基于SVM的自密实水泥土强度预测模型研究

Research on Strength Prediction Model of Self-compacting Cement Soil Based on SVM

自密实水泥土是一种靠自重就可以自密实的新型填筑材料,可以解决管道沟槽难以压实问题.为了确定特定强度的最佳配合比,通过15组配合比试验数据代入支持向量机(SVM)预测模型中训练,得到自密实水泥土强度与水泥掺入比、含水量之间的关系,并分别用网格寻优算法(GS)、粒子群算法(PSO)、模拟退火算法(SA)对模型参数进行优化,并采用10组新的配合比试验进行验证.结果表明,用SA算法优化的SVM预测模型中,预测结果和实际结果的均方误差最小,准确率最高,可以为配合比设计提供参考范围,大大缩减试验工作量.

Self-compacting cement soil is a new type of filling material that can be self-compacted by self-weight,which can solve the problem that pipeline grooves are difficult to compact.In order to determine the optimal mix ratio for a specific intensity,the relationship between the strength of self-compacting cement soil and the mixing ratio and water content of cement was obtained by substituting 15 sets of mix ratio test data into the support vector machine(SVM)prediction model.The model parameters were optimized by grid optimization algorithm(GS),particle swarm algorithm(PSO)and simulated annealing algorithm(SA),and 10 groups of new mix ratio experiments were used for verification.The results show that in the SVM prediction model optimized by SA algorithm,the mean square error between the prediction result and the actual result is the smallest,and the accuracy rate is the highest,which can provide a reference range for the mix ratio design and greatly reduce the test workload.