Evaluation of fracture behavior of Yellow River ice based on three-point bending test and PSO-BP model

Abstract

The fracture toughness of river ice is the primary mechanical characteristic that affects the dynamic behaviors of ice such as fracture, overlap, and accumulation. It can provide a scientific basis for the forecast of the break-up date of ice and the prevention and control of ice run. In this work, taking Yellow River ice as the research object, the fracture toughness under different types of ice (columnar ice and granular ice), temperatures (-2℃, −4℃, −8℃, and −10℃), and strain rates (10-6 s−1 ∼ 10-2s−1) was measured by three-point bending test. And the fracture toughness of the Yellow River ice was predicted and analyzed by back-propagation (BP) model and particle swarm optimization (PSO)-BP model. The experimental results exhibit that the fracture toughness of Yellow River ice mostly ranges from 30 kPa·m1/2 ∼ 130 kPa·m1/2. The fracture toughness of columnar ice and granular ice decreases with the increase of strain rates and temperatures, and the fracture property of columnar ice is better than that of granular ice. The probability of predicted values of train set and test set for BP model and PSO-BP model within 20 % error is 78.57 % and 85.71 % respectively. The analysis of the evaluation indicator presents the prediction performance of the model for the fracture toughness of Yellow River ice is promoted using PSO optimized BP neural network.