Abstract: The prediction of reduction ratio for pore closure in heavy forgings stock is limited to the finite element numerical simulation and some complex mathematical formula.This paper proposes a new forecasting method，using the SVM model to predict the critical reduction ratio of pore closure.Several major factors are selected that affecting pore closure as the input features of SVM.SVM’s nuclear parameters with PSO are optimized in order to enhance its accuracy，LIB-SVM toolbox is combined to train a SVM model.The model can predict the critical reduction ratio of pore closure in heavy forging quickly，compared its predictions with computer simulation results，the correlation coefficient reaches almost 85%，it has a good prediction performance.

Key words: pore closure, Particle Swarm Optimization（PSO）, Support Vector Machine（SVM）, critical reduction ratio

摘要： 预测大型锻件内部空洞锻合时的压下率，目前都是局限于有限元数值模拟和一些复杂的数学公式。提出一种新的预测方法，用SVM回归模型预测空洞闭合的临界压下率。选取几个影响空洞闭合的主要因素作为支持向量机的输入特征，用PSO优化SVM的核参数以提高其精度，结合LIB-SVM工具箱，训练出一个SVM模型。该模型可以快速预测锻坯内部空洞锻合临界压下率，将其预测结果与计算机模拟结果相比较，相关系数几乎达到了85%，具有较好的预测性能。

关键词: 空洞锻合, 粒子群优化, 支持向量机, 临界压下率