【作者】 程联军；

【导师】 赵国群；

【作者基本信息】 山东大学 ， 材料加工工程， 2008， 博士

【摘要】 汽车工业是国民经济的支柱产业,作为制造汽车受力件和关键件主要手段的锻造工艺在汽车零部件制造中占有重要地位。转向节是汽车上的重要零件,品种多,用量大,目前主要通过锻造工艺进行生产。转向节是汽车锻件中最难生产的锻件之一,其锻造设计水平代表了汽车锻件的最高设计水平。转向节在锻造分类上属于枝杈类锻件,此类锻件锻造工序多,锻造工艺复杂,对其锻造工艺与模具设计要求高。目前,转向节锻造生产面临的主要问题是:锻件材料利用率低,锻造能耗高;转向节新品开发主要依靠设计者的经验,开发周期长,费用高,风险大。针对以上问题,本文重点研究转向节的锻造工艺设计和锻造智能化设计。通过有限元分析和实验研究相结合的方法,以省材节能为目的,进行转向节锻造工艺优化与开发;开发转向节锻造设计专家系统,用来指导转向节的锻造工艺与模具设计,实现转向节新品的短周期、低费用和高质量的开发模式。本文采用有限元软件DEFORM-3D对典型转向节锻造过程进行了数值模拟。通过锻造过程金属流动分析、锻件和模具应力/应变场分析、锻件温度场分析和成形过程力能曲线分析,获得了转向节锻造过程的成形规律。同时分析了摩擦因子等工艺参数对转向节锻造过程的影响。通过以上研究,给出了转向节锻造过程中可能出现的缺陷及原因,为转向节锻造工艺优化与模具设计提供指导。本文开发了一种复杂枝权类锻件的锻造新工艺,能够实现该类锻件在热模锻压力机上的一火锻造。通过数值模拟和实验相结合的方法,以SK2锻件为例,设计了合理的制坯工序,开发了封闭飞边闭式模锻工艺用于预锻工序,实现了该锻件的一火锻造,降低了能耗。本文同时针对长杆类转向节卧式锻造工艺进行研究,以SK1锻件为例,开发了一种新的锻造工艺。该工艺采用直接挤出杆部的制坯工艺,使预制坯的材料分配更为合理;同时开发了控制飞边闭式模锻工艺用于预锻工序。新工艺显著提高了材料利用率。本文针对转向节锻造设计神经网络的关键技术进行研究。提出了单元和形状因子的概念,用来描述锻件特征,建立锻件特征信息模型。所建锻件特征信息模型既包含了特征尺寸,又包含了特征形状。为建立转向节锻件不同特征形状对锻造过程影响的统一评判标准,定义了六种单元类型,并对不同类型单元的成形过程进行了研究。在保证体积不变的前提下,建立了不同类型单元的成形模型,通过对比成形过程最大打击能量,得到了不同类型单元对成形过程影响的程度。在此基础上,计算得到单元形状因子的大小,通过形状因子对单元尺寸的修正得到锻件特征信息。将转向节锻件和模具特征信息分别作为神经网络的输入和输出向量,建立了转向节锻造设计神经网络模型,该模型不仅考虑了锻件特征尺寸对成形过程的影响,也考虑了特征形状的影响,同时将特征形状的影响数值化。本文同时针对转向节锻造设计神经网络的类型、控制参数和训练算法进行了研究。研究了在不同网络类型、控制参数和训练算法下神经网络的收敛情况和预测精度,给出了确定神经网络参数的建议。本文研究了神经网络专家系统的基本结构、知识表示和推理方式,在此基础上,将转向节锻造设计理论和经验知识同人工智能理论有机地结合在一起,开发了汽车转向节锻造设计专家系统FDesign-SK。研究了三十多组不同型号的转向节锻造实例,总结了其中的设计经验,并将这些经验知识存储在系统中。该系统实现了神经网络和专家系统的协同式集成,克服了传统的专家系统具有的推理单调性、知识获取比较困难及不存在统一的知识表示方法等缺陷。该系统采用模块化设计,界面友好,操作简单。系统采用开放式设计,具备良好的自学习功能,可以根据需要不断扩充和完善训练样本库,用来扩大系统的应用范围,提高预测精度和稳定性。本文最后通过锻造实例,给出了利用FDesign-SK系统、数值模拟和反求技术相结合的方法,进行转向节锻造设计。该方法是一种有效的锻造设计方法,可以大幅缩短转向节新品开发周期,降低开发费用,减少设计人员的工作量。更多还原

【Abstract】 The automobile industry is the pillar industry in national economy. The forging process, which is the main method of manufacturing the key and the load-bearing automobile parts, plays an important role in the manufacture of automobile components. The steering knuckle, with many varieties and usage quantities, is mainly made from the forging process now. The steering knuckle is one of the most difficulty automobile forging parts, and the forging design level of the steering knuckle is regard as the highest of automobile forging parts. In the forging classification the steering knuckle belong to the category of fork parts. The forging process of the fork parts is muti-stages and complex forging process, so the forging design of the fork parts is very difficult.At present there are two problems in the forging manufacture of the steering knuckle. The forging process of the steering knuckle has low material utilization and high energy consumption; The new product development of the steering knuckle mainly depends on the designers’ experience, which will bring the long cycle, the high cost and the large risk. This paper focuses on the studies of the forging process design and the forging intelligent design of the steering knuckle. The new forging process of the steering knuckle is developed to decrease the material utilization and energy consumption by using FEM simulations and experimental investigations. In order to find a new product development method of short-cycle, low-cost and high-quality, the expert system of the forging design of the steering knuckle is developed to supervise forging process and die design.Using finite element code DEF0RM-3D, the forging process of the typical steering knuckle is analyzed in this paper. The metal flow, the stress/strain fields, the temperature field and the load/energy curves are provided, and the forming discipline of the forging process of the steering knuckle is achieved. Moreover, the influence on forming process of the friction factor is analyzed. On the base of the analyses, the forming defect of the steering knuckle is provided to supervise forging process optimization and die design.In this paper the new forging process of a type of complex fork part is proposed to realize one-time forging on the hot die-forging press. Taking the case of SK2 forging part, the proper preforming is arranged, the blanking impression is designed, and the closed-die forging with closed-flash process for the preforging is proposed. This new process design realizes one-time forging and decreases energy consumption. Meanwhile, the horizontal forging process of the steering knuckle is studied. Taking the case of SK1 forging part, the new forging process is proposed. The new process is composed of preforming process of extruding staff and preforging process of the closed-die forging with controlled-flash. The material utilization increases significantly in the new forging process.The key technique of the neural network (NN) of the forging design of the steering knuckle is studied in this paper. The unit and the shape factor are defined to describe the feature of the forging part. The feature information of forging part is composed of both feature size and feature shape. The judging criterion is proposed to evaluate the influence on forming process of the different feature shape. Six kinds of units are defined, and the forging process of these units is studied. The forming model of the different kinds of the units is made in terms of equal volume. The shape factor is calculated by contrasting with the max energy in the forging process. The unit size revised by the shape factor is used to describe feature information of the forging part. The feature information of forging part and die are regarded as the inputting and outputting vector respectively to make the NN model of the forging process of the steering knuckle. The model includes the influence on forming process of the different feature size and feature shape. Architectures, controls parameter and learning algorithms of the NN of the forging process of the steering knuckle are also studied in the paper. The convergence and the prediction precision of the NN under different architectures, controls parameter and learning algorithms are provided, and the suggestion how to choose the parameter of the NN is proposed.Basic architecture, knowledge express and reasoning mode of the NN-based expert system is studied in this paper. The design theories and experience knowledge of the forging process and artificial intelligence technology are unified to develope the self-developed software FDesign-SK for expert system of the forging design of the steering knuckle. The experiential knowledge extracted from more than thirty kinds of forging samples of the steering knuckle is stored in this system. The system is realized with the cooperation integration of the NN and the expert system, which could avoid the default of the monotonous reasoning, difficult knowledge acquisition, no uniform knowledge representation, and et al. The framework of the system is modularization, which makes the interface operate easily. FDesign-SK is a system of opening design, and it has the good self-learning function. Therefore, the learning samples could be expanded to improve forecast accuracy and stability of the system. At last, the forging design method of application of FDesign-SK software, FEM simulation and reverse technology is proposed, which could shorten development period, lower cost and reduce workload of the designer.更多还原