基于粒子群优化支持向量机的地下洞室支护设计

Design of Underground Powerhouse Cavern Support based on Particle Swarm Optimization Support Vector Machine

水电站地下洞室支护设计因其环境复杂性而面临重大挑战,现有方案受限于主观经验和低精度等问题,难以满足设计需求。为提高地下洞室设计效率和可靠性,通过引入粒子群优化(PSO)优化支持向量机(SVM)参数,开发地下洞室支护智能设计模型。模型将洞室跨度、洞室高度、洞室高跨比、洞室埋深、围岩类别、岩石饱和单轴抗压强度、最大主应力值、岩石强度应力比作为输入指标。通过对100个国内外水电站地下洞室支护案例的训练测试。结果表明:该模型在各项输出指标上显示了高度准确性,其中喷混厚度、锚杆直径、锚杆间排距的定类准确率分别达到90%、85%、90%,锚杆长度的定量预测拟合优度为0.843。研究成果可为地下洞室支护设计提供一种新方法。

The support design of underground powerhouse caverns in hydropower stations faces major challenges due to the complexity of the underground environment.Existing solutions are limited by subjective experience and low accuracy,making it difficult to meet design requirement.In order to improve the efficiency and reliability of underground powerhouse cavern design,an intelligent design model for underground powerhouse cavern support is developed by introducing particle swarm optimization(PSO)to optimize support vector machine(SVM)parameters.The model uses the cavern span,cavern height,cavern height-to-span ratio,cavern burial depth,surrounding rock type,rock saturated uniaxial compressive strength,maximum principal stress value,and rock strength-to-stress ratio as input indicators.Through the training and testing of 100 domestic and foreign underground hydropower station cavern support cases.The results show that the model shows a high degree of accuracy in various output indicators,among which the classification accuracy of spray-mix thickness,anchor diameter,and anchor spacing reaches 90%,85%,and 90%respectively,and the quantitative prediction goodness of fit of anchor length is 0.843.The study can provide a new method for underground powerhouse cavern support design.