Prediction for Mechanical Performance of Cold Rolled Ribbed Steel Bars Based on BP Neural Network with Dividing Variable Space According to Original Materials' Tensile Strength

摘要

This paper proposes a predictable method for mechanical performance of cold rolled ribbed steel bars based on BP network with dividing variable space according to original materials' tensile strength. It builds a sample variable space partitioning model according to the original materials' tensile strength. It also studies the performance prediction of cold rolled ribbed steel bars based on the 4-in & 1-out BP network and the performance prediction of cold rolled ribbed steel bars based on the 4-in & 2-out BP network. The results show that this method can reliably predict the mechanical performance of cold rolled ribbed steel bars, and the predictive effect of the 4-in & 1-out BP network model based on dividing variable space according to original materials' tensile strength is superior to the 4-in & 2-out BP network model.