WIDENER UNIVERSITY School of Engineering

Third Edition
Tradition & Excellence Since 1821 Mechanical Engineering

TOPIC

ME 354

03 Computations in
Mechanical Engineering

Matrices

© Tulong Zhu, All rights reserved.

That’s THE MATRIX

What is a MATRIX ?

is
Matrix

The

03 - 2
Matrices: Definition
Definition • A matrix is a rectangular array of elements written as

 a11 a12  a1n  • A matrix with m rows and n

a  a2 n 
 21 a22  columns
A
      • aij is the element of the
a  amn  matrix A.
 m1 am 2

 An m n matrix
 A matrix of size m n
 A matrix of order m n
 Sometimes A is denoted as A = [aij] or (aij)

Equivalent • Two matrices A = [aij] and B = [bij] are equal if

Matrices  They are of the same size, and
 The corresponding elements are equal, i.e.,
aij = bij

03 - 3
Matrices Operations: Addition & Scalar Multiplication
Addition • Two matrices of the same size can be added by performing
element-by-element addition, i.e.,
 a11 a12  a1n   b11 b12  b1n 
a  a2 n  b  b2n 
 21 a22 
b22
A B   21 
           
a  amn  b
 m1 am 2  m1 bm 2  bmn 

 a11  b11 a12  b12  a1n  b1n 

a b a22  b22  a2n  b2n 
C AB   21 21 
     
a  b  amn  bmn 
 m1 m1 am 2  bm 2

Scalar  ka11 ka12  ka1n 

Multipl.  ka  ka2n 
21 ka22
• kA   
     
ka  kamn 
 m1 kam 2
03 - 4
Matrices Operations: Matrix Multiplication
Matrix • If A is m  n and B is n  p , the C = AB is m  p
Multiplica-
tion  a11 a12  a1n   b11 b12  b1 p 
a  a2 n  b  b2 p 
 21 a22   21 b22 
A B
           
a b  bnp 
 m1 am 2  amn   n1 bn 2
 b1 j 
 c11 c12  c1 p  b 
cij  [ai1 ai 2  ain ] 
2j
c  c2 p 
 21 c22    
C b 
     
c  nj 
 m1 cm 2  cmp   ai1b1 j  ai 2b2 j    ain bnj

Notes  Unless m = p, BA is not defined.

 AB  BA even if both AB and BA are defined.
03 - 5
Matrices Operations: Transpose
Transpose • If A is m  n , the transpose AT, (A’) is n  m, given by

 a11 a12  a1n   a11 a21  am1 

a  a2 n  a  am 2 
 21 a22 
a22
A AT  A   12 
           
a  amn  a  amn 
 m1 am 2  1n a2 n

Notes  A T T
A

 ( AB )T  BT AT

Row and • A row vector, x = [x1, x2, …, , xn ]can be considered as a matrix

Column of size 1  n.
Vectors
 y1 
y 
• A column vector, y   
2
  
y 
 m
03 - 6
can be considered as a matrix of size m  1.
Matrices Operations: Properties
Properties • In the following properties, all operations are assumed to be
definable.

• A+B=B+A

• (A + B) + C = A + (B + C)

• A(B + C) = AB + AC

• (A + B)C = AC + BC

• k(A + B) = kA + kB

• A(BC) = (AB)C

• (AB)T = BTAT

• AB  BA for most matrices.

03 - 7
Matrix Operations: MATLAB© Commands
Types Operation Command Math Form
Basic Addition (subs) AB AB
Operations Scalar Multipl. k*A kA
Transpose A’ AT or A’
Multiplication A*B AB
Matrix Single Element A(i,j) aij
Index
Row A(i,:) [ai1, ai2, …, ain]
Column A(:,j) [a1j, a2j, …, amj]T
Sub-matrix A(i:j,k:l) A sub-matrix of A from rows i to
j and from columns k to l
Define Empty matrix [] Empty matrix
Special Matrix of 1s ones(m,n) m  n matrix of ones
Matrices
Matrix of 0s zeros(m,n) m  n matrix of zeros
Matrix of random rand(m,n) m  n matrix of random
numbers numbers
03 - 8
MATLAB© Element-by-Element Operations
Operation Command Math Form
Multiplication A.*B  a11b11  a1nb1n 
    
 
am1bm1  amnbmn 
Power A.^k  a11k
 a1kn 
 
    
a k  k 
amn
 m1 
Function f(A)  f (a11 )  f (a1n ) 
    
 
 f (am1 )  f (amn )
Add Scalar A+k  a11  k  a1n  k 
    
 
am1  k  amn  k 

See Ex03_01.m for matrix operations.

03 - 9
Square and Symmetric Matrices
Square • A square matrix has the same numbers of rows and columns.
Matrices  a11 a12  a1n 
a a22  a2n  • aii : diagonal elements
A 21 
      • aij (i  j ) off-diagonal elements
a 
 n1 an 2  ann 
Symmetric • A square matrix A with elements aij = aji is called a symmetric
Matrix matrix.
 a11 a12  a1n 
a a22  a2 n 
A   12 
     
a  ann 
 1n a2 n

Fact  A = AT for a symmetric matrix A.

An  xx  xy  xz 
Example  
 The stress matrix σ   xy  yy  yz 
  xz  yz  zz 
 
03 - 10
Square Matrices: Diagonal and Identity Matrices
Diagonal • A special symmetric matrix in which all the OFF-DIAGONAL
Matrices elements are zero.
d11 0  0 
 0 d  0 
D 22 
     
 0 0  d nn 


Identity • A special diagonal matrix with all diagonal elements being 1.

Matrix I
1 0  0
0 1  0
I 
   
0 0  1


 >> eye(n)

Note  For any square matrix Q, QI = IQ = Q

03 - 11
Determinants
Definition • The determinant of a square matrix A (n  n) is a number, denoted
as det(A) or |A|.
How to • The determinant can be computed using the Laplace expansion.
Compute? To do that, the minor and the cofactor need to be defined.
Minor  The minor of the element aij designated Mij is the n–1  n –1
submatrix of A by deleting row i and column j. For example
 a11 a12  a1n  a21 a23  a3n 
a a22  a2n  a a33  a3n 
A 21  M12   31 
           
a   a  
 n1 a n2 a nn   n1 a n3 a nn 
Cofactor  The cofactor of aij, designated Aij is value (–1)i+j det(Mij).
Laplace Exp • The determinant of A can be determined using all elements and
their cofactors of a row OR a column as follows:

 
n n
k j
Expand by
 detA    akj Akj    1 akj det M kj 
Row k j 1 j 1
n n
ik
 detA    aik Aik    1 aik detM ik 
Expand by
Column k 
03 - 12 i 1 i 1
Determinants of Order 2 and 3: Expand by 1st Row
a a 
• A 2nd order matrix A   11 12 
a21 a22 

A   111 a11a22   11 2 a12 a21  a11a22  a12 a21

 b11 b12 b13 

• A 3rd order matrix B  b21 b22 b23 

b31 b32 b33 

11 b22 b23 21 b12 b13 31 b12 b13

B   1 b11   1 b21   1 b31
b32 b33 b32 b33 b22 b23

B  b11 b22b33  b32b23   b21 b12b33  b32b13   b31 b12b23  b22b13 

B  b11b22b33  b12b23b31  b13b21b32  b31b22b13  b32b23b11  b33b21b12

– – –
b11 b12 b13 b11 b12
b21 b22 b23 b21 b22 This scheme is applied
to matrix of any order

b31 b32 b33 b31 b32
03 - 13 + + +
Determinants: Properties and MATLAB©
Properties • If A has a row or column of zeros, det(A) = 0
• If any two rows (or columns) of A are interchanged, the sign of
det(A) is changed.
• If all elements of a row ( or column) of A are multiplied by the
same number k, then the determinant becomes k det(A), e.g.,
ka11 ka12  ka1n a11 a12  a1n
a21 a22  a2 n a21 a22  a2 n
k
       
am1 am 2  amn am1 am 2  amn

 det(kA) = kn det(A) ( A is or order n)

• The determinant of a diagonal matrix is the product of the
diagonal elements.
• det(AT) = det(A)
• det(AB) = det(A) det(B)

03 - 14
• >> det(A) 
Square Matrices: Inverse
Reciprocal • For a number a  0, aa–1 = a–1a = 1, a–1 is the inverse of a.
Inverse • Inverse of a square matrix A, designated A–1 is such that
Matrix AA–1 =A–1A = I
 The inverse A–1 may not exist.
 If the inverse A–1 exists, A is said to be invertible or
nonsingular. Otherwise, A is noninvertible or more
commonly singular.
 If A is invertible, then det(A)  0, vice versa.
 Only a square matrix can have an inverse.
 I–1 = I
 (AB)–1 = B–1 A–1
How? • Using cofactors. 
• Solve for each element in A–1 from AA–1 = I 
• Using elementary row operations. 
• >>inv(A), >>rref(A) 
03 - 15
• …
Inverse Matrix: Using cofactors
Cofactors • Recall the cofactor Aij
 The minor of the element aij designated Mij is the n–1  n –1
submatrix of A by deleting row i and column j. For example
 a11 a12  a1n  a21 a23  a3n 
a  a2 n  a a33  a3n 
 21 a22 
A M12   31 
           
a  ann  a 
 n1 an 2  n1 an3  ann 
 Aij = (–1)i+j det(Mij).

Inverse T
 A11 A12  A1n 
Matrix A A22  A2n 
1 adjA    21 
1
• A  adjA 
 detA   
A
   
 n1 An 2  Ann 

 For A to be invertible, det(A) must not be zero.

03 - 16
Inverse Matrix: Elementary Row Operations
Elementary • Interchange two rows (Ri  Rj).
Row • Multiply a row by some nonzero constant (kRi).
Operations • Add a multiple of a row to another row(kRj + Ri).
Inverse • Form the augmented matrix that has A on the left and the
Matrix from identity matrix (same size as A), I, on the right, [A | I ].
Elem. Row
Operations • Perform elementary row operations on the augmented matrix
until the left part (at A) becomes an identity matrix, or a row of
zeros appears on the left part.
 If the procedure produces a row of zeros, then the matrix A
is singular (noninvertible).

 If it produces an identity matrix, then the matrix on the right

of the augmented matrix is the inverse, A–1, i.e., the original
augmented matrix [A | I ] becomes [I | A–1 ]
• MATLAB has commands to do this procedure:
>>rref[A eye(n)] [more accurate than >>inv(A)]
>>rrefmovie[A eye(n)] (step by step display)
03 - 17
Inverse from Ele. Row Operations: An Example
3  1 0  3  1 0 1 0 0
A  1 1 2 The augmented matrix 1 1 2 0 1 0 R1  R2
   
0 2 4 0 2 4 0 0 1

1 1 2 0 1 0 1 1 2 0 1 0
3 –3R1 + R2 0  4  6 1  3 0 R2  R3
1 0 1 0 0  
 
0 2 4 0 0 1 0 2 4 0 0 1

1 1 2 0 1 0 1 1 2 0 1 0  –R + R
0 2 0 1 2 0 0 1 / 2  2 1
4 0 0 1 1/2 R2   4R2 + R3
 
0  4  6 1  3 0 0  4  6 1  3 0 
1 0 0 0 1  1 / 2 1 0 0 0 1  1 / 2
0 1 2 0 0 1 / 2  1/2 R3 0 1 2 0 0 1 / 2  –R3 + R2
   
0 0 2 1  3 2  0 0 1 1 / 2  3 / 2 1 

1 0 0 0 1  1/ 2   0 1  1/ 2 
0 1 0  1 3  3 / 2  A 1    1 3  3 / 2 See Ex03_02.m.
   
0 0 1 1 / 2  3 / 2 1  1 / 2  3 / 2 1 
03 - 18
Square Matrices: Orthogonal Matrices
Orthogonal • An orthogonal matrix Q is an n  n matrix that is invertible and
Matrices Q–1 = QT or QTQ = QQT = I

Example cos  sin  

 The vector transformation matrix Q   is an
orthogonal matrix.  sin  cos 

T  cos sin   cos  sin  

Q Q
 sin  cos   sin  cos 
 cos2   sin 2   sin  cos  sin  cos 
 
 sin  cos  sin  cos cos   sin 
2 2

1 0 
  I
0 1 
Similarly, QQT = I

Theorem • The n  n matrix Q is orthogonal if and only if the columns (and

rows) of Q forms an orthonormal system.
03 - 19
Square Matrices: Triangular Matrices
Triangular • Lower Triangular Matrix • Upper Triangular Matrix
Matrices
 a11 0  0  a11 a12  a1n 
a  0   0 a  a2 n 
 21 a22 
A A 22 
           
a  ann   0 
 n1 an 2  0  a nn 

 The diagonal matrix and the identity matrix are special

triangular matrices. They are both lower and upper triangular
matrices.

Properties • Det(A) = |A| = a11a22…ann

• The inverse of an upper (lower) triangular matrix is up (lower)
triangular.
• The product of two upper (lower) triangular matrices is upper
(lower) triangular.

• The eigenvalues of a triangular matrix are the diagonal elements.

03 - 20
LU Factorization (Decomposition)
LU • Any nonsingular square matrix A may be decomposed into the
Factoriza- product of two triangular matrices as,
tion
A = LU

1 0  0
l 1  0
• A unit lower triangular
L   21  matrix with all diagonal
     elements equal to 1
l  1
 n1 ln 2

u11 u12  u1n 

 0 u  u2n 
U 22  • An upper triangular
      matrix.
 0 0  unn 


03 - 21
LU Factorization: Equations
Manual  a11 a12  a1n   1 0  0 u11 u12  u1n 
a a22  a2n   l21 1  0  0 u22  u2n
A   21   
                
a  ann   ln1
 n1 an 2 ln2  1  0 0  unn

How to Get • (1st row of L)  (All columns of U)  1st row of U, u1j = a1j
lij and uij?
• (2nd row of L)  (All columns of U)  2nd rows of L and U:
l21  a21 / u11, u2 j  a2 j  l21u1 j

• (ith row of L)  (All columns of U)  ith rows of L and U:

j 1 i 1
lij  (aij   lik ukj ) / u jj , uij  aij   lik ukj
k 1 k 1

 When calculating lij, ujj is divided. This will generate

computational error if |ujj| is very small.
 To reduce the error, the LU factorization with pivoting
(exchanging rows and/or columns in A) should be used
(MATLAB uses row exchange!!).
03 - 22
LU Factorization: MATLAB©
>>[L,U,P]=lu(A)
• Provides the L, U and P (permutation) matrices. The
permutation matrix is basically an identity matrix with rows
and columns reordered. P reveals the information about how
rows in A are changed before the factorization is performed.
 The process from the identity matrix I to P is the same as
the process of exchanging rows in A, e.g.,
1 0 0 
P  0 0 1 
Shows that the 2nd and 3rd rows are
  exchanged
0 1 0

• The actual factorization in MATLAB (and it should be the way

factorization is performed to reduce errors) is
PA = LU, not A = LU

See Ex03_03.m

03 - 23