【作者】 沈南燕；

【导师】 方明伦； 何永义；

【作者基本信息】 上海大学 ， 机械制造及其自动化， 2011， 博士

【摘要】 大型曲轴是船舶、机车和发电设备内燃机上的关键零件,其加工质量与内燃机的耐磨损性、耐疲劳性、振动、噪声等性能关系密切,并直接影响内燃机的可靠度和使用寿命。船舶、机车制造业的快速发展、内燃机产品的更新换代,对大型曲轴的制造工艺提出了高速、精密、复合化的要求。针对切点跟踪磨削技术的工艺特点和加工对象大型曲轴的特性,结合“高档数控机床与基础制造装备重大专项”——大型数控切点跟踪曲轴磨床项目(编号:2009ZX04001-111)的实施,本文以提高大型曲轴切点跟踪磨削过程控制水平为研究主旨,围绕着大型曲轴智能磨削的关键技术及工艺展开系统研究,主要对曲轴弹性变形、自动定位及磨削余量分配优化、磨削参数智能决策、砂轮磨损对连杆颈几何形状和表面质量的影响、误差智能补偿等问题进行深入研究分析,并引入传感器检测及人工智能技术为解决这些难题提出了相应的方法,论文的主要研究工作和成果如下:研究了曲轴在重力、夹具夹紧力和磨削力作用下产生的弹性变形及其对连杆颈尺寸、圆度的的影响。在此基础上,分析了中心架辅助支撑及优化磨削工艺、合理安排工步顺序对减少曲轴弹性变形的作用。根据垂直、水平方向弹性变形的不同特点,重点针对中心架辅助支撑,研究了基于档距变化调整垂直方向支撑块位置的变形控制方法,并提出了基于主轴颈切深误差调整水平方向支撑力的变形控制方法,实现了对曲轴不同方向上弹性变形的有效控制。为了快速、精确地确定曲轴工件坐标系与机床坐标系之间的关系并且保证加工余量的均匀分布,以曲轴自动定位测量系统为基础,首先制定了触发式测头自动跟踪测量连杆颈表面的控制策略,通过测量数据确定了各档连杆颈圆柱面方程。在此基础上,构建了基于各档连杆颈磨削余量分布优化的加工零点定位模型,并建立了约束条件以避免“负余量”现象。求解此模型时采用杂交粒子群算法,引入基于不可行度的竞争选择机制处理约束,实例分析表明该模型和算法在求解磨削余量、确定磨削点位方面具有准确、快速的特点。探讨了曲轴切点跟踪磨削参数选取问题,设计了磨削参数智能决策系统。先对曲轴切点跟踪磨削系统参数进行分类,在此基础上将参数决策任务进行了分解,着重为磨削用量和砂轮修整参数的选取,设计了以范例推理Agent为基础,以模型推理Agent为核心,以规则推理Agent为补充的三种决策Agent。在此基础上,以黑板结构作为多Agent之间通讯与相互作用的媒介,构建了由交互层、决策层和资源层组成的基于多Agent的参数智能决策系统,实现了初始磨削用量的优化选择。砂轮的状态在一定程度上决定了被加工工件的磨削质量,为此详细分析了砂轮半径变化对连杆颈几何形状的影响以及砂轮磨损对连杆颈表面波纹度和粗糙度的影响。以此为基础,结合曲轴切点跟踪磨削过程控制的需要,研究了基于接触传感器的砂轮半径测量方法;并利用砂轮磨粒破碎、剥落产生的声发射信号,根据其振铃计数和均方根与设定阈值的比较结果进行磨削接触或修整过程监控;又以当量磨削厚度、声发射信号均方根、砂轮主轴功率信号的多项式回归曲线均值作为输入,砂轮修整信号作为输出,构建了基于RBF神经网络的砂轮磨损识别模型,并通过实验验证了该模型在砂轮磨钝监测中的有效性。研究了曲轴切点跟踪磨削加工误差补偿策略和智能补偿方法,并设计了相应补偿系统。首先提出一种适应切点跟踪磨削特点的在线误差补偿策略:通过向数控系统提供附加脉冲修正量的方式消除曲轴连杆颈加工误差。在此基础上,针对连杆颈综合加工误差,研究了在线智能预补偿方法,给出相应的补偿算法与推理规则,并利用RBF神经网络选取补偿调节因子,对补偿力度进行控制。又引入了基于模糊自学习的误差补偿方法,以连杆颈半径误差及其变化率作为模糊推理输入量,并利用自学习算法将模糊推理输出与以往补偿经验相结合作为砂轮架跟踪运动的附加修正量。磨削实验结果显示,两种补偿方法都能有效地缩小连杆颈的圆度误差,但前者更适用于“边加工边补偿”的在线补偿,而后者则具有更好的补偿精度和更快的误差收敛速度。更多还原

【Abstract】 As the key part of the internal-combustion engine used in ship, locomotive and electric power equipment, the large-scale crankshaft has close relationship with its wearing resistance, fatigue resistance, vibration and noise characteristic and has the direct influence on its reliability and useful life. With the rapid development of ship and locomotive building manufacturing, the older generations of internal-combustion engines have to be replaced by the new ones, which ask the high-speed, high-precision and complex processing technique of the large-scale crankshaft.Supported by National Science and Technology Major Projects to research and develop the large-scale tangential point tracing grinding machine for crankshaft(No. 2009ZX04001-111), this dissertation does systematical research on some key problems of tangential point tracing grinding to enhance the control level of grinding process for the large-scale crankshaft. Keeping in mind the special characteristics of this grinding technique and the huge crankshaft, the deep studies are conducted on the elastic deformation of crankshaft, the automatic positioning of crankshaft, the optimization of grinding allowance distribution, the intelligent decision of grinding parameters, the influence of grinding wheel wearing on machining quality and the intelligent compensation of machining error. The sensor detecting technology and the artificial intelligence technology are introduced in this dissertation to resolve the technical difficulties as mentioned above. The main researches and results of this dissertation are summarized as follows:The influence of the elastic deformation on the ground crank pin’s dimension and roundness due to gravity, clamping force and grinding force is analyzed. Based on this analysis the methods including the auxiliary support from steady rest, the optimization of grinding parameters and the sound order of grinding process, for reducing crankshaft’s elastic deformation is discussed. According to the different characteristics of vertical and horizontal deformations, the emphasis is putted on the study of corresponding control of steady rests’supporting force in each direction. The process for position control of steady rests’vertical supporting pads is based on the change of crank span. And based on the error of grinding depth for crank journal, a dynamic adjustment method is proposed to control the horizontal supporting force. Thanks to these control methods, crankshaft’s elastic deformation can be decreased effectively.To position the crankshaft in machine coordinate system with the uniform distribution of grinding allowance automatically and exactly, the strategy for controlling touch trigger probe to realize the dynamic tracking measurement for cylindrical surface of crank pin is designed on the basis of measurement device grounded on coordinate measuring principle. Using formulae of all the crank pins’cylindrical surfaces obtained by measurement data, the method for crankshaft’s automatic position is presented based on building optimization model of grinding allowance distribution. And the constraints are also established to avoid the negative grinding allowance of semi-manufactured crank pins. The hybrid particle swarm algorithm is adopted to solve the optimization model and the competition strategy based on unfeasible degree of solution is utilized to handle the constraints. The case study demonstrates that crankshaft can be located in machine coordinate system with uniform distribution of grinding allowance quickly and exactly through this optimization model and its solution algorithm.The intelligent decision system is designed to select the parameters of crankshaft tangential point tracing grinding. According to the classification of all the parameters, the parameter decision-making task is analyzed. CBR Agent,RBR Agent and MBR Agent are respectively designed as the basic, key and supplementary modules for inferring the main grinding parameters and grinding wheel dressing parameters. Using blackboard to mediate the communications and interactions among all the agents, the intelligent decision system consisting of HMI layer, decision layer and resource layer based on multi-agent framework is founded to realize the selection and optimization of the initial parameters of crankshaft tangential point tracing grinding.The state of grinding wheel has decisive influence on grinding quality of workpiece to a certain extent. Thus this dissertation analyzes the influence of change in grinding wheel dimesnsion on the ground contour of crank pin as well as the influence of grinding wheel wearing on surface waviness and roughness of the ground crank pin in details. According to the demand of controlling crankshaft tangential point tracing grinding, the measurement process based on contact senor is studied to survey grinding wheel radius. The ring-down count and root mean square of acoustic emission signal produced by the fragmentation and exfoliation of abrasive grains are applied to detect the contacting state of the grinding wheel and the workpiece or the dressing state of the grinding wheel. The grinding wheel wearing identification model is build using radical basis function neural networks (RBF NN)which takes the equivalent grinding thickness, the root mean square of AE signal and the average value of grinding wheel spindle power signal after polynomial regression analysis as inputs and the dressing signal of grinding wheel as output. The validity of this identification model is then verified by the experiment results.The compensation strategy, method and system for machining error of crank pin in tangential point tracing grinding are researched deeply in this dissertation. The additional impulses as the displacement correction of grinding carriage are given to numerical control system for reducing the errors, which is an effective compensation strategy apt for tangential point tracing grinding. Based on this strategy, an intelligent machining error on–line precompensation system is studied and its corresponding compensation algorithms and reasoning rules are also introduced. The RBF NN is used to decide the compensation regulation factor by which the intensity of error compensation can be controlled. Considering both the error compensation experience and its developmental trend, a new compensation method based on fuzzy reasoning and self-learning for crank pin’s machining error is proposed. According to the radius error of crank pin and its change, the fuzzy reasoning module infers the compensation value for crank pin’s machining error. The grinding experimental results show that the roundness error can be reduced effectively by both two methods, but the former method is more suitable for on-line compensation system and the latter one has higher compensation precision and efficiency.更多还原