Soft computing-based prediction models for interference effect of twin C-shape tall building

Wind interference affects the pressure coefficients (C_p) on tall buildings and wind analysis is typically conducted using a wind tunnel or computational fluid dynamics (CFD). However, identifying the critical arrangement that significantly changes the C_p on the principal building (PB) poses a significant challenge because it requires numerous simulations to be conducted, with various PB and interfering building (IB) configurations. To address this, machine learning (ML) models are introduced here to optimise CFD simulations. The CFD simulations are done for 16 arrangements (based on height and location of IB) of the PB and IB. The CFD simulation results are used in eight ML models for training and testing, including fine tree, medium tree, bagged tree, boosted tree, medium neural network, wide neural network (WNN), bilayered neural network and trilayered neural network. The fine tree model performed exceptionally well. Rank analysis and statistical parameters are used to examine the efficiency of these models. The fine tree is the best-performing model (R²_Training = 0.9872 and R²_Testing = 0.9732), followed by bagged tree (R²_Training = 0.9825 and R²_Testing = 0.9653) and WNN (R²_Training = 0.9671 and R²_Testing = 0.9729). The fine tree effectively identified critical PB and IB locations corresponding to maximum C_p, offering an efficient alternative to traditional wind analysis methods.