Developing case-based reasoning systems: a case study in diagnosing building defects

Fault diagnosis is a classic expert systems (ES) domain and thus the diagnosis of building defects would seem an ideal domain for an ES. However, the diagnosis of building defects is complicated by the following factors: numerous materials are used in numerous combinations within a building, similar symptoms can be caused by different problems, there is a limited causal information for why materials fail, and the cause of a defect may not be discovered until destructive tests are performed. All of these factors suggest that implementing a diagnostic ES for this domain would be non-trivial. This paper presents a possible method of overcoming these problems by using case-based reasoning (CBR). This paper contrasts the CBR paradigm with that of model-based reasoning (MBR), it describes characteristics that make a problem domain suitable for CBR, and describes the implementation of a CBR system highlighting the good and bad features of the software tool used.