基于CMS-SSA-BP模型的混凝土碳化深度预测性能对比

Comparative on Prediction Performance of Concrete Carbonation Depth Based on CMS-SSA-BP Model

为了提高SSA-BP模型的预测准确性,分别使用三类混沌映射序列(CMS)初始化麻雀位置,帮助SSA-BP模型跳出局部极值,从而提高解的质量。利用200组实际混凝土碳化深度试验数据作为运行数据,以黏接剂剂量、粉煤灰置换水平、水胶比、CO_(2)体积分数、相对湿度、暴露时间作为输入变量,混凝土碳化深度作为输出变量,分别运行2次后得出各项指标数值,对比分析三类CMS-SSA-BP模型各自的优化点。结果表明,经过混沌映射序列(CMS)优化的SSA-BP模型预测性能更佳,其中,Tent-SSA-BP模型的预测精度最佳,Logistic-SSA-BP模型的预测稳定性最佳,Sine-SSA-BP模型的收敛速度最快。

In order to improve the prediction accuracy of the SSA-BP model,three types of chaotic mapping sequences(CMS)were used to initialize the sparrow position respectively,which helped the SSA-BP model to jump out of the local extremes,thus improving the quality of the solution.Using 200 sets of actual concrete carbonation depth test data as running data,with adhesive dosage,fly ash replacement level,water-cement ratio,CO_(2) volume fraction,relative humidity,exposure time as input variables,and concrete carbonation depth as output variables,the values of each index were obtained after two runs,and the optimization points of each of the three types of CMS-SSA-BP models were compared and analyzed.The results showed that the SSA-BP models optimized with chaotic mapping sequences(CMS)had better prediction performance.Among them,Tent-SSA-BP model had the best prediction accuracy,Logistic-SSA-BP model had the best prediction stability,and Sine-SSA-BP model had the fastest convergence speed.