Fault Diagnosis using a Combined Model and Data Based Approach: Application to a Water Cooling Machine

In this paper, the problem of fault diagnosis of an industrial water cooling system is addressed using a combined data-driven and model based approach. Using the energy balance equations, the design of the fault diagnosis system is based on structural analysis. As result of this analysis, a set of Minimally Structurally Overdetermine Sets (MSOs) are obtained presenting the desired fault detectability and isolability properties. Since the mathematical expressions of such MSOs are very complex and highly non-linear, and there are an important number of parameters that should be estimated from data, a system identification approach based on machine learning techniques is used. Not only the nominal model but also the error model is estimated. Finally, the proposed approach is tested with the data obtained from a water cooling machine.