Lecture 8 Paper
๐ฏ 1. Big Picture: What is PID Tuning?
PID controllers are the most widely used controllers in industry (over 50%!)
๐ Goal: Make a system behave the way you want by minimizing error between:
- Setpoint (target)
- Process value (actual output)
๐ก Controller tuning = choosing the right parameters so the system:
- responds fast โก
- is stable ๐ง
- doesnโt overshoot too much ๐ฏ
๐ง 2. The PID Equation (Core Idea)
The controller combines three actions:
- P (Proportional) โ reacts to current error
- I (Integral) โ reacts to accumulated error over time
- D (Derivative) โ predicts future error
๐ Variables (from page 2):
- ( u(t) ): control signal
- ( ฮต(t) ): error
- ( K_c ): gain
- ( ฯ_i ): integral scaling
- ( ฯ_d ): derivative scaling
- ( b ): bias (offset)
โ๏ธ 3. Understanding Gain (SUPER IMPORTANT)
๐ฝ Too low gain:
- System responds slowly
- Oscillations die out (damped)
๐ Seen in the graph on page 2:
- oscillations shrink over time โ stable but sluggish
๐ผ Too high gain:
- System becomes unstable
- Oscillations grow larger and larger
๐ Seen in the graph on page 3:
- oscillations increase โ system โblows upโ
๐ฏ Perfect tuning:
- Oscillations are steady and consistent
This gives: ๐ Ultimate gain (Ku) ๐ Ultimate period (Pu)
These are key for tuning methods.
๐ 4. Open vs Closed Loop Systems
๐ Closed-loop (feedback):
- Output is compared to input
- Used in most real systems
๐ Open-loop:
- No feedback
- Just observe system response
๐ Important difference:
- Ziegler-Nichols โ works for both
- Cohen-Coon โ mainly open-loop
๐งช 5. Trial-and-Error Method (Simple but crude)
๐ก Idea:
โTry values โ see what happens โ adjustโ
Steps:
- Start with P control
- Increase (K_c)
- Add I and D to refine
๐ Typical values (VERY exam-relevant)
๐ Flow:
- Use P or PI
- Avoid D (too noisy!)
- (K_c = 0.4โ0.65), (T_i = 6s)
๐ข๏ธ Level:
- PI preferred (reduces offset)
- Example:
- (K_c = 2)
- Bias = 50%
- PI:
- (K_c = 2โ20)
- (T_i = 1โ5) min
๐จ Pressure:
- Depends on medium:
| Type | Kc | Ti |
|---|---|---|
| Liquid | 0.5โ2 | 6โ15 s |
| Gas | 2โ10 | 2โ10 min |
๐ก๏ธ Temperature:
- Use full PID (slow system!)
- (K_c = 2โ10), (T_i = 2โ10) min, (T_d = 0โ5) min
๐ 6. Process Reaction Curve Method (Open-loop)
This is VERY important conceptually
๐งช What you do:
- Let system reach steady state
- Apply a step disturbance
- Record response
๐ What you measure:
From curve (page 6 diagram):
- โฑ๏ธ Dead time (ฯ_dead) โ delay before system reacts
- โณ Time constant (ฯ) โ how fast it responds
- ๐ Final value (Mu) โ steady-state value
- ๐ Step size (Xo)
๐ The tangent line at inflection point (shown in diagram page 6) is used to estimate these.
๐ Key ratios:
- ( R = ฯ_ / ฯ )
- Used to calculate controller parameters
๐ฏ Goal:
Convert these values into:
- (K_c), (T_i), (T_d)
๐ฅ 7. ZieglerโNichols Method (MOST IMPORTANT)
Two versions:
๐ A. Closed-loop Method
๐งช Steps:
- Turn OFF I and D
- Increase (K_c)
- Wait until oscillations are:
- constant amplitude
๐ This gives:
- Ultimate gain (Ku)
- Ultimate period (Pu)
๐ Visual (page 7):
- Increasing gain โ oscillations stabilize
- Measure oscillation period
๐งฎ Then:
Plug into formulas:
| Controller | Kc | Ti | Td |
|---|---|---|---|
| P | Ku/2 | โ | โ |
| PI | Ku/2.2 | Pu/1.2 | โ |
| PID | Ku/1.7 | Pu/2 | Pu/8 |
โ Advantages:
- Simple
- Uses real system dynamics
โ Disadvantages:
- Can push system into instability
- Time-consuming
๐ B. Open-loop Method (Process Reaction)
๐งช Steps:
- Apply step input
- Measure:
- (ฯ_), (ฯ), (Mu)
- Use formulas to compute PID parameters
โ Advantages:
- Faster
- Safer (no instability testing)
โ Disadvantages:
- Less accurate
- Based on approximations
๐ 8. Quarter Decay Ratio (QDR)
Used mainly in open-loop tuning
๐ฏ Definition:
Amplitude ratio = 4:1
๐ First peak is 4ร larger than second peak
๐ก Meaning:
- System oscillations decay nicely
- Good balance between:
- speed โก
- stability ๐ง
โ๏ธ 9. CohenโCoon Method (Alternative)
Mentioned but not deeply detailed here.
๐ Key idea:
- Another open-loop tuning method
- Uses process reaction curve
- Often more accurate than Z-N for some systems
๐ง Final Mental Model (IMPORTANT)
Think of PID tuning like this:
| Component | Role |
|---|---|
| P | Reacts NOW |
| I | Fixes past errors |
| D | Predicts future |
| Method | When to use |
|---|---|
| Trial & Error | Quick/simple systems |
| Ziegler-Nichols | No model, need fast tuning |
| Process Reaction | Safer open-loop systems |
| Cohen-Coon | More refined open-loop |
๐ Key Takeaways
- PID tuning = balancing speed vs stability
- Too low gain โ slow ๐ข
- Too high gain โ unstable ๐ฅ
- Ziegler-Nichols is the most important classical method
- Process reaction curve is essential for open-loop understanding
- Real systems often require experimental tuning