CAEECentre for Automotive Embedded Engineering · HSR Layout, Bengaluru
Module 18 of 20 · Advanced

Optimisation and Constrained Engineering Design

Optimisation chooses design variables that minimise cost or maximise performance while respecting physical constraints. It supports controller tuning, power dispatch and circuit sizing.

Advanced80 min study timePrerequisite: Differential and multivariable calculus
Module learning progress0 of 3 stages

1 · Why this mathematics matters

Tune controllers and engineering designs while respecting physical constraints.

2 · Prerequisites and learning approach

Prerequisites: Differential and multivariable calculus. Begin with the physical meaning, identify every symbol and unit, study the fixed-scale graph, then solve and verify the examples.

3 · Symbols and units

f(x)objective function
∇fobjective gradient
αstep size
g(x)constraint function

4 · Concepts and textbook formulas

1. Objective and constraints

A design problem separates what should be improved from what must remain feasible.

Objective and constraints equation

How to read it: identify the output quantity first, then the variables, parameters, units and assumptions. Test zero, unity and limiting values before using the equation in a design.

2. Gradient descent

The negative gradient gives a local downhill direction. Step size must balance speed and stability.

Gradient descent equation

How to read it: identify the output quantity first, then the variables, parameters, units and assumptions. Test zero, unity and limiting values before using the equation in a design.

3. Convexity

For a convex objective over a convex feasible set, every local minimum is global.

Convexity equation

How to read it: identify the output quantity first, then the variables, parameters, units and assumptions. Test zero, unity and limiting values before using the equation in a design.

4. Lagrange multipliers

Equality constraints are included using multipliers; stationary points satisfy gradient conditions.

Lagrange multipliers equation

How to read it: identify the output quantity first, then the variables, parameters, units and assumptions. Test zero, unity and limiting values before using the equation in a design.

5 · Step-by-step worked examples

Example 1: Foundation

Minimising f(x)=(x-3)^2 gives df/dx=2(x-3)=0, so x=3.

Verification: check the unit, sign, order of magnitude and physical trend.

Example 2: Two-variable minimum

Given: Minimise (x-1) squared plus (y+2) squared.

Two-variable minimum solution equation

Engineering meaning: The stationary point is the global minimum because the Hessian is positive definite.

Example 3: Equality constraint

Given: Minimise x squared plus y squared subject to x+y=1.

Equality constraint solution equation

Engineering meaning: The multiplier enforces the constraint while locating the least-norm point.

6 · EE/ECE applications

  • analog sizing — use the governing equation to predict parameter sensitivity, confirm a simulation and interpret measured behaviour.
  • optimal power flow — use the governing equation to predict parameter sensitivity, confirm a simulation and interpret measured behaviour.
  • filter design — use the governing equation to predict parameter sensitivity, confirm a simulation and interpret measured behaviour.
  • control tuning — use the governing equation to predict parameter sensitivity, confirm a simulation and interpret measured behaviour.

7 · Common mistakes

  • Finding an unconstrained minimum that violates the design limits.
  • Using a step size that causes divergence or oscillation.

8 · Practice problems

  1. Take two gradient-descent steps on a quadratic objective.
  2. Formulate an engineering design with one objective and two constraints.
Verification habit: predict the direction of change before calculating, then check units, sign, limiting cases and engineering reasonableness.