4/10 AUTOMATED INTERACTION BETWEEN A DT WITH REAL ASSET Flashcards
Examples of automated interaction, explain, remember the loop of how its the interaction between DT and Physical model
What are the different kind of controls?
RULE BASED CONTROL:
Based on classical logic
- If-then-else, yes, no, store, alarm, …
- First application of Programmable
Logic Controllers (PLC)
HYSTERESIS CONTROL
see image.
What is the PID controller
Feedback loop
- PID acts on tracking error
- P = Proportional action
- I = Integral action
- D = Differential action
- I takes care of steady state error
- D can cause instability with noise
- Great for Linear Time-Invariant systems!
How to tune a PID controller
What are the advantages of the DT
How to tune a PID controller?
- Manual & intuitive with step response shaping
- Tune for desired bandwidth and phase margin
- Ziegler-Nichols, Astrom-Hagglund, Kaiser-Rajka
- Root locus method
- Automatic tuners
- Frequency domain: Bode, Nyquist & Nichols
****
You can upgrade the model on the simulation software and then transfer the data to the physical system
What is cascade controller, where you use it
Each feedback loop will have different speed (Hz)–here you cascade the execution time because other wise the feedback lope will conflict with each other
**
Three step cascaded control: Position (P), Speed (PI) & Torque (PI+FOC)
- Discrete execution with increasing step size & bandwidth for each loop
- Intermediate saturation & rate limits: Accounts for physical limits, but anti-windup needed!
- Intermediate feedforwards to help control with system-knowledge
- Sometimes pre-filter to smooth position setpoint
How does a Model Predictive Control Acts (14)
“MPC uses a model to interact with the
physical system, so is it a Digital Twin?”–>Inverse digital shadow, explain why?
A linear cost function are easy to solve!! because are convex
What are the cost function?
Uses a model of the system dynamics
- Can handle MIMO systems
- Determine ideal control action by
minimizing a quadratic cost function J
J is convex, so has absolute minimum
- Control horizon: How many time steps
do you apply control input?
- Prediction horizon: How many time
steps do you predict in the future?
weighting factors!! –You tell the controller what you find important
What is Model predictive control (15).
What if your system is no linear?, how to linearize it
Linear MPC
System, constraints and cost function are linear, so solving is fast
Adaptive MPC
System not linear. Linearize around current operating point,
update the model, and then use linear MPC
Gain scheduled MPC
System not linear. Constraints, control- or prediction horizon vary, but linearly. Switch between several linear MPC’s
Explicit MPC
Solve cost function offline and interpolate with linear fit (faster)
Non-linear MPC
If system, constrains and/or cost function are non-linear. Do full blown non-linear optimization (difficult, due to local minima)
What are the Motion profiles 17
**= The input of the control for motion applications
Point to point motion
̶ Only end point is important
̶ Setpoint can be simple step function
̶ Overshoot is often undesired
Path motion
- A desired path must be tracked
- Path = Series of discrete points in space
- Interpolation: Linear or spline
What is Adaptive Control, how does Classical control techniques work great for time-invariant systems 21
Classical control techniques work great for time-invariant systems
̶ However, many systems have complex time-varying parameters and dynamics!
How to create or adapt a control that is working to systems that are complex time-varying parameters and dynamics
*Choose the worst case condition
Solution 1: Robust control
̶ Make control sufficiently robust in tuning
̶ Stable response even if system dynamics deviate
̶ But you lose optimality …
Solution 2: Adaptive control
- Adapt control based on model info
- Model: Digital Twin!
22
- Gain scheduling
- Adaptive feedforwards
- Model Reference Adaptive Control (MRAC)
- Model Identification Adaptive Control (MIAC)
- Learning control
What is Gain scheduling 23
What is adaptive cascaded control 24
Explain MPC and relation with Digital Twin 25
Conclusions. For the exam, specific application and ask which kind of controller do you use and how you will use the DT for a condition
