Silvio Savarese, Martin Fischer
Worldwide governmental imperatives impose new energy performance goals to facility owners that require increasingly accurate Building Energy Performance Simulations (BEPS), even in post-occupancy stages. Advances in BIM and Energy Systems promise a seamless integration of one into the other, however actual applications showcase that there is still a wide gap to bridge between the two data representations (BIM vs BEPS model). Current practice is limited by the non-standardized way BIMs are developed, tailored to fit the construction industry purposes. In addition, most of these BIMs do not reflect the as-built status of the facility. We propose to automate the geometric modeling of BEPS by leveraging existing approaches that generate point cloud based BIMs, and specifically, the Building Parser system. Due to the automatic nature of the BIM generation, modeling conventions can be incorporated into the process, ensuring a simple conversion step to BEPS with minimum human intervention. This will allow the establishment of a process from scan to BIM to BEPS, that can be easily applied to existing facilities, even without prior documentation.
BEPS models are often developed for new buildings as a means of showing compliance with whatever code or regulation is required in that geographical location. In many cases, the models are discarded after that compliance purpose has been served. The models’ accuracy is very seldom tested post-occupancy. When it is tested, it is usually done as part of an academic exercise and commonly shows the model to forecast about 50% of the actual energy usage, post-occupancy. Furthermore, the cost associated with their development is high. All of these factors have combined to give the impression of a lack of quantified accuracy and reliability.
Current practice for energy simulation is based on manually converting an existing BIM to a BEPS model, which isn't an easy process. It is indicative that time savings of 75% for the creation of building geometry in small to medium buildings can be achieved through the appropriate application of automated processes.
1. Develop a method to automatically populate a BIM suitable for input to energy simulation systems with the geometry of a complex commercial building, quickly and at low cost, using the Building Parser system.
2. Test the accuracy of the geometry importation from the Building Parser in generating a BEPS model for a commercial building.
As indicated in the workflow diagram below, the raw point cloud data of the buildings is collected which is converted into the BIM model with the help of the Building Parser system (developed here at CIFE). The BIM model is built taking care of the appropriate modeling conventions (for energy simulation software).
Further, the thermal, material and HVAC properties can be added to the model to develop an input model for the BEPS software as indicated by the last block of the workflow. As a first step of this research and for this project, the above properties will be manually entered.
1. The research is expected to provide robust test methods to examine the accuracy of the geometric and semantic output from the Building Parser BIM, in terms of a BEPS input.
2. The work will also provide confirmation of the Building Parser’s geometry export module’s ability to enable a significant reduction in the effort required to develop a calibrated and best design BEPS model. This will have implications for many facets of the Energy Simulation community.
3. Given the breakthrough nature of the Building Parser itself, it is important that several test cases geared towards practical use, are implemented.
4. We believe this may be the first instance of an automated internal geometry system directly populating a BIM of a large-scale building with accurate geometric and semantic as-built semantic data readily tailored for energy simulation purposes. This will have implications for the global future use of BIM, particularly in existing buildings.
5. Likewise, if we are successful in building a simplified and accurate BEPS model from a correctly configured BIM, it may have significant effects on the uptake of energy modeling in existing buildings.
6. In addition to the practical value of this project, we expect that the findings from specifying and developing interface modules can be useful to standardize the automatic generation of BIMs from visual data, as well as their conversion to a valid and suitable representation for BEPS.
We have currently met the majority of our research objectives and verified most expected results:
An algorithm has been developed that leverages the output of 3D object detection algorithms on indoor point clouds to automatically generate a 3D model that is suitable for BEPS purposes. The algorithm addresses the geometric and semantic automatic population of such models.
Qualitative analysis on several complex commercial buildings showcased promising results and a robust and fast mechanism.
Appropriate quantitative methods are currently being explored. All analysis includes comparisons to ground truth data, as well as currently industry developed BEPS models.
To further enhance the understanding of the accuracy and limitations of the developed algorithm, we will perform additional tests on new point cloud data that have substantial differences to the ones in our current databases and on which the algorithm was never fine-tuned on.