基于KNN算法的中心带孔圆板拉深-翻孔变形方式的研究

Study on drawing-flanging deformation mode in circular plate with center-hole based on KNN algorithm

为了准确、快速判别中心带孔圆板拉深-翻孔过程中的变形方式,选取第3代先进超高强钢QP980材料,以预制孔直径、凹凸模圆角半径和板料厚度为特征参量,通过AutoForm成形仿真获取样本数据集,以Python作为编程语言,基于机器学习中的KNN算法构建中心带孔圆板变形方式预测分类器,并利用随机数据集验证其准确度。结果显示,当K=3、p=14且考虑距离权重时,该分类器的预测效果最佳,总体分类准确率达到90.2%,对随机数据集的分类准确率为88.9%。对比传统理论计算和计算机仿真预测,该分类器能以较高的准确度在短时间内同时预测多组样本,能够为实际生产提供参考和借鉴。

In order to discriminate the deformation mode of the circular plate with center-hole in the drawing-flanging process accu-rately and quickly,for the third-generation advanced high strength steel QP980 material,taking diameter of pre-manufactured hole,fillet radius of punch and die and sheet thickness as characteristic parameters,the sample datasets were obtained by forming simula-tion software AutoForm,and Python was used as the programming language.Then,based on KNN algorithm in machine learning,a deformation mode predictive classifier of the circular plate with center-hole was built,and its accuracy was verified by a random datasets.The results show that when K=3,p=14 and the distance weight are considered,the classifier has the best prediction effect,the overall classification accuracy reaches 90.2%,and the classification accuracy of the random datasets is still 88.9%.Compared with the traditional method of theoretical calculation and the computer simulation prediction,the classifier is able to simultaneously predict multiple groups of samples in a short time with high accuracy,which can offer experience and reference for actual production.