Establishment of a combined predictive model for accurate predicting the wind drag coefficient at car tail

The optimization of car appearance is the key to reduce energy consumption and increase the energy efficiency. This study examines the wind resistance performance of a specific CAR vehicle model through the use of numerical simulation, fluid theory analysis, tail flow field analysis, critical parameter selection, and prediction model design. By independently establishing models based on radial basis function neural network (RBFNN) and particle swarm optimization based least squares support vector machine (PSO-LV-SVM) for predicting rear wind resistance of automobiles, a combined wind resistance prediction model using BP neural network optimization can be designed and established. The findings demonstrate that, while the accuracy of the combined model is increased to 96% due to the BP neural network's clever adjustment of the single model's weight, the prediction model based on a single approach is only 94% and 93.5% accurate, respectively. Additionally, the accuracy of the combined prediction model has increased to 98% by refining the genetic algorithm to maximize the BP neural network's starting weight values. The combination prediction model proposed in this article can accurately predict the drag coefficient at the rear of a car during driving, which is beneficial for designing and improving the external structure that reduces car energy consumption.