Enhancing BIM with Bayes Point Machine from Early Design Phases for HVAC Maintenance-Oriented Design

This study presents the application of a BIM-based methodology for assessing the maintainability of HVAC systems from the early design phases. The research aims to limit the negative impact of design decisions that overlook maintenance, thus affecting a building’s quality and costs. The BIM model is enhanced with a Bayes Point Machine (BPM) evaluator, which develops a probabilistic model to support the design team in making more effective decisions. The BPM is trained on data from technical specifications of HVAC components (e.g., frequency of maintenance), real-world settings (e.g., height of installation), and implicit knowledge of technical operators (e.g., level of difficulty of tasks). The field of application is HVAC design for existing and heritage buildings, which represents the most challenging activity due to spatial constraints and system complexity. By iterating the BPM, we can systematically analyse and predict the ease of maintenance for each proposed design solution. This allows for proactive adjustments to enhance feasibility, ensure cost-effectiveness, and reduce operational rework. The combination of BPM and BIM improves decision-making during the design process and contributes to the long-term sustainability and performance of HVAC systems in existing buildings. The methodology was tested using a small model based on an existing library room inside the Sapienza University of Rome. Future developments will focus on enhancing the dataset with more data from interviews with HVAC experts.