Thermoelastic Forum Vol.1, No.3, December
1992
First Words
R&D Side
Tech Tips
Events
UW Corner
New Products
We are excited about the excellent response that we have had to this
newsletter. Many SPATE users have told us that they find the Thermoelastic
Forum very useful. Whether good or bad, we like to receive your feed back.
If you find an article in the newsletter that raises questions, give us
a call. We are always happy to discuss your SPATE applications or questions.
Let us remind you that Stress Photonics would like this newsletter to be
an open forum. If you have a topic you would like covered, or if you want
to contribute a thought, suggestion, article, or even a complaint, we are
all ears.
Need extra copies of this or either of the the two previous newsletters?
We have an ample supply. Just mark the appropriate box on the reply card
and we'll include them with your next mailing.
Many of us will be vacationing during the winter holidays so we would like
to take this opportunity to wish you a safe, happy, and refreshing holiday
season.
By Jon Lesniak
Stress Photonics is currently developing software aimed at helping the SPATE
user more accurately and conveniently quantify stress intensity parameters.
A key part of this software is improving edge data, where a majority of
stress intensity measurements are made. In this article, I will show how
optical blur can be corrected through a deconvolution of the SPATE image
in order to achieve more accurate stress measurements.
Hole in Plate
Let's look at an example situation. Figure 1 shows the SPATE signal near
a hole normalized to the nominal stress. The data was collected from a 4"
wide plate with a 5/16" center hole. The scan was performed at a camera
distance of 24". The theoretical stress concentration for this geometry
is approximately 2.75, yet the data would suggest an intensity factor of
2.45. This represents an 11% error. Errors of 20% or more are common in
this situation. These inaccuracies can be a critical impediment when making
fatigue life predictions.
Deconvolution
The concept behind the deconvolution is that all the signal from the stress
concentration is accounted for, but is spatially blurred. The detector spatial
sensitivity profile can be measured to describe exactly how the signal is
blurred and this information can be used to put the signal back where it
belongs. Therefore, this technique is a deconvolution not an extrapolation.
The detector sensitivity profile for this study was obtained by scanning
across a slit placed in front of a chopped heat source.
The deconvolution is accomplish by first
selecting a mathematical basis that can reliably describe typical stress
states. For the 1-D work I chose a very general Chebyshev basis, the first
few terms of which are:
The stress distribution is assumed to be a linear combination of these fundamental polynomials.
Of course SPATE doesn't see it this way, it measures a blur of the linear combination. To model this, each polynomial is blurred. The SPATE data S(x) is equivalently the linear combination of the blurred basis T'.
For example, figure 2 shows the T1 polynomial and the blurred base T'1
which is obtained by convoluting T1 with the detector sensitivity profile.
If the blurred basis is fit to the SPATE data the magnitudes of Cn can be
found. Using these coefficients, the first equation can be solved for the
stress distribution and a stress intensity factor can be estimated.
Soft Edge
To make the application of this routine easier, two types of edges can be
defined, a soft edge and a hard edge. A soft edge is defined as the edge
of the data set that does not correspond to a geometrical edge. It is simply
where the scan starts or stops in the middle of the stress field. The soft
edge works best if set in a smooth stress field as demonstrated in figure
1. For the purpose of creating the blurred basis, the routine assumes that
the base polynomial continues on past the soft edge (fig. 2.)
Hard Edge
A hard edge is set to correspond to a geometrical edge. In this case the
basis is constructed to extend a detector radius beyond this edge in order
to model the signal that is blurred over the edge. It is not always easy
for the user to determine exactly where the geometrical edge is, so the
software searches both sides of the user set hard edge to see if a better
fit can be made with a refined edge position. The user only needs to get
in the ball park.
Two Dimensional Deconvolution
In many cases a simple 1-D deconvolution will be adequate, but if the signal
gradient is too severe normal to the scan line a 2-D convolution is necessary.
Here the signal is described by a 2-D basis. The basis can be selected to
impose compatibility requirements.
I am very interested in your stress intensity or other stress parameter
needs. I would appreciate some feed back about your specific applications.
Please give me a call.
By Brad Boyce
If you are interested in improving your color hard copy capability then
you should learn about the HP DeskJet 550C. It's a low cost, high resolution
printer. Part of the DeskJet line, it uses ink jet printing to produce laser
printer quality outputs. It has a cut-sheet paper feed, parallel and serial
ports, and conforms to Hewlett Packard's PCL (Printer Control Language)
just like HP's LaserJet printers.
The Deskjet 550C offers an interesting combination of features:
The only drawback to the printer is that color outputs can take as much
as five minutes per page.
To use the printer for SPATE applications you can use the HP PaintJet printer
driver provided with your HP Basic computer. A simple modification to the
driver (available free from Stress Photonics) is needed to take advantage
of the resolution range of the DeskJet. The printer can either be connected
directly to the HP computer's serial port, or the HPIB through an HPIB/parallel
interface to speed printing. If you have a DOS computer or a Mac nearby
it works great for them too! If you are considering a printer purchase give
us a call, we'd be happy to discuss the options with you.
By Brad Boyce
Are you constantly out of disk space? Well, there are some inexpensive things
you can do about it.
One of the things you can do to improve disk space is to use a utility program
that HP provided you with your HP BASIC language. The "MASS_STOR"
program can be used to repack a disk. Due to the way HP BASIC uses disk
space (under the Logical Interchange Format, LIF) the disk tends to become
fragmented. Small areas of unused disk space pile up as files are added
and deleted from your drive. Repacking the disk recovers this unused area.
How much area you can recover using "MASS_STOR" depends on how
you use your system. A recovery of 10-20% of the space on your disk may
be possible.
If you still don't have enough space you should look into buying some more
disk space. Although new HP equipment can be expensive, Stress Photonics
has found that used HP equipment is inexpensive and reliable. Generally,
10Meg will go a long way to solving your space problems. We recently priced
a used 20Meg drive at $550 with overnight delivery. Give us a call if you
need the names of some sources of used HP computer equipment.
The next SPATE Application Working Group Meeting will be held in Detroit
in conjunction with the SEM Spring Conference. Mark your calendars for Monday
7 June 1993. Two technical paper sessions, a thermoelastic tutorial, and
the SAWG meeting/round-table discussion will all be held that day. And of
course there will be the manufacturers exhibits as well. The conference
will be held at the Hyatt Regency, Dearborn, Michigan.
Publications:
"Thermoelastic Determination of Individual Stress Components in Loaded
Composites," Feng, Z., Zhang, D., Rowlands, R.E., Sandor, B.I., Experimental
Mechanics, June 1992.
By B.J. Rauch & R.E. Rowlands
Department of Engineering Mechanics, University of Wisconsin-Madison
Excerpts from a paper to be presented at the Spring 1993 SEM conference,
Dearborn, Michigan.
This paper describes a finite element based scheme which accurately and
efficiently separates the isopachic data into the individual in-plane stress
components. The governing elliptic partial differential equations require
complete boundary conditions to be well-posed, however, it is shown that
the stress separation algorithm can be employed with incomplete boundary
conditions and yield accurate results. Examples with theoretical and experimentally
measured isopachic data are given to demonstrate the technique.
Complete boundary conditions are not always available for a problem. This
paper demonstrates the amount and type of information that is necessary
for an engineering quality solution. Particular attention is given to the
use of additional information that can be obtained by natural boundary conditions
or additional experimentation, such as strain rosettes.
Figure 1 and 2 are the separated normal stresses in the horizontal and vertical
directions respectively for a plate with a central hole. The specimen was
loaded in the vertical direction by both a force and a small moment. The
isopachic stresses were experimentally measured and further processed with
smoothing and edge enhancement algorithms before stress separation was performed.
The correlation with strain gage measurements was excellent.
By Brad Boyce
Stress Photonics has developed the Multi-Contour Software to meet the special
needs of turbine blade SPATE stress analysis. The software is designed to
aid in the location of high stress regions common to a number of vibrational
modes of a single part. To use the software, several vibrational modes of
a part are scanned. The scans are then smoothed and normalized to the highest
stress within each scan. The index feature of the Filer is used to group
the scans. The Multi-Contour Software is then used to overlay the highest
contour lines from each frequency. By keying the color of the contours to
the frequency of the scan, not the stress level as is normally done, a valuable
plot is constructed for determining the strategic location of strain gages
for subsequent testing.
By Brad Boyce
With the new Box Averager Software from Stress Photonics you can quickly
find the average value of SPATE data within a rectangle drawn on the screen.
The Box Averager Software is a handy little routine for calibration, damage
assessment, or for finding nominal stress levels. The routine provides the
capability to draw rectangular areas on the screen and can either find the
average value within a rectangle or the average within a number of rectangles.
Both the Multi-Contour and Box Averager Software were written by Stress
Photonics as the result of requests from users. If you would like to incorporate
these routines or others of your own design into your SPATE or VPI system
contact Stress Photonics.
