Open-Loop and Feedforward Control Systems Notes

Open-Loop and Feedforward Control Systems

Goals and Objectives

• Describe the purpose and properties of transfer functions, including their DC gain.
• Construct and simplify block diagrams involving static and dynamic elements.
• Analyze the structure and function of feedforward control systems.
• Identify limitations and practical challenges in feedforward control implementations.

Entrance

• Transfer Functions: Mathematical descriptions of system dynamics.
• Block Diagrams: Visual representations of system relationships.
• Together, they provide powerful tools for:
  • Analyzing complex dynamic systems
  • Designing effective control strategies
  • Predicting system behavior
  • Simplifying multi-component systems
• These concepts form the foundation for understanding and designing modern control systems across all engineering disciplines.

Quick Overview on Transfer Functions

• A transfer function is a mathematical representation that describes the relationship between a system's input $u(s)$ and output $y(s)$.
• It is based on the Laplace transform. It converts differential equations into algebraic equations.
• General form: $G(s) = \frac{y(s)}{u(s)}$
• DC Gain: The DC gain $K$ of a transfer function $G(s)$ is the value of the transfer function when $s = 0$, i.e., $G(0)$. This value is crucial for understanding the steady-state behavior of the system.
  • Given a transfer function $G(s) = \frac{b<em>m s^m + b</em>{m-1} s^{m-1} + … + b<em>0}{a</em>n s^n + a<em>{n-1} s^{n-1} + … + a</em>0}$, the DC gain $K$ is calculated as $K = G(0) = \frac{b<em>0}{a</em>0}$.
• Example of a 1st-order ODE and its Transfer Function:
  • Given the differential equation: $\tau \frac{dy(t)}{dt} + y(t) = Ku(t)$
  • Applying the Laplace transform: $\tau s y(s) + y(s) = Ku(s)$
  • The transfer function $G(s)$ is then: $G(s) = \frac{y(s)}{u(s)} = \frac{K}{\tau s + 1}$

Advantages of Transfer Functions

• Provide a simple representation of process dynamics.
• Make signal manipulation and analysis easier (compared to ODEs).
• Create direct relationships between inputs and outputs.
• Allow the use of multiplication to represent connections between components.
• Enable block diagram representation for complex systems.

Block Diagrams

Introduction

• Block diagrams represent the flow of information in a system.
• Blocks represent systems or components with transfer functions.
• Signals are represented by arrows connecting blocks.
• Block diagrams provide a visual representation of mathematical relationships.
• Each block could contain a static gain, a simple dynamic function, or even a differential equation.

Block Diagrams with Static Gains

• Each block in a block diagram represents a transfer function.

• These transfer functions may simply be constant values, called "static gains":

• $y(s) = Ku(s)$

• The output $y(s)$ is simply a scaled version of the input $u(s)$, and $K$ is indeed referred to as the static gain because it represents a fixed proportionality between input and output with no frequency or time dependency.

• The static gain $K$ has a physical unit:

  • $K \text{ [unit]} = \frac{y(s) \text{ [unit]}}{u(s) \text{ [unit]}}$

• Examples of static gains and their units:

  • If $u(s)$ is in volts (V) and $y(s)$ is in meters (m), then $K$ is in meters per volt (m/V).
  • If $u(s)$ is in degrees Celsius (°C) and $y(s)$ is in pascals (Pa), then $K$ is in pascals per degree Celsius (Pa/°C).
  • If $u(s)$ is in liters per minute (L/min) and $y(s)$ is in liters (L), then $K$ is in minutes (min).
  • If $u(s)$ is in amperes (A) and $y(s)$ is in radians per second (rad/s), then $K$ is in radians per second per ampere (rad/s/A).
  • If $u(s)$ is in Newtons (N) and $y(s)$ is in meters per second squared (m/s²), then $K$ is in meters per second squared per Newton (m/s²/N).
  • If $u(s)$ is in watts (W) and $y(s)$ is in degrees Celsius (°C), then $K$ is in degrees Celsius per watt (°C/W).
  • If $u(s)$ is in kilometers per hour (km/h) and $y(s)$ is in kilometers (km), then $K$ is in hours (h).

• The output response $y(t)$ to a step input of amplitude $A$ is $y(t) = K \cdot A \cdot (1 - e^{-t/\tau})$

Linearity and Superposition Principle

• A system is linear if it satisfies the following two properties:
  1. Additivity (Law of Superposition): If $H(u<em>1) = y</em>1(t)$ and $H(u<em>2) = y</em>2(t)$, then $H(u<em>1 + u</em>2) = H(u<em>1) + H(u</em>2)$.
  2. Homogeneity (Scaling Property): If $H(u) = y(t)$, then $H(\alpha u) = \alpha y(t)$ for any scalar $\alpha$.
• Superposition Principle: For a linear system, the response to a sum of multiple inputs is the sum of the responses to each input acting alone.
  • Mathematically, for inputs $u<em>1, u</em>2, …, u<em>n$, the output is: $H{\sum</em>{i=1}^{n} \alpha<em>i u</em>i} = \sum<em>{i=1}^{n} \alpha</em>i H{u<em>i}$ where $\alpha</em>i$ are constants.

Time Invariance

• A system is time-invariant if a time shift in the input results in an identical time shift in the output. Mathematically, if $y(t) = H{u(t)}$, then for any time shift $\tau$, $y(t - \tau) = H{u(t - \tau)}$.

Feedforward Control System

Introduction

• A control system consists of subsystems and processes assembled to achieve a desired output with specified performance.
• Open-loop control system: operates without feedback (i.e., no output correction), with a feedforward controller as its compensator.
• Closed-loop control system: uses sensors to feed the output back for correction, with a feedback controller as its compensator.
• Performance is measured based on: transient response (rise time, overshoot, settling time, etc.) and steady-state response (steady-state error).

What is the Feedforward Control System?

• A feedforward controller is a proactive control strategy placed before the plant transfer function in a series connection.
• It anticipates the input and adjusts the control signal $u(s)$ to enhance the output response $y(s)$.
• Key advantage: Improves system performance by acting preemptively, unlike feedback which reacts to errors.

Feedforward Block Design

• Static Gain: A simple constant (e.g., $K_f$) can reduce steady-state error by scaling the input to match the plant's inverse gain.
• Dynamic Block: Incorporates dynamics (e.g., a lead-lag compensator) to enhance transient response, reducing settling time and oscillations.

Feedforward Controller as a Static Gain

• $K_f = \frac{1}{G(0)} = G^{-1}(0)$
• Optimal Static Gain for Feedforward Controller: For an ideal steady-state system, the step response reaches the set-point value when the controller's proportional gain equals the reciprocal of the system's DC gain.

Feedforward Controller as a Dynamic Transfer Function

• For the transient response, the feedforward controller must be upgraded from being just a static gain $K_f$ to being a dynamic transfer function $C(s)$.
• Feedforward control is model-based and focuses on preemptively compensating for measurable disturbances or set-point changes by leveraging knowledge of the system's dynamics.
• Feedforward controllers often use lead-lag compensators to adjust the timing and magnitude of the control action.
• If the system has a known time constant or lag, the feedforward can include a lead component to preemptively cancel that lag.
• General form of a Lead-Lag Compensator: $C(s) = K<em>f \frac{s + z</em>1}{s + p<em>1} \cdot \frac{s + z</em>2}{s + p_2} \cdot …$
  • Lead compensator: $\frac{s+z}{s+p}$ where z < p
  • Lag compensator: $\frac{s+z}{s+p}$ where z > p

Limitations and Challenges of Feedforward Control

Feedforward Control Scheme Recap

• Enhancing Steady-State Response: Inverse of the DC gain is used to correct the steady-state error.
• Enhancing Transient Response: A lead-lag compensator is implemented.
• Although effective in theory, practical applications often reveal several challenges.

Introduction to Practical Limitations of Feedforward Control

• While feedforward control improves steady-state accuracy and transient response, real-world imperfections introduce challenges.

Overview

• Parameter Uncertainty: Model inaccuracies during design caused by different parameters (e.g., inertia, resistance, damping, etc) than what's modeled.
• Disturbances: Unanticipated external influences, like changes in load or noise.
• Aging Effects: Gradual drift in system behavior, for example, due to wear, material fatigue, chemical depletion of components.

Parameter Uncertainty

• Models are approximations; actual system parameters may differ.
• Impact:
  • Steady-State Error: Incorrect DC gain inverse leads to residual error.
    • Example: A motor's torque constant $K_t$ is modeled as 0.1 Nm/A, but the real value is 0.09 Nm/A → 10% steady-state error.
  • Transient Degradation: Lead-lag compensators become misaligned with system dynamics.
    • Example: Compensator zeros/poles placed incorrectly due to wrong time constants → Oscillations or sluggish response.
• Mitigation: Robust design (e.g., H∞), adaptive control, or hybrid feedback-feedforward architectures.

Disturbances

• Feedforward control cannot counteract unmodeled or unmeasured disturbances (e.g., load changes, noise).
• Impact: Persistent errors due to lack of corrective action.
  • Example: A conveyor belt's feedforward speed controller cannot adjust for sudden added weight.
• Types of Disturbances:
  • Step: Sudden load changes.
  • Periodic: Vibrations from rotating machinery.
  • Stochastic: Sensor noise.
• Mitigation: Integrate feedback control (e.g., PID) for disturbance rejection. Use disturbance observers or feedforward sensors (e.g., measuring wind gusts in drones).

Aging Effects

• Systems degrade over time (e.g., wear, material fatigue, chemical depletion).
• Impact:
  • DC Gain Drift: Aging alters steady-state behavior.
    • Example: Battery internal resistance increases → reduced actuator output.
  • Dynamic Shift: Time constants change (e.g., lubricant degradation increases friction).
    • Example: Lead-lag compensators designed for "new" systems fail to dampen aged resonances.
• Mitigation: Scheduled re-calibration or re-identification of system parameters. Adaptive control with online parameter estimation (e.g., recursive least squares).