Digital Twin-Based HVAC Model Generation Using Point Cloud and Image Data for Building Retrofitting

Previous studies have explored the 3D reconstruction of indoor environments from point cloud and image data. However, these studies have mainly focused on the reconstruction of structural elements. There is a lack of research on HVAC system reconstruction. This research aims to develop a method for automatically generating a digital twin-based HVAC model. A custom-labelled dataset of HVAC component images, along with a public image dataset, is used to train a region-based convolutional neural network to identify HVAC components. These identified components are then used to generate the digital twin model with their spatial relationships. The proposed method is validated through a real-world case study using point cloud and image data, demonstrating its potential to streamline the generation of digital twin-based HVAC models. The results confirm that the method is both reliable and efficient for capturing the current state of existing HVAC systems.