Automatisierte Klassifikation und Bewertung von Betriebszuständen raumlufttechnischer Anlagen

Behind the operation of heating, ventilation and air conditioning (HVAC) systems hides due to operational faults often great savings potential. Fault detection and diagnostics in the field of air handling play an important role. An algorithm for automated fault detection and classification of operating states is developed following the approach of U. Habib [1] and M. Bär [2]. This algorithm uses the hourly average of the operating data and classifies the operating hours into different clusters by using hierarchical clustering. The clusters correspond to different operating conditions or faults. This algorithm is applied to three HVAC systems with similar structure in a shopping center. The 4 expected operating conditions were identified, and a few operational faults were coordinated, namely the time and location at which the errors occur. In addition, the robustness of the algorithm with incomplete sensor equipment and the influence of the normalization procedure on one system of the three were examined. The algorithm works very robustly in the absence of unimportant sensors or with different normalization methods