DeepBIM

Towards Automatic Extraction and Prediction of Annual Heating and Cooling Demand from Floor Plans

Creation of 3D floorplans with roof and floor :

• Follow the setup and requirements in revit-batch-transformation, then run the Revit Addin to create all the Revit files
• A Success file is created indicating the floorplans where roofs and floors have been added correctly

Batch Analysis :

• In EnergyReports, use Files.py to rename all the successul files (change local path)
• Gather all these files with different names into one folder manually
• For each file :
  • Open Revit
  • Open Dynamo and load SetSpaces.dyn to automatically change above level and upper limit of spaces, let it run in background : need of archi-lab package
  • Place spaces automatically with the tool
  • Run heating and cooling load (change the type of materials/profile of building and HVAC sytem used if wanted)
  • Get the temporary htm file containing the report created at file:///C:/Users/"UserName"/AppData/Local/Temp and transfer it into a common folder (only existing when the report is visualised in Revit)
  • Close the report and floorplan
• Repeat for 400 files
• When the folder of html reports is complete, in EnergyReports, run parser_main.py to extract all the information (parameters and heating and cooling load) and write the dataset in bim_train.csv (change local path)

Model : Training and Predicting new heating and cooling loads :

• RandomForest : run train_predict.py to train and test the model on the dataset, and predict new heating and cooling loads for new values put in data/bim_prediction.csv (for further use)
  • Requirements : scikit-learn
• KerasRegressor : run train.py to train and test the model for each load. The best models are then saved and can be loaded to predict new heating and cooling loads with predict.py , for new values put in data/bim_prediction.csv (for further use)
  • Requirements : Tensorflow, Keras