Diagnosis [Conditioned on Failure]

Note that in this figure, the node "LPCI-Sys" is red. Because of our assertion its failure probability is now one. Similarly, the failure probability for "Lpci-Tr-A" and "Lpci-Tr-B" are now also 1.0 (both must fail for the system to fail), therefore they are red as well. The node "LPCI-Sub-B" is also dark red, but its failure probability is not 1.0, just very high (0.99). This makes logical sense, the LPCI subsystem only has one motor operated valve, but the pump subsystem has two redundant pumps. Thus the pump subsystem is more reliable and is less likely to be the failure cause. "LPCI-Sub-A" has a slightly lower failure probability (conditioning on system failure) of 0.98 and hence it is a slightly pale red. This is because the test data for "Pump-A" and "Pump-C" indicate that they are less reliable than "Pump-B" and "Pump-D". (See the data example for more information on how Graphical-Belief incorporates test data into the model.)

Next we look at the individual components. "MOV-25-A" has failure probability 0.92 and it is now a light red. "CV-46-A" has failure probability 0.061 and it is now a pale blue. This makes sense because the Check valves (CV) are much higher reliability components than the motor operated valves (MOV). Thus we are quickly led to believe that "MOV-25-A" is the most likely problem in branch A of the fault tree.

Looking at the node "MOV-25-A" we can now see the purpose of the black bars on either side of the node. They give the probability of the positive state of the node (in this example no failure). The bar on the left hand side gives the baseline probability (before any changes) and the bar on the right gives the active probability (after the current changes). This makes it simple to perform comparisons on the nodes of the graph. The history browser allows the analyst to change either the baseline or active state and thus go back and see previous states of the model. Probes are available to provide a more detailed view of the changes at each node.

Suppose we are diagnosising a problem with this system and we follow the information provided above and check MOV-25-A and discover it is working properly. We can enter this new finding into the model and get a revised estimate of the probabilities. Figure~3 shows the result.

Return to the main example page.

Back to overview of Graphical-Belief.

View a list of Graphical-Belief in publications and downloadable technical reports.

The Graphical-Belief user interface is implemented in Garnet.

Get more information about obtaining Graphical-Belief (and why it is not generally available).

get the home page for Russell Almond , author of Graphical-Belief.

Click here to get to the home page for Insightful (the company that StatSci has eventually evolved into).