【作者】 曹宏强；

【导师】 赵军； 官英平；

【作者基本信息】 燕山大学 ， 材料加工工程， 2023， 博士

【摘要】 铝合金型材滚弯件在航空航天、汽车、船舶和电厂等主体设备中应用广泛,目前,定曲率滚弯件的生产制造技术已趋于成熟,但变曲率滚弯件由于易出现回弹量难以预测、截面易畸变、表面起皱等问题,导致制造精度难以控制,是滚弯件制造技术领域尚待解决的瓶颈问题。本文基于对型材滚弯成形中回弹量与影响因素之间的内在关系深入研究的基础上,采用数据挖掘的方法建立影响回弹量与各因素之间的关联关系,建立型材滚弯成形回弹预测模型,开发型材滚弯成形工艺控制智能分析系统,目的是提升滚弯零件的制造精度和滚弯装备的智能化水平。对型材变曲率滚弯工艺进行研究,通过理论解析得到了回弹量的计算公式,给出了用离散法解决变曲率滚弯回弹量的预测方法。建立了型材滚弯数值模拟仿真模型,分析了摩擦系数、辊轮转速等工艺参数对型材滚弯质量的影响,得到了铝合金型材变曲率滚弯合理的工艺参数。设计制造了滚弯机并开发了实验控制系统,进行了铝合金型材变曲率滚弯实验,建立了数据测量系统,对于实验数据进行了处理,得到了不同条件下的实验数据集,在此基础上建立了数据挖掘模型所需的样本库,为搭建变曲率滚弯回弹预测模型提供了数据支撑。分析了材料参数、几何参数和工艺参数对回弹的耦合作用和型材滚弯成形工艺机理,对适用于型材滚弯回弹预测的数据挖掘算法进行了研究。分别建立了定曲率和包含变曲率影响因素的型材滚弯成形回弹预测模型,用XGBoost和图遗传编程算法构建了定曲率滚弯成形回弹量预测模型,并在定曲率模型的基础上,设计了基于XGBoost和条件随机场的变曲率滚弯成形回弹量预测模型,并开发了新型算法,解决了传统算法难以精确计算变曲率滚弯回弹量的问题。将预测结果与实验结果进行对比发现,预测模型的精度随着数据量的增加不断提升,证明算法具有较高的智能化水平。提出了包括时间依赖关系、空间依赖关系和动态演变关系的多维特征空间表征和度量方法,设计了滚弯成形多维特性的空间分析器和时间分析器,开发了滚弯成形控制工艺智能分析系统,实现了变曲率型材滚弯回弹的精确智能预测,并成功应用于通用滚弯成形设备,提高了设备的智能化水平。更多还原

【Abstract】 Aluminium alloy profiles roll is widely used in the aerospace,automotive,Marine and power plant equipment.At present,the constant curvature roll bending parts manufacturing technology has been mature,but the change of curvature roll bending parts due to easy to appear the springback is difficult to predict,cross section distortion and surface wrinkling problems such as easily,cause manufacturing precision is difficult to control,is roll manufacturing technology bottleneck problems remain to be solved.Based on the in-depth study of the internal relationship between the elastic amount and influencing factors in the rolling bending of profiles,this paper uses the method of data mining to establish the correlation between the influencing elastic amount and various factors,establish the elastic prediction model of the rolling bending of profiles,and develop the intelligent analysis system of the rolling bending process control.The purpose is to improve the manufacturing accuracy of rolling parts and the intelligent level of rolling equipment.This paper studies the rolling and bending process of the profile with variable curvature,obtains the calculation formula of the springback through theoretical analysis,and gives the prediction method to solve the springback by discrete method.The numerical simulation model of profile roll bending was established,and the influence of process parameters such as friction coefficient and roller speed on profile roll bending quality was analyzed,and the reasonable process parameters of aluminum alloy profile roll bending with variable curvature were obtained.The rolling bending machine was designed and manufactured and the experimental control system was developed.The rolling bending experiment of aluminum alloy profiles with variable curvature was carried out,and the data measurement system was established.The experimental data sets under different conditions were obtained by processing the experimental data.On this basis,the sample library required for the data mining model was established,providing data support for the construction of the elastic prediction model of rolling bending with variable curvature.The coupling effect of material parameters,geometric parameters and process parameters on springback and the process mechanism of rolling bending were analyzed,and the data mining algorithm suitable for prediction of rolling bending springback of profiles was studied.The elastic prediction models of roll bending with constant curvature and those with variable curvature were established respectively.The elastic prediction models of roll bending with constant curvature were constructed by using XGBoost and graph genetic programming algorithm.On the basis of the constant curvature model,the elastic prediction models of roll bending with variable curvature were designed based on XGBoost and conditional random field.A new algorithm is developed to solve the problem that the traditional algorithm is difficult to accurately calculate the rebound of rolling with variable curvature.By comparing the prediction results with the experimental results,it is found that the accuracy of the prediction model is continuously improved with the increase of the amount of data,which proves that the algorithm has a high level of intelligence.A multi-dimensional characteristic space characterization and measurement method including time dependence,space dependence and dynamic evolution relationship was proposed,a space analyzer and time analyzer of multi-dimensional characteristics of roll bending was designed,and an intelligent analysis system of roll bending control process was developed to realize accurate and intelligent prediction of roll bending rebound of variable curvature profiles,and was successfully applied to general roll bending forming equipment.Improve the intelligent level of equipment.更多还原