LAUSR

Abstract Central heating system faults affect building energy consumption and indoor thermal comfort significantly. The interdependencies among system components and multiple failure modes present a challenge for system health diagnostics and prognostics. A reliable diagnosis and prognosis can only be ensured when all component conditions are monitored with minimum uncertainty. In this regard, sensors should be selected based on their priority in providing system health information. Currently, most of the research on sensor optimization models optimize sensors position and orientation. However, in this study sensor type is optimized as well. In addition, the proposed method is based on the Bayesian network model and considers failure interdependencies of components, which leads to more realistic results. The objective of the optimization model is to obtain lowest entropy on system health information given financial and practical constraints. The proposed methodology is applied to a central heating system as a case study. The Bayesian Network model of the central heating system is constructed based on knowledge regarding the component interdependencies and conservation laws, as well as the gathered historical data. Results indicate that the Bayesian network model represents component interdependencies, and provides valuable information for sensor combination optimization. Sensors are optimized to obtain the minimum system information uncertainty. In addition, the sensitivity analysis on data generated by sensors demonstrates that the proposed method diagnosis system faults correctly, and predicts system health status effectively.