Home Overview Objectives Applications Target Market Product Details Background Publications		Product
	Technical Details and Innovations

The case-base enables automated exchange of fault-failure information between operational and supporting departments. The case-base definition is specialised for large-scale diagnosis applications:

In separate data bases, descriptions are defined for symptoms, tests and actions.

• The symptoms are defined with information about alarm levels, operational status and textual explanations.
• Tests are defined with explanations on how to perform the tests, possible results, time and efforts required to perform the tests, operationals status during which the test can be performed, and the operator levels that are allowed or enabled to perform the tests.
• Actions are defined with descriptions about how to perform the actions, what tools and spare parts to use, estimated time and costs, operator level and operational status to perform the action.

The cases are defined as the faults, with features for symptoms, test results and actions. The features for symptoms and tests are defined with factors identifying the uncertainty of the occurance. Alarm levels can also be defined for the faults individually, as a refinement on the alarm levels for symptoms.

1. The diagnosis response in on-line situations should clearly identify actions and faults. The network structure in on-line diagnosis systems should be the result of and induction of symptom-fault-action relations.
2. General domain knowledge and specific examples are defined seperately in data bases. General domain knowledge is defined for symptoms, tests and actions. Fault specific definitions from technical and operational knowledge and experience are clearly seperated in the definition. This includes the definition of uncertainty in occurance of faults, symptoms and the outcome of tests.

Failure stories about previously occuring situations should not be interpreted or adapted in an on-line situation. This is a task for off-line development of the on-line diagnosis systems. The uncertainties in the fault-cases define a problem space around the fault in which experienced problems should lie. Induction of these uncertainties (in any form) and fault case definitions is performed in an off-line case-base development environment. The case-base contains all validated actions and known faults, for which actions and occurances can be verified in the development of the on-line systems. For on-line diagnosis, only actions should be presented that are allowed to be executed under the current operational conditions. Unknown faults or multiple occuring fault situations cannot be distinguished. It is the ultimate operators responsibility to solve a problem and to perform tests and actions when necessary. Automated adaptation of actions and fault cases cannot be predicted if the occuring faults are unknown, and therefore not allowed. During development of the case-base, on the other hand, analysis tools for detectability and coverage of the diagnosis systems should be provided rather than automated adaptation tools.

During diagnosis application, the response time has to be restricted and should be guaranteed for a number of operational applications. This requirement contradicts with standard case-based retrieval engines, on-line adaptation of failure cases and actions. The response time and accuracy requirements concern the retrieval of actions during operational situations. The requirements can only be guaranteerd for a predefined and deterministic search structure. Obviously, a set of search trees for different symptoms are not an efficient structure, and consultation of the original case-base is not allowed. For this reason, BRIDGE automatically develops the search structures for the retrieval task in operational diagnosis systems. The search structure represents all cases and relations in the fault case-base. The search structure is translated in a procedural programme of which the search process and time characteristics can be guaranteed.

The direct outcome of the project will be a new software tool, delivered in fully documented and tested form, available for rapid commercialisation via one of the industrial partners. This project has therefore a very sharp focus with the explicit intention of producing, within a 24 month period, a commercially exploitable software tool for development and maintenance of fault diagnosis systems for large technical systems.

The complete BRIDGE software will have the following features:

• a development environment for unique and simplified description for faults, failures, tests, results and actions;
• development and maintenance tools different diagnosis systems, consisting of
  • a parser for off-line generation of the fault network, and
  • a real-time kernel for on-line diagnosis;
• an environment for on- and off-line diagnosis, with user interfaces and inference engines.