MIL-HDBK-1823
"Nondestructive Evaluation System
Reliability Assessment"
Overview:
MIL-HDBK-1823 tells how to plan an NDE experiment, design
and fabricate reliability demonstration specimens, acquire the system
performance data, and provides statistical methods for analyzing the data to produce POD(a) curves, 95%
confidence bounds, noise analysis, and noise vs detection trade-off curves.
It presents worked-out examples using real Hit/Miss and ?/i> data, and serves
as a user?s manual for the mh1823 POD software.
These methods are statistical best-practices and have
universal applicability ? NDE of engines, airframes, ground vehicles ?
subject to the following limitations:
-
The NDE systems must produce output that can be reduced
to either a quantitative signal, ?/i>, or a binary response, hit/miss.
(Images therefore will require some pre-processing to provide either ?/i>
or Hit/Miss as input to these analysis methods.)
-
The specimens must have targets with measurable
characteristics, like size or chemical composition. This precludes
amorphous targets like corrosion unless a specific measure can be
associated with it, such that other corrosion having that same measure
will produce the same output from the NDE equipment.
-
This mh1823 POD software assumes that the input data are
correct. That is, if the size is X, then that is the true size.
If the response is Y, then that is the true response. Situations
where these conditions cannot be ensured (e.g. where target sizing is
only approximate) will necessarily provide only approximate results.
(The problem of accurate crack sizing is discussed in Handbook Appendix
I.1 Departures from Underlying Assumptions ? Crack Sizing and POD
Analysis of Images.)
-
This mh1823 POD software assumes that a POD curve goes
to zero on the left, and to one on the right. Data for which min(POD) >
0 (perhaps due to signal contamination by excessive background noise),
or max(POD) < 1 (resulting from random misses not related to target
size) cannot be correctly represented by a model for which min(POD) = 0
and max(POD) =1. (See MIL-HDBK-1823, Appendix I-4 "Asymptotic POD
Functions.")
http://www.statisticalengineering.com/mh1823/QNDE/POD-floor-ceiling.htm
If the input data do not meet MIL-HDBK-1823 requirements, the mh1823 POD
software may still produce an answer, but it will be WRONG.
Background:
| NOTE: The mh1823
POD algorithms use R,
the most powerful statistical and graphics engine available anywhere,
for all data manipulation, statistical analysis, and graphics.
Because R is open-source (and free),
and because all of the algorithms and methods
developed here are based on modern, well documented
statistical best practices described in the open
literature, there is nothing proprietary in this
mh1823
POD software. Since there are no restrictions
on its use, the mh1823
POD software can be used as a
universal standard for performing Probability of
Detection (POD) analysis. |
|
