, but the controlled variable does not feel it, so in other words a disturbance to the variable will happen, but the control process will already handle it. Because feedback control does not work until there is a deviation to the controlled variable, it is impossible
• 01:00 - 01:30 for a disturbance to have no effect on this variable, which connects number two, because you are only measuring the controlled variable at the end of the process, there is no way to predict what's going to happen to this variable. An additional problem is that processes are slow or have large dead time can lead to control, which can take a while. Why? Because if a system has a very large dead time, for example, a disturbance will occur, the process
• 01:30 - 02:00 won't feel it for a while, but once it feels it, it's going to take a while to actually act on it because of the slowness of the process. So feedforward control is a way to handle all three of these drawbacks. So to try to represent this we are going to look at a somewhat rudimentary example of a blending tank, and in our blending tank we are trying to control the output concentration, and we have two streams, so therefore an example of a feedback only scheme would be one where we would try to directly control the composition coming
• 02:00 - 02:30 out by the flow rate of the first one. However, one example here would be let's say that the flow rate of the first scheme is known to fluctuate a bit, so therefore if that is a potential disturbance our process won't recognize that disturbance, and therefore if we see a change in this flow rate that will then cause a change in the composition, which isn't necessarily good when we are trying to control this outlet composition. So what we can do is develop a control scheme where we combine this feedback control with a feedforward controller,
• 02:30 - 03:00 which would be able to better control this process. Why? Well the reason why is that if I see a change in the flow what I can do is have this feedforward controller send a direct signal to the valve saying that a change is going to happen, therefore the flowrate of the manipulated stream can be adjusted in advance notice of the fact that this flow is going to be changing. So now what we need to do is discuss some of the things about
• 03:00 - 03:30 feedforward control. So the first is that the variable, or variables, being used for feed forward control must actually vary to some degree, because if they are not fluctuating at all there would be no point to have a controller for it, and if they're only fluctuating in a small amount it might not be worth it to introduce additional equipment and what not, which would introduce costs, and also potentials for equipment for breaking and what not, so
• 03:30 - 04:00 therefore you want to make sure these variables actually can vary in some way. The second part of interest here is that a model must exist that can relate the feed forward variable or variables to the controlled variable. So the beauty of what feedforward control does is if I know the fact that by changing my flow rate by 1 gallon per minute of this stream will cause my composition to go down by 0.02, let's say mole fraction. What I can do is
• 04:00 - 04:30 I know how to adjust the other stream, the second stream where the valve is, the manipulated variable, in order to adjust for that change, so therefore you must be able to have a model with some reasonable accuracy, otherwise if you don't know how the variables are related you're really not going to gain anything from this, so therefore having a model is important. The third obviously is that the variable, or variables of interest must be measurable. Depending
• 04:30 - 05:00 on the process this may not be as easy as it sounds, particularly if you're in a process with high pressure, high temperature, highly corrosive materials, etc., it may not be as easy to measure these, but in order for feed forward control to properly work you need to actually measure these variables to see how much they are actually fluctuating from their assumed steady state value. So this process here represents a feedback/feedforward controller, because the part in blue is indicative of the feed forward process, where as the part of
• 05:00 - 05:30 the control scheme in black represents the feedback processing. With the feedforward control scheme the variable of interest, the controlled variable, is not measured. So therefore this feedforward controller here is conducting analysis without knowing whether or not the composition here is too high, too low, or exactly where it should be, so that's kind of an important aspect here. By putting these together the question is from a block diagram
• 05:30 - 06:00 standpoint what does this look like, and as a start, we'll start with the part in black which is the feedback scheme, and really the feedback scheme is no different than what we saw before where we had the transmitter, we'd have a set point with an appropriate gain. It would then go to the controller, the valve, the process, which here is the tank, and then out would come the composition. So therefore this part has been maintained by this process. However, the question to ask ourselves is how does the feedforward
• 06:00 - 06:30 controller fit in, and the way we can gain some insight on this is kind of looking at the diagram. What we notice here is that right before the valve instead of it receiving one controller signal it now receives two. Both sum together, and the other one is the feedforward controller. The feedforward controller gets the signal from the flow transmitter, which means the fact the flow transmitter has to take into account the measurement of the flow
• 06:30 - 07:00 rate of that stream, which we'll call here F1. By adjusting F1 I am also adjusting the composition, so therefore there is a second tank transfer function here that represents the relationship between F1, the disturbance variable, and the composition coming out of the tank, so therefore we have another block which shows that relationship, so therefore the composition here changes because of two things, the change of the flow rate of the
• 07:00 - 07:30 first stream, and the flow rate of the second stream, which is being manipulated by the valve. This relationship would come out through a material balance. Now that we have developed our feedback and feedforward control scheme in a block diagram, and therefore we could conduct a good amount of analysis or put this into Simulink to kind of see how the processes act together. So to leave this we'll have a few parting shots here of well why not have just a feedforward only control scheme, and in general feedforward only control schemes
• 07:30 - 08:00 are not very beneficial by themselves, the major reason why is something that was noted earlier is that there is no measurement of the controlled variable, so therefore the feedforward controller is acting the same way regardless of whether or not your desired variable is way too high to little too high, way too low, a little too low, or just right, and there's no way for the feedforward part to know. Therefore feedforward often works
• 08:00 - 08:30 very well with feedback, because feedback is measuring that value and just making sure that your scheme is online with what you're expecting it too be. Another issue with feedforward only is it's very model dependent, so if the model that you have is not very strong, or is not a great fit to the process then that means that you're going to have some problems when it comes to actually handling the control of this process. Again feedback control can handle that because it measures the actual value and does comparisons, so therefore if
• 08:30 - 09:00 the model is slightly off it can adjust the manipulated variable to counteract that. And a final point, freedforward is not needed if the disturbance is minimal, or has minimal effect. You'll just be introducing more softward/hardware to your process, which could be prone to breaking, could more trouble than it's worth, particularly when you have a control scheme already present from the feedback standpoint. So in this screencast we took a look at the basic idea of feedforward control, why a process may benefit from feedforward control, and what it looks like from a block
• 09:00 - 09:30 diagram standpoint.