Framework for Automated Change Propagation from Building Information Models to Building Energy Models

<p dir="ltr">The design and creation of Building Energy Models (BEM) based on Building Information Models (BIM) are increasingly replacing traditional BEM modeling methodologies. This shift is primarily attributed to the cost-efficiency and time savings associated with BIM-based BEM, which eliminates the need for redundant modeling efforts. However, design changes frequently occur to BIM during the building design phase, and propagating such design changes from BIM to BEM is challenging in current practice. Due to the lack of connection that can transmit change information from BIM to BEM, change propagation from BIM to BEM currently relies on manual operation, which is error-prone and time-consuming. The case study first introduced in this thesis identifies three key engineering challenges for automated change propagation from BIM and BEM: (1) localizing components in BEM based on changes identified in BIM, (2) assessing the impact of BIM changes on BEM, and accordingly (3) devising strategies for implementing the BIM changes in BEM. </p><p dir="ltr">To address these challenges, this study proposes a three-step framework for automated change propagation from BIM to BEM. In this framework, BIM and BEM are represented using graphs where building components are modeled as nodes and spatial or functional relationships between them are modeled as edges. With BIM and BEM graphs, the framework firstly constructs instance-level mapping relationships between BIM and BEM. Subsequently, the framework interprets changes identified by comparing BIM and translates them into corresponding changes in the counterpart BEMs. Lastly, appropriate BEM updating strategies are generated and implemented to finalize the automated model updating process. </p><p dir="ltr">Based on literature review and implementations, this study proposed and answered three research questions corresponding to three steps in the proposed framework. In the step of constructing mapping relationships, this study highlighted that current ways to map BIM and different types of BEM mainly utilize semantic information, but developing invariant signatures from this information cannot be done reliably. Therefore, this study proposed a topology-informed method creating invariant signatures based on topological information and class information for instance-level mapping between BIM and BEM, improving generalizability of instance-level mapping method. In the step of interpreting changes from BIM to BEM, this study identified the issues of lacking training data and low scalability in existing studies. To fill this gap, this study highlights that the essence of BIM-to-BEM change interpretation problem is to evaluate the impact of BIM changes on heat transfer paths among building components in BEM. Furthermore, this study identifies four fundamental scenarios that collectively represent all possible changes to heat transfer paths and proposes three basic rules to address them. In the last step, for automated BEM updating, this study formalized it as a model repair problem using reinforcement learning (RL). Meanwhile, the implementation results indicate that quantified characteristics are missing for guiding the RL agent to select proper actions for BEM updating, therefore this study proposed method quantifying the level of alignment between BIM and BEM based on parameter identification and weights optimization. The validation of the proposed method in each research question is conducted based on a set of matched pairs of BIM and BEM from real-life projects and online resources, and the validation results demonstrated feasibility and effectiveness of proposed methods. </p><p dir="ltr">In summary, this thesis makes the following contributions: (1) A topology-informed method for mapping BIM with different types of BEMs; (2) A context-aware method with high generalizability to interpret changes from BIM to BEM; (3) A reinforcement learning based framework for automated BEM updating. Practical implementation of these contributions can significantly strengthen the interoperability between BIM and BEMs and reduce the reliance on manual labor for BEM updating. Thus, the BIM-based BEM modeling process can be more accurate, scalable and efficient. Future work of this study could extend the scope of the framework to include other types of changes, such as parameter-level change, and other types of targeted simulation models, such as structural analysis models.</p>