Dynamic Engineering Corp.
Government Research
Problem:
A US Government Agency contracted DYNAMIC Engineering in a three(3) year contract to review Nuclear
Reactor Containment testing practices with advanced computer modeling and numerical analysis
techniques. With the newer techniques in data acquisition and calculating safety leakage limits
of each containment vessel, the US NRC had serious concerns about testing practices.
The NRC was concerned about accuracy, repeatability, and overall test integrity. One presented
theory and concern was the ability of computers to take data in speeds far beyond those tests
previously performed by manual methods. Such speeds could be multiple times a second, causing
round off errors and other statistical problems not anticipated in the accepted numerical
analysis.
Solution:
DYNAMIC Engineering broke the project down into several phases. The first phase was to determine
instrumentation accuracy and integrity.
- Phase-I
-
Then an investigation was performed to take a sample
group of perfromed nuclear containment tests and statistically fabricate data, as though the
data was taken at a high rate of speed. Then the data was re-analyzed to test the statistical
methods for aberrations.
Data smoothing techniques were then applied to the intrumentation data to determine if there
was any aberrant behavoir in the equations.
-
Phase-II
-
Visit a specified group of nuclear utilities and perform back to back tests. The utility
performed its own Reactor Vessel Test and DYANMIC would perform a simultaneous Reactor Vessel
test. The results were independantly reviewed and comapared. A separate end criteria was
investigated to determine wether enough data was collected to statistically have sufficient
data to have confidence in the test results
Results:
It was determined that acquired data when taken at high rates would not effect the statistics
abnormally. As one would conquer with logic the more data the more accurate the test results.
However, When data was "smoothed" either using polynomials or forrier tranforms the data could
be preconditioned to fool the statistics into a non numerically supported conclusion.
From Phase-II it was concluded that the test methods in-place were adaquate to evaluate reactor
containment vessels. Data acquired at fast rates would not affect the test results. Instrumentation
data smoothing, conditioning, or manipulation in any way must be prohibited.
It must be stated that this research contract was too extensive and far beyond the scope of this
description.