Optimized and reliable operation of heating, ventilation, air conditioning, and refrigeration (HVAC) systems with regard to maintenance and energy management through predictive, data-driven, and self-learning fault detection. Conceptual design and prototype implementation of an AI (Artificial Intelligence) tool for automated data analysis and recommendation generation for technical building management.
Initial Situation/Motivation
Throughout the building lifecycle, up to 70% of total costs arise during the operational phase, making this stage the largest opportunity for economic optimization. Technical building equipment (TBE) plays a significant role in the construction, inspection, maintenance, and repair costs of a building. Additionally, it accounts for a substantial portion of annual energy consumption and, consequently, a significant share of CO2 emissions.
Content and Objectives
The mAIntenance project aims to reduce both energy and maintenance costs of HVAC systems by leveraging predictive and self-learning algorithms, while simultaneously achieving more efficient and reliable operations. On the system level (combined analysis of building and HVAC systems), time-series forecasting with neural networks will be used to account for future needs to meet building heating or cooling loads within the control strategy. On the component level of technical building equipment, abnormal behavior will be detected, analyzed, and highlighted through modeling and machine learning.
Methodological Approach
The project involves the development of a mockup for an AI-supported tool for fault detection and diagnosis, as well as its functional validation through prototype implementation in an FM control room. This allows the building manager or maintenance technician to be supported with data-driven recommendations for action.
Expected Results
The insights gained will provide information on the performance of the chosen self-learning algorithms in conjunction with minimal preparation effort for energy management, maintenance, and repair processes. Additionally, conclusions will be drawn regarding the required operational monitoring (e.g., number of data points, measurement periods, etc.). Furthermore, transfer learning approaches will be implemented to explore innovative data-driven competencies for digital building operations in cases of missing or insufficient datasets.
Project Framework
The project is being conducted as industrial research by AIT Austrian Institute of Technology GmbH in close collaboration with PKE Facility Management GmbH. The initiative aligns with the primary goal of developing predictive energy and maintenance services for technical facility management to enhance energy and resource efficiency. The development of innovative business models for marketing digital "Technical FM Services" is also a key aspect of this initiative.
