【作者】 张元；

【导师】 王桂兰；

【作者基本信息】 华中科技大学 ， 材料工程， 2011， 硕士

【摘要】 电弧直接制造技术(ADRPM)是一种金属零件直接制造的低成本新方法。该方法材料利用率高,能耗低,设备投资小,成本低,成形效率高,材料制备简单,能够快速响应市场需要,有广泛的应用前景。由于ADRPM制造的零件为全焊缝组织,其技术关键是选择合适的工艺参数保证制造零件的形状尺寸和成形性能质量。电弧直接制造过程十分复杂,影响成形质量的因素众多。为了提高电弧直接制造工艺水平,对电弧直接制造进行过程监测、工艺特点分析、成形质量预报和工艺参数智能优化有着非常重要的意义。本文以电弧直接制造的工艺实验为基础,结合模式识别方法对现有参数样本的工艺特征进行研究,并提出了对工艺参数进行质量预测和智能优化的原理和算法。选择遗传算法设计新的工艺参数指导工艺实验,扩充了工艺参数样本容量,基于主成分法对扩充后的参数样本进行分析,提取了工艺特征,并在此基础上建立了质量预测模型。基于特征抽提法分析各参数对成形质量的影响程度,并结合模式识别调优法提出了线性规划判据,从而确定了参数优化的方向和参数取值范围。本研究对电弧直接成形参数样本进行分析,结果表明:由主成分分析处理,参数样本由5维降为3维,得到的映射图,能够将成形质量优和非优的样本点以一条判别直线明显区分开,充分反映了现有参数样本的工艺特征。根据不同目标量的映射图特点,分别建立了隶属度函数,以宽度稳定性为目标量的隶属度为好区半径与参数对应映射图上的点到好区中心的距离之比,以0.714为阈值,若预报点隶属度值大于0.714则判别为优类,否则为非优;以高度稳定性为目标量的隶属度为参数对应映射图上的点到表示优区的球心的距离与球的半径之比,以0.653为阈值,若预报点隶属度值大于0.653则判别为优,否则为非优。用建立的质量预测模型对3组未知工艺参数的成形质量进行预测,经实验验证,预测值与实验结果相符,进一步说明了质量预测模型的合理性。对各参数对成形质量的影响程度进行了分析。以宽度的标准差为目标量,则影响成形质量的程度从大到小为:送丝速度、频率、机器人行走速度、弧长、脉宽比;以高度的标准差为目标量,则影响成形质量的程度从大到小为:频率;机器人行走速度;送丝速度;枪板距;脉宽比。对工艺参数进行优化调控,依照线性规划判据确定的优化方向和参数取值范围依次设计了3个优化样本点,经实验验证,3组参数对应的成形质量逐步提升,实现了参数优化的功能。综上述结果,说明论文提出的参数样本特征分析方法能迅速挖掘参数工艺特征,在此基础上建立的质量预测模型和工艺优化方案较为可靠,具有一定实用性,可用于指导建立合理的电弧直接制造工艺。更多还原

【Abstract】 Arc-Direct Rapid Prototyping Manufacturing (ADRPM) is a new technique for direct manufacture of metal components. It stands out for its high efficiency of utilizing the materials and shaping the components, as well as its low cost of energy usage and equipment expense. Besides, due to simple material preparation and quick response to the market demands, ADRPM shows its value of various applications in the future. Since the component made by ADRPM is complete welding microstructure, the core technique of ADRPM is to choose proper technological parameters to ensure the shape, size and the formability of the components.The process of ADRPM is very complex, since there are a number of factors which influence the forming quality. To improve the technology level of ADRPM, the process monitoring of ADRPM, the analysis of technology characteristics, the prediction of forming quality and the intelligent adjustment of the main processing parameters have great significance.Based on the technology experiments of ADRPM and combined with pattern recognition, this paper made scientific research on the technology characteristics of existing parameters samples, and proposed the principles and algorithms of the quality predication and intelligent optimization of the technology parameters. The author chose Genetic Algorithms to design new technology parameters so as to guide the technology experiment, which expanded the amount of technology parameters.Then the author analyzed the expanded parameters samples based on Principal Component Analysis, extracted the technology characteristics, and then built a quality prediction model.In addition, the author analyzed each parameter’s impact on the forming quality on account of feature extraction, and proposed the criterion of linear programming combined with pattern recognition evolutionary method to determine the direction of the parameters optimization and the range of parameters. This research analyzed the samples of parameters, and the results indicated: the samples analyzed by Principal Component Analysis have been decreased from 5 dimensions to 3 dimensions, and the mapping obtained from Principal Component Analysis could obviously distinguish the samples in high forming quality from the ones with relatively low quality , which fully reflected the technology characteristics; According to the Organizing Map’s features of different desired value,the record of degrees’ funtion can be established .The record of degree weighing width’s stability is defined as the ratio of the radius of good area and the distance from the point in the Organizing Map to the centre of good area.Setting 0.714 as the standard value,if the forecasting point’s record of degree is greaterthan 0.714 ,then we can judge the point as good point ,or no-good point.The record of degree weighing height’s stability is defined as the ratio of the radius of sphere and the distance from the point in the Organizing Map to the centre of sphere.Setting 0.653 as the standard value,if the forecasting point’s record of degree is greater than 0.653,then we can judge the point as good point ,or no-good point.the research used the quality prediction model to predict the forming quality of three groups of unknown technology parameters, and the experiment verified that the predictive values were consistent with the result of experiment, which further explained the rationality of the quality prediction model; The influence degree from various parametars to forming quality has been analysised .Weighing the width’s standard deviation,parametars can be ranged for the forming quality’s degree from much to little:wire feed speed、frequency、robot’s speed of travel、Gun-board distance、pulse width ratio;Weighing the height’s standard deviation,parametars can be ranged for the forming quality’s degree from much to little:frequency、robot’s speed of travel、wire feed speed、Gun-board distance、pulse width ratio; the research conducted the optimization control of the technology parameters, according to the direction of optimization which was determined by the criterion of linear programming and the range of parameters, three optimization sample points were designed. The experiment verified that the forming quality of the three groups of parameters increased gradually, which achieved the function of parameters optimization. In conclusion, the analysis of the characteristics of parameter samples could excavate the characteristics of parameter samples quickly, and the quality prediction model and technology optimization scheme were more reliable and practical, which could be used to guide to establish reasonable technological parameters in ADRPM .更多还原