构造了一类小波核函数,并利用小波核函数支持向量机对预应力混凝土碳化深度进行了仿真和预测。将小波函数用于SVM得出一类小波核函数,并将其与径向核函数SVM、神经网络在预应力混凝土碳化深度预测中的应用进行了对比。研究发现,SVM对预应力混凝土碳化深度的预测精度要远远高于神经网络,且小波核函数SVM比径向核函数SVM精度高。小波核函数在一定场合下精度要比径向核函数高,SVM用于预应力混凝土碳化深度的预测是可行的,且具有较高精度。 A kind of wavelet kernel function is constructed and the carbonization depth of prestressed concrete is simulated and predicted by using wavelet kernel function support vector machine. The wavelet function is applied to SVM to obtain a class of wavelet kernel function, and compared with radial kernel function SVM and neural network in predicting carbonation depth of prestressed concrete. The results show that SVM predicts the carbonation depth of prestressed concrete much more than the neural network, and the accuracy of wavelet kernel function SVM is higher than that of radial kernel function SVM. The accuracy of wavelet kernel function is higher than that of radial kernel function in some cases. SVM is feasible for predicting the carbonation depth of prestressed concrete and has higher accuracy.