基于支持向量回归机的颗粒阻尼减振结构阻尼特性实验

Prediction of Particle Damping Ratio Based on GA-SVR

将基于遗传算法(GA)的支持向量回归机(SVR)用于建立颗粒阻尼减振结构阻尼特性预测模型。应用基于结构风险最小化准则的SVR,建立颗粒阻尼减振结构阻尼特性—影响因素SVR预测模型,对颗粒阻尼减振结构的阻尼特性进行预测,并通过实验进行了验证。结果表明:在选择适当的参数和核函数的基础上,利用该方法建立的预测模型,平均相对误差在10.3%左右;颗粒阻尼器的减振性能随填充率的增加而增加,颗粒填充率75%时,减振性能最好;颗粒密度是影响颗粒阻尼器减振效果的重要因素,颗粒密度越大系统的减振性能越好;颗粒阻尼器的减振性能随着颗粒直径的变化不明显。

In order to research the effective prediction method of particle damping ratio,genetic algorithm-support vector regression machine（ GA-SVR） is used to establish model to predict particle damping ratio. On the basis of GA,we optimize the SVR parameters,apply the support vector regression machine which is based on structural risk minimization criterion to establish the ＂particle damping ratio-influence factors of the SVR forecasting model ＂,and then predict particle damping ratio. Experiment analysis shows that on the basis of selecting the appropriate parameters and kernel function,the particle damping ratio prediction model is established and works well. Based on this method,the average relative error is about 10. 3%.