Sequential Diagnosis Tool
Web Interface Help
System operation in the presence of faults can be presented by different system failure states. The goal of the
diagnostic procedure is to identify the actual failure state by executing a sequence of tests which provide some
information on system behaviour. In principle, any measurement, signal, or other observable event can be viewed as a test.
Determining the sequence of tests required to reach a diagnostic conclusion at minimum cost is known as the
test sequencing problem.
More precisely, for a system described by
- the set of the faulty system states,
- the probabilities of faulty system states,
- the set of available tests,
- the costs, and
- the test matrix presenting the diagnostic capabilities of the tests,
the problem is to generate the diagnostic procedure such as that the average cost of the decision tree is minimized.
Test sequencing problem has been originally defined for symmetrical and binary tests. It was later generalized to
include asymmetrical and multi-valued tests.
(More details on this issue are given in the tutorial.)
SDT generates solutions for the generalized Test Sequencing Problem.
In order to run SDT the following input data are required:
Test matrix dimension is specified by
- number of system states: m + 1
- number of tests: n
Algorithm:
You can select one of the following algorithms for generation of the diagnostic procedure:
- AO* HEF1
- AO* HEF2
- AO* HEF3
- PQ
(Default selection is AO* HEF1)
Test matrix - the following input data are required:
- test names,
- system state names,
- a-priori probabilities of system states (The sum of the probabilities of all the states
must be equal to 1.),
- test costs,
- matrix elements are test outcomes for given system states. (If test tj is
symmetrical for system state si: test outcome is 1 if tj fails and 0
otherwise. In the case of an asymmetrical test, multiple outcomes are separated by semicolon and include outcome
probabilities in brackets. For example r1(0.3);r2(0.7) means that test outcome r1 has probability
0.3 and outcome r2 has probability 0.7.
The sum of probabilities for the given test and system state must be equal to 1.)
Note: By default, tests are denoted by (t1, ..., tn), system states by (s0, ..., sm) and test costs are all
equal to 1. Otherwise, notice that state names, test names and matrix elements should not contain spaces and colons.
Any space or colon in system state or test name will be replaced with the '_' character,
or removed in the outcome name. Names for tests, system states and outcomes should not exceed 12 character limit.
Longer names will be shortened to the first 12 characters.
Display error messages
- Select this option if you want Web SDT to display messages about errors in test matrix. (default selection is off)
Submit values
- Click the 'SUBMIT VALUES' button each time you want to apply changes.
Errors reported by web SDT
Web SDT checks and reports these errors:
- Error: Tests must have at least 2 outcomes!
SDT accepts binary and multivalued tests. Tests with less than two outcomes are not valid.
- Error: Sum of a-priori probabilities in test matrix must be 1!
A-priori probabilities are listed under the p(s) column. Their sum must be equal to 1.
- Error! A-priori probability of system state s must be greater than 0!
System states must have probabilities in range (0..1). - (greater than zero and less than one)
- Error! Tests have equal names!
You have entered two or more test with equal names. All tests must have unique names.
- Error! System states have equal names!
You have entered two or more system states with equal names. All system states must have unique names.
- Error: Sum of conditional probabilities for state s and test t must be 1!
You have entered asymmetrical test, but sum of outcome probabilities is not equal to 1.
- Error! Invalid test cost (test t)!
Test costs must be greater than 0.
links:
SDT Web Interface |
SDT Examples |
SDT main page |
IJS Computer Systems Department