基于SVM的船体外板冷弯回弹预测模型研究

Research on prediction model of hull plate bending springback based on SVM

研究方形压头非对压成形装置成形原理的基础上,分析准确预测回弹对板材数控冷弯成形的意义。针对预测样本较少的情况,采用支持向量机技术进行回弹预测分析,并对模型中的核函数及其损失函数参数ε和惩罚因子C进行分析研究,建立板厚与回弹量的预测模型,并将其预测结果与试验结果及采用BP神经网络预测结果进行比较验证其准确性,为船体外板数控成形加工解决回弹问题提出新的思路。

Accurate springback prediction is very important for NC bending machine by studying on the forming principle of the square head non symmetric bending forming device. According to the prediction samples is less, the support vector machine （SVM） technology was used to the springback prediction. And the kernel function and loss function parameter 6 and penalty factor C in the model were analyzed. Prediction models about plate thickness and springback was established. And the predicted results were compared with the experimental results and the prediction results got by BP neural network to verify the accuracy. Put forward a new idea for the hull plate CNC forming to solve the springback problem.