【作者】 董静薇；

【导师】 李大勇；

【作者基本信息】 哈尔滨理工大学 ， 测试计量技术及仪器， 2006， 博士

【摘要】 型砂质量控制一直是铸造领域的重要研究课题。由于缺乏有效粘土含量的快速检测方法,目前的湿型粘土砂质量控制系统多采用由性能检测到组分控制的间接控制模式。本文在型砂组分快速检测方法的基础上,研究了一种湿型粘土砂质量直接控制技术。研究内容主要包括湿型粘土砂导电特性与有效粘土含量和含水量关系的试验研究,型砂主要组分快速检测方法测试电路参数的确定,求解型砂组分的人工神经元网络模型的构建,湿型砂质量分布式控制系统的研制。研究结果表明,基于组分快速检测方法的由组分测试到组分控制的直接控制模式可以实现型砂质量优化控制。通过试验研究与理论分析,证明了湿型粘土砂的导电特性与其主要成分——有效粘土含量和含水量有确定的关系。提出了砂样导电的RC串联等效模型,并定性分析了电极形状、外加激励源的幅值、频率、型砂组分等因素与砂样的等效电阻Rs、等效电容C_d的非线性关系。砂样电极采用嵌入样筒内壁的弧形电极,交流激励源为频率1kHz、峰值10V的正弦波,直流激励源的电压幅值为+9V。研究了湿型砂紧实率与有效粘土含量、含水量及水土比的关系,并分析了回用砂中死粘土对紧实率的影响。研制了由自动制样机构和测试下位机构成的湿型砂组分及紧实率快速测试单元,可完成标准砂样的制作和紧实率、导电参数的自动测量。自动制样机构采用压缩空气驱动,压力设定值为4.5×10~5pa,在制样过程中同步进行实砂压力及压头位移的测量,由测试下位机自动计算出砂样紧实率。紧实率的测量范围为:10～60%±1%。测试单元可在10秒钟时间内与上位机配合完成有效粘土含量及含水量的快速求解。构建了型砂有效粘土含量和含水量的人工神经元网络求解模型。该网络为三层结构的BP网络,输入节点为4个,分别是湿型粘土砂交流导电能力、初始直流导电能力、直流导电能力变化率和紧实率;输出节点为2个,分别是型砂的有效粘土含量和含水量;网络包含一个隐含层,隐含层节点为6个。为了克服原材料来源不稳定对求解精度产生的影响,对BP算法进行了改进以提高网络模型的泛化性能和鲁棒性。有效粘土含量的求解精度为±0.3%,含水量的求解精度为±0.2%。提出了基于组分快速在线测试方法的数字PID加料控制算法,对混砂机的加料量进行控制。根据混砂试验的经验数据建立了模糊规则表,采用模糊式参数自整定方法对PID参数进行在线自整定。研制了混砂机加料控制单元,包括自动加料机构和加料控制下位机,可实现手动、自动加料及闭环联网加料控制。自动加料机构由螺旋给料器、称料单元、加水单元等部分组成。加料控制下位机完成加料控制算法的运算,控制加料机构动作完成加料过程。构建了由两级计算机组成的分布式控制系统。上位机完成现场测试数据的采集、实时显示、存储及分析,人工神经元网络运算以及系统各项参数的设置。下位机控制机构动作,完成标准砂样制作、信息参数采集与上传以及型砂组分的定量添加。计算机系统具有在线故障自动诊断与处理功能,对系统关键部件的运行状态进行实时监测、分级管理,使系统的稳定性、实用性达到铸造车间生产实际的要求。试验运行结果表明,本文研制的型砂质量直接优化控制系统的型砂组分控制误差分别为:有效粘土含量<±0.5%,含水量<±0.3%。更多还原

【Abstract】 The quality control of green sand has been the important task in domain of foundry all the while. Because there is no fast testing method for active clay, most of the green sand quality control systems at present work in direct mode, from performances testing to ingredient controlling. Based on a fast ingredient testing method, a direct control technique of green sand quality is investigated in this dissertation. The research includes experimentally studying of the relationship between conductive capabilities of bentonite-bonded molding sand and its active clay and water content, selecting the parameters of ingredient testing circuit, modeling of the artificial neural network for green sand ingredient calculation, developing the distributed control system for green sand quality. The investigation results have proved that the optimization control of green sand quality can be realized in direct mode, from ingredient testing to ingredient controlling, based on fast ingredient testing method.The relations between conductive capabilities of bentonite-bonded molding sand and its main ingredient, active clay and moisture, are confirmed by experimental research and theoretical analysis. The equivalent model of series RC for sand specimen conductivity is put forward, the nonlinear relationships among the model, equivalent resistor Rs and equivalent capacity Cd, and various factors such as electrode shape, amplitude and frequency of external exciting power source and green sand ingredient are analyzed qualitatively. A pair of electrodes in arc-shaped are embedded in inner wall of specimen tube, the frequency and peak value of AC exciting power source with sine wave are 1kHz and 10V respectively, the amplitude of DC exciting power source is +9V. The relations among green sand compactability, active clay, moisture and ratio of water to clay are studied, and the influence of dead clay on compactability in returned sand is also analyzed. The green sand ingredient and compactability fast testing unit, made up of an automatic sampling machine and a testing slave, is developed to prepare a standard sand specimen and test conductive parameters automatically. The automatic sampling machine works under the drive of compressed air and the working pressure is 4.5×105pa. During the process, the compacting pressure and displacement of compressing head are measured simultaneously and compactability of green sand is calculated by the testing slave. The testing range of compactability is 10~60%±1%. Cooperating with host computer, the testing unit can display and print the active clay and moisture of green sand within 10 seconds.An artificial neural network model to solve the active clay and moisture of green sand is designed. The model is a three-layer BP network with 4 input nodes and 2 output nodes. The input nodes represent green sand AC conductivity, original DC conductivity, variety rate of DC conductivity and compactability. The output nodes represent green sand active clay and moisture. The network includes a hidden layer with 6 nodes. In order to overcome the influence of instability of moulding materials on precision of ingredient calculation, some innovations of BP algorithm are taken to improve the generalization and robustness of BP network. The precision of solving active clay is±0.3%, and the moisture is±0.2%.Based on the fast ingredient testing method, a digital PID feeding control algorithm is put forward to control the feeding operation of the sand muller. Fuzzy rules are set up by expirical data from mixing experiment, and PID parameters are adjusted by fuzzy self-tuning method. A feeding control unit, including an automatic feeding machine and a feeding control slave, is developed and it is able to feed ingredients in manual, automatic and closed-loop modes. The automatic feeding machine consists of screw feeders, weighing filler and water adding unit. The feeding control slave is used to run the PID algorithm, and control the operation of feeding machine.A distributed control system including two-level computers is constructed. The host computer is used to collect and display field measuring data in real time, save and analyze them, run artificial neural network program and set all parameters of the system. The tasks of slave computers are controlling sampling machine, collecting and uploading information parameters and feeding green sand ingredients quantificationally. The system has the function of automatic on-line trouble shooting, monitoring the operation states of system key parts in real time and supervising them in grade, insuring the stability and practicability of system to meet the requirement of real production in foundry workshop. Experimental running result indicates that the control error of green sand ingredients with optimizing direct control system developed in this thesis are active clay<±0.5%, and moisture content<±0.3%.更多还原