Fault diagnosis of air conditioning systems based on qualitative bond graph

The bond graph method represents a unified approach for modeling engineerin g systems. The main idea is that power transfer bonds the components of a s ystem. The bond graph model is the same for both quantitative representatio n, in which parameters have numerical values, and qualitative approach, in which they are classified qualitatively. To infer the cause of faults using a qualitative method, a system of qualitative equations must be solved. Ho wever, the characteristics of qualitative operators require specific method s for solving systems of equations having qualitative variables. This paper proposes both a method for recursively solving the qualitative system of e quations derived from bond graph, and a bond graph model of a direct-expans ion, mechanical vapor-compression air conditioning system. Results from dia gnosing two faults in a real air conditioning system are presented and disc ussed. Occasionally, more than one fault candidate is inferred for the same set of qualitative values derived from measurements. In these cases, addit ional information is required to localize the fault. Fault diagnosis is ini tiated by a fault detection mechanism which also classifies the quantitativ e measurements into qualitative values; the fault detection is not presente d here. (C) 1999 Elsevier Science S.A. All rights reserved.