Generation of a digital twin application model for construction projects

Thesis study aims to generate a digital twin model for construction projects. In the first phase of this study, the digital twin models were generated during the project's lifecycle and then validated using a questionnaire survey. Then, the study was implemented in a reinforced concrete beam designed for the laboratory experiment. Moreover, the strain happened in the reinforced concrete beam due to the applied load, temperature, and relative humidity data collected using sensors in real-time. The model was established using mathematical concepts and machine learning algorithms. Neural network, NARX, NARNN, and regression models were used in this study to establish the digital twin model. Furthermore, the study uses a two-phase analysis method. The purpose of the first phase is to evaluate the models' efficiency. At this phase, the dataset has been divided into seventy percent as a training set and thirty percent as a testing set. The analysis was done in the six-week cascade phase over each week separately and then over the six weeks. Moreover, the analysis was done over one week, 15 days, one month, two months, three months, and six months of the stepless phase. In the second phase analysis, the models were implemented to forecast the values of the strain during the next week by training a set of one week, a set of six weeks, and a set of three months. The results of training a one-week dataset show that GPR was the best model with R2 of 0.67508. NARX represents the best results during training a six-week dataset with R2 of 0.816135. While training a three-month dataset, neural network and NARNN indicate the best results with R2 of 0.879067 and 0.880006 respectively. The results indicate that increasing the training dataset size leads to an increase in the accuracy of the testing results. Finally, considering factors other than applied load, temperature, and relative humidity while forecasting the strain's future value may lead to more accurate results.