To investigate the damage law and influencing factors of mechanical property degradation of hydraulic con-crete structures after early freezing damage in the cold and arid regions of Northwest China,the indoor accelerated test of early freezing and freeze-thaw cycle of concrete was designed.The deterioration law of freeze-thaw damage of concrete structure with a temperature of-10℃and a freezing moment of 3.5 h was investigated through the freeze-thaw cycle of concrete specimens.The impacts of water-cement ratio,fly ash and air entraining agent on the post-freezing mechanical properties of early-stage damage-induced concrete were studied.The back propagation neural network improved by gray wolf optimization algorithm was used to simulate and predict the mechanical properties and frost resistance of early-stage damaged concrete.The sensitivity analysis of each influencing factor was carried out.The results show that low water-to-cement ratio concrete has significantly better frost resistance,air-entraining agent can improve the early frost resistance,and the optimal dosage is 0.01%;The fly ash has significantly improved the frost resistance of concrete in the late freeze-thaw cycle,and the optimal substitution amount is 20%;Four regression evaluation indexes of the GWO-BPNN model are better than those of the traditional neural network model,and it is able to predict the mechanical properties of early-staged concrete more accurately.It was found that the largest variable affecting the durability of early-stage damaged con-crete was the water-cement ratio,and the smallest variable was the air-entraining agent.
关键词
灰狼优化算法/神经网络/早期带伤混凝土/抗压强度/影响因子分析
Key words
grey wolf optimizer/neural networks/early banded concrete/compressive strength/influencing factor analysis
维普
万方数据