7.6. COMPUTATION OF UNIFORM RESIDUAL STRESSES 153
Initial optical measurements for hole-drilling residual stress evaluations used calculation
methods analogous to those for use with strain gauges. Typically, they involved visually pick-
ing a small number of opportune points within the measured images, interpreting their fringe
orders, and then doing a strain gauge style calculation. Such fringe pattern interpretation is an
important need when working with the unsigned analog data that comes from traditional holo-
graphic measurements with photographic or thermoplastic plates. Although effective results are
achieved, the performance of these methods can be further enhanced by including the contri-
butions of the substantial quantity of additional data available beyond the few selected points.
e ESPI, phase-stepped Moiré and DIC methods share the advantage that they provide
signed numerical data. A typical image taken with a modern video camera contains several hun-
dred thousands, often millions of pixels, thus a very large number of independent measurements
are produced. e availability of such large amounts of quantitative data makes these methods
well suited for more in-depth data analysis.
In contrast to the traditional strain gauge style measurements, all optical techniques indi-
cate surface displacements, not strain. Estimation of surface strains from displacement measure-
ments involves numerical differentiation, which is an inherently noisy process, and consequently
is to be avoided. us, it is desirable to choose computation methods that work directly with
displacement data. In addition, linear methods are particularly desirable to achieve effective and
compact data processing. Nonlinear procedures can also be effective, but they are much more
computationally intensive, potentially less stable, and thus should be used only when essential.
A computation that has a greater number of data than unknowns is called “over-
determined.” e number of available optical data is much greater than the number of un-
knowns, so there exists a good opportunity to extract some further results from the measure-
ments. Here, it is very effective also to include in the calculation the effects of systematic artifacts
that can occur in practical experimental measurements. For example, in each of the various op-
tical methods, drifts often occur that cause the data from all image pixels to shift by the same
amount. In the ESPI and Moiré methods such drifts occur from temperature changes within
the apparatus, with DIC they occur from small position changes of the optical components.
Similarly, image stretching can occur due to specimen temperature changes, or to magnification
changes in the case of DIC measurements within a scanning electron microscope. Even with
these additional factors included in the calculation, the residual stress computation is still highly
overdetermined. e excess data can usefully be exploited to provide an averaging effect that
acts to reduce the effect of random measurement noise.
It is generally impossible to achieve a numerical solution of an overdetermined measure-
ment that exactly fits all measured data. Consequently, a “best-fit” solution is sought that is
as close as possible to the majority of the measured data. is can conveniently be done using
the Least-Squares method. is approach was introduced for ESPI measurements by Steinzig
and Ponslet (2003) and further developed by subsequent authors. Figure 7.10 schematically de-
scribes the procedure. e starting point is the measured image shown on the left. is could be