论文部分内容阅读
在管材数控(NC)弯曲过程中,可能出现起皱、过度减薄的质量缺陷,同时会不可避免地发生回弹,都将严重影响成形质量。为了对数控弯曲成形质量进行预测,提出了使用有限元模拟与机器学习相结合的方法,并建立了快速的成形质量预测方法。首先,建立了有效的管材数控弯曲的参数化有限元模型,在工艺参数取值范围中随机选择进行大量的模拟实验作为样本,完成学习数据的挖掘。随后,基于径向基函数(RBF)神经网络建立壁厚减薄与回弹程度的预测模型并使用支持向量机(SVM)建立管材起皱的预测模型。最后,使用模型对新的实例进行预测,并利用模拟与数控弯曲实验对预测模型进行验证。该方法可以对大直径薄壁管材数控弯曲质量进行有效的预测,提高弯曲管件零件设计效率。
During tube bending (NC), wrinkling, excessive thinning of the quality defects and inevitable rebound may all be caused, which will seriously affect the forming quality. In order to predict the quality of NC bending forming, a method combining FEM with machine learning is proposed, and a rapid method of forming quality prediction is established. First of all, an effective parametric finite element model of CNC bending was established. A large number of simulation experiments were randomly selected as the samples in the range of process parameters to complete the learning data mining. Subsequently, a prediction model of wall thickness reduction and rebound degree was established based on Radial Basis Function (RBF) neural network and a prediction model of tube wrinkling was established by using Support Vector Machine (SVM). Finally, the model is used to predict the new instance, and the simulation model and numerical control bending test are used to verify the prediction model. The method can effectively predict the NC bending quality of large-diameter thin-walled tubes and improve the design efficiency of the curved tube parts.