Math for ME · Chapter 9 of 19 · Intermediate
Matrices and Systems of Linear Equations
Real structures produce dozens of equations at once. Matrices are how engineers organize them; elimination is how computers and humans solve them.
The thread: Engineering rarely hands you one equation. A truss or a circuit gives dozens at once, and matrices are how you hold them and solve them together.
Readiness check
From earlier chapters. Tick only what you can do closed-notes.
- Solve two equations with two unknowns by substitution or elimination.
- Keep subscripted symbols (a₁₂, x₃) organized on paper.
- Write a particle equilibrium as two component equations (Vectors or Statics).
- Multiply and add fractions and decimals reliably.
- Check a solution by substituting it back.
The core idea
Ax = b: a whole system of equations as one object, solved by elimination.
Ax = bx = A⁻¹b (concept, not method)Gaussian elimination subtracts multiples of rows to create a triangle of zeros, then back-substitutes. The determinant answers one question only: is the system solvable uniquely (det ≠ 0)? Rank counts the independent equations you actually have.
The skills, taught in order
9.1 Matrices and the one multiplication rule
A matrix is a grid of numbers that stores a whole system at once. Addition is entry by entry, but multiplication is row times column: each entry of AB is one row of A dotted with one column of B. Because of that, order matters and AB ≠ BA in general. A linear system packs into a single object:
Ax = bA holds the geometry and properties, x holds the unknowns, and b holds the loads or sources.
9.2 Gaussian elimination: the universal solver
Elimination subtracts multiples of one row from others to place zeros below the diagonal, producing an upper triangle; back-substitution then solves from the bottom row upward. Three row operations are legal and leave the solution unchanged: swap two rows, scale a row, or add a multiple of one row to another. Every solver, by hand or inside FEA software, runs this same process.
9.3 Determinant and inverse: solvability
The determinant answers one question: does a unique solution exist? For a 2×2,
det [a b; c d] = ad − bc[a b; c d]⁻¹ = (1/det)[d −b; −c a]If det ≠ 0 the inverse exists and the solution is unique; if det = 0 the inverse does not exist and the system is redundant or contradictory. The inverse is mainly a tool for reasoning; in practice you eliminate rather than invert.
9.4 Rank and the three outcomes
Rank counts the genuinely independent equations. Comparing it with the number of unknowns tells you which of three things happened:
| Elimination ends in | Meaning |
|---|---|
| a pivot in every unknown | one unique solution |
| a row reading 0 = 0 | redundant equation: infinitely many solutions |
| a row reading 0 = nonzero | contradiction: no solution |
In statics language, these are properly constrained, under-constrained, and over-constrained.
9.5 Why engineers live in Ax = b
Equilibrium at the joints of a truss, currents in a circuit, springs in a network, and the stiffness equations of finite elements are all linear systems. A six-joint truss already gives twelve equations, well past hand substitution, which is why the matrix method is the one that scales.
Engineering connection: Structures, FEA, Numerical Methods, Controls, data fitting.
Worked example: the traffic light, this time as Ax = b
Statics Module 3 solved a 20 kg light on two cables (30° and 45°) by substitution. Write that same equilibrium as a matrix system and solve it by Gaussian elimination.
- ProblemSolve the system in Figure 1 by elimination.
- Given / findA = [−0.866, 0.707; 0.500, 0.707], b = (0, 196.2) N. Find x = (T₁, T₂).
- AssumptionsThe equilibrium equations are linear in the tensions, which they always are.
- Modelcos 30° = 0.866, cos 45° = sin 45° = 0.707, sin 30° = 0.5. Row 1 is ΣFx, row 2 is ΣFy.
- Equations−0.866 T₁ + 0.707 T₂ = 0 0.500 T₁ + 0.707 T₂ = 196.2
- SolveSubtract row 1 from row 2: (0.500 + 0.866) T₁ = 196.2, so 1.366 T₁ = 196.2 and T₁ = 143.6 N. Back-substitute: T₂ = 0.866(143.6)/0.707 = 175.9 N.
- Checkdet A = (−0.866)(0.707) − (0.707)(0.500) = −0.966 ≠ 0: a unique solution exists. The answers match Statics Module 3 exactly.
- ConclusionSubstitution and elimination are the same mathematics, but elimination scales: a 6-joint truss gives 12 equations, and only the matrix method survives that. FEA assembles and solves systems like this with millions of rows.
Worked example 2: solve a circuit with the matrix inverse
Two mesh currents in a resistor network satisfy 3I₁ − I₂ = 5 and −I₁ + 2I₂ = 3 (in consistent volt and amp units). Solve for the currents using the 2×2 inverse, then verify.
- Given / findA = [3, −1; −1, 2], b = (5, 3). Find x = (I₁, I₂).
- Determinantdet A = (3)(2) − (−1)(−1) = 6 − 1 = 5. It is nonzero, so a unique solution exists.
- InverseA⁻¹ = (1/5)[2, 1; 1, 3], using the swap-and-negate rule for a 2×2.
- Multiplyx = A⁻¹b: I₁ = (1/5)(2·5 + 1·3) = 13/5 = 2.6 A; I₂ = (1/5)(1·5 + 3·3) = 14/5 = 2.8 A.
- CheckSubstitute back: 3(2.6) − 2.8 = 5 and −2.6 + 2(2.8) = 3. Both original equations hold.
- ConclusionFor a 2×2 the inverse is quick and instructive, and it shows plainly why det = 0 breaks everything: the 1/det factor would divide by zero. For larger systems, elimination replaces this.
Misconceptions and diagnostics
| Mistake | Symptom | Diagnostic question | Correction |
|---|---|---|---|
| Multiplying matrices entry by entry | AB computed like addition | "Row times column: did I sum the products?" | Each result entry is one row dotted with one column. Practice until automatic. |
| Assuming AB = BA | Factors swapped freely mid-derivation | "Did I just commute two matrices?" | Matrix order encodes operation order. Swapping is illegal unless proven safe. |
| Dividing by a matrix | "x = b/A" written on paper | "What operation does the inverse actually perform?" | Multiply by A⁻¹ (when it exists), or better, eliminate. Division does not exist. |
| Ignoring det = 0 | Elimination produces 0 = 0 or 0 = 5 and panic follows | "Are my equations independent and consistent?" | 0 = 0 means redundancy (infinite solutions); 0 = nonzero means contradiction (none). Both are answers. |
Practice ladder
Solve by elimination: 2x + y = 5 and x + 3y = 10.
Show answer
R2 − ½R1: 2.5y = 7.5, y = 3; back-substitute: x = 1. Check: 2(1) + 3 = 5.
Then find the determinant of A = [4 2; 3 1] and state whether Ax = b has a unique solution.
Show answer
det = 4(1) − 2(3) = −2 ≠ 0, so a unique solution exists for any b.
Two springs in parallel hold a bar: k₁u + k₂u = F with k₁ = 300, k₂ = 500 N/mm, F = 4000 N. Then two springs in series carry the same 4000 N. Set up and solve both as linear equations for the deflections.
Show answer
Parallel: (300 + 500)u = 4000, u = 5 mm. Series: each spring carries 4000 N: u₁ = 13.33 mm, u₂ = 8 mm, total 21.33 mm. Stiffnesses add in parallel; flexibilities add in series. This is the seed of the FEA stiffness matrix.
Multiply A = [1 2; 0 3] by x = [4; 5].
Show answer
Ax = [1·4 + 2·5; 0·4 + 3·5] = [14; 15]. Each entry is one row dotted with the column.
Solve the 3×3 system: x + y + z = 6; 2x − y + z = 3; x + 2y − z = 2. State det's role before solving.
Show answer
Elimination gives x = 1, y = 2, z = 3 (substitute back to verify all three). A nonzero determinant guaranteed this unique solution before any work.
Write the joint equilibrium equations of the Statics Module 6 triangular truss (three joints, two equations each) as one matrix system in the member forces and reactions. Identify the matrix size and which physical fact each row states.
What good work looks like
A 6-equation system with each row labeled (joint, direction), the unknown vector listed, and a sentence noting that solving it reproduces the 8.49 kN and 6 kN answers from Statics.
Working with AI, and proving it yourself
Use AI as an examiner, not a solver
Portfolio task
Build a small Gaussian elimination routine (spreadsheet steps or Python) for 3×3 systems. Validate on the Level 3 system and on the traffic-light system from the worked example, and make it report when a pivot hits zero.
Retrieval and spaced review
Closed notes. Answer out loud, then reveal.
1. What do A, x, and b each hold in an engineering Ax = b?
A: the system's geometry and properties (angles, stiffnesses). x: the unknowns (forces, displacements). b: the applied loads or sources.
2. Describe Gaussian elimination in two sentences.
Subtract multiples of rows to create zeros below the diagonal. Then solve from the bottom row upward by back-substitution.
3. What single question does the determinant answer?
Whether the system has a unique solution: det ≠ 0 yes, det = 0 no (redundant or contradictory).
4. What is the rank of a matrix, in engineering words?
The number of genuinely independent equations (or directions) the system contains.
5. What does a homogeneous system (b = 0) ask physically?
Whether nonzero internal states exist with zero load: the question behind buckling and vibration modes (Eigenvalues and Modes).
Textbook mapping
| Item | Mapping |
|---|---|
| Main source | Kreyszig, Advanced Engineering Mathematics, Ch 7 (matrices, vectors, determinants, linear systems) |
| Core topics | 9.1 Notation · 9.2 Addition and multiplication · 9.3 Linear systems · 9.4 Gaussian elimination · 9.5 Inverse · 9.6 Determinants (minimum) · 9.7 Rank and independence · 9.8 Homogeneous systems · 9.9 Springs, trusses, circuits |
| Engineering connection | Structures and FEA (stiffness systems), Controls, Numerical Methods, least-squares data fitting. |
| Skip on first pass | Cofactor expansions beyond 3×3, abstract vector spaces, proof-based treatments. |
| Read next | Eigenvalues, Eigenvectors, and Modes. |