Decentralized Fault Detection in Building Services by Means of Tensor Decomposed Qualitative Models

Fault detection and diagnosis (FDD) methods are most of the time deployed in buildings as supervisory solutions on a management level or as cloud computing solution. The deployment of Internet of Things devices will enable to embed FDD methods as edgecomputing solutions directly on subsystems like heat pumps or air handling units. This paper shows how qualitative models can be used for fault detection in building services and how tensor decomposition methods can enable their integration on decentralized subsystems as edge-computing solution.