7.7. COMPUTATION OF NON-UNIFORM RESIDUAL STRESSES 159
1. Drill a hole in a series of small depth increments down to the final desired depth.
2. Make an optical measurement of the surface around the hole after each hole depth incre-
ment has been completed.
3. Discretize the residual stress profile into a series of steps based on the hole depth incre-
ments (see Figure 5.7).
4. Use the Integral Method to determine those stresses from the combination of all the mea-
sured data.
In concept, the procedure for implementing the Integral Method with full-field optical
data substitutes Equation (7.1) for the analogous strain gauge Equations (5.18). Figure 5.8 in
Chapter 5 schematically illustrates the computational concept of the Integral Method. e dia-
grams represent the responses of the material around a drilled hole to the residual stresses within
various hole depth increments in holes of various depths. e measured response at a given hole
depth is the sum of all the responses along the row corresponding to that hole depth. is ar-
rangement can be expressed directly in matrix format, which for strain gauge calculations gives
the lower triangular matrices
N
a and
N
b described in Chapter 5.
For strain gauge measurements, single-strain values are measured to identify each of the
stress components P; Q, and T . us, a single number represents each loading case in Figure 5.8
within the matrices
N
a and
N
b. However, when making full-field optical measurements, the num-
ber of measured displacement values for each loading case is in the 10
4
–10
6
range. Under these
circumstances, each number in matrices
N
a and
N
b is replaced by the much larger grouping shown
in Figure 7.13 (or a reduced version for single-axis measurements) within a much larger com-
bined G matrix. Figure 7.14 schematically illustrates the resulting matrix equation, where the
rectangles represent entire matrices and vectors from Figure 7.13, each corresponding to its as-
sociated load case in Figure 5.8.
e G matrix in Figure 7.14 is very large, so significant care is required to handle it ef-
fectively. Fortunately, the corresponding G
T
G matrix for Equation (7.2) is much smaller, with
dimensions 9n 9n (or 6n 6n for a single-axis measurement), where n is the number of hole
depth increments used. is is a much more tractable size and is comfortably handled by mod-
ern computers. e various columns in the G matrix and ı vector, respectively, represent the
theoretical and measured full-field images for the various load cases. ese do not all need to be
held in computer memory at the same time, thus avoiding the need to store the entire G matrix
simultaneously. Instead, the various images can be loaded a few at a time and the G
T
G matrix
progressively assembled by incrementing the dot products of corresponding parts of columns.
Of note is that the parts of the G matrix corresponding to the artifacts w
4
–w
9
are the same at
each hole depth, so one copy of these numbers suffices for all hole depths. On a conventional
2017 desk computer, the computation time for a 16-increment single-axis measurement with
100,000 active pixels was approximately one minute. e required computation time varies ap-
proximately with the square of the number of hole depth increments and proportionally to the