There are several types of objects to formalise and supervise the knowledge. The most important are: diagnosis objects, question-class objects and rule objects.

The different diagnostic objects are defined and organised in the diagnostic hierarchy. They are represented in a tree structure which enables to distinguish them either in roughly or in finer diagnosis in arbitrary depth. They represent all the given diagnosis for possible error sources in solar installations. Concerning diagnostic objects many attributes can be set, e. g. the „a priori
frequency" or a proposal for the resolution of an error.

We have eventually found about 60 different diagnosis of possible errors. The hierarchy of question classes includes all question classes and the corresponding question objects. We tried to enter all possible and relevant error symptoms and we reached about 60 symptoms in 25 question classes.

Rule objects link up answers to the questions with their possible solutions, e. g. in very simple form such as “IF A AND B THEN C”, or also such bounds as “N from M” in arbitrary depth. Each kind of classification has its own rule objects.

Out of a number of approximately 60 different diagnoses, about 150 rules in heuristic and safe classification are necessary.

Application

After the start of D3 a dialog interface appears. On this interface the user interacts with the system, which answers questions to indicate the characteristics. First the user must indicate a rough symptomatology of the problem.

Depending on the evaluation given by the system, further question classes appears.

When desired, the user can find help for each question, or the cause of the current question class. If all questions are answered, the system evaluates the characteristics and indicates to the user the most plausible solution(s).

The user can consult the different propositions of solution given by the computer thanks to graphs or other special figures so that the diagnosis is easier to understand.

3. Conclusion

In its present structure the solar expert system enables users to easily find error sources thanks to the knowledge of solar installations which are put on disposal.

In the future the collection of the characteristics will be completed with explanations, pictures and propositions of solutions so that even a user with a few or without knowledge in solar installations can use the system.

We have also planed to integrate the solar expert in a WWW-browser which offers the advantage that changes and actualizations in the knowledge basis are more available. Thanks to this, the use of the system will become much easier. Presently we are just testing the second version of the software. After its probation it should be free-of-charge distributed.