【作者】 黄春跃；

【导师】 周德俭；

【作者基本信息】 西安电子科技大学 ， 机械制造及其自动化， 2007， 博士

【摘要】 采用表面组装技术(Surface Mount Technology，SMT)形成的焊点的可靠性是SMT产品的生命。焊点组装故障检测与组装质量的控制技术是保障SMT产品质量和可靠性的关键技术。在组装过程中及时检测、发现组装故障，并予以实时反馈，进行及时的组装工艺参数调整和消除故障处理，实现组装过程中的质量检测和反馈控制，以求达到高的产品合格率或一次通过率，是SMT产品生产中刻意追求的目标。本文基于焊点虚拟成形技术，对SMT焊点组装质量检测与鉴别技术进行了较为全面地研究，解决了焊点三维表面形状重构、焊点三维质量信息提取、焊点缺陷智能鉴别、焊点缺陷原因智能分析等关键技术，在此基础上形成了相应的系统及其软件，研究成果具有焊点组装质量信息获取、反馈和控制实时性强，组装工艺参数调整时间短等特点，能适用于要求快速完成组装工艺参数调整进入稳定组装生产、达到高的组装一次通过率的小批量SMT产品组装生产场合。论文研究具有很强的工程背景和现实需求，研究结果具有很强的实用价值和经济、技术价值。基于最小能量原理建立了在表面势能、重力势能和外力作用下的片式元器件焊点、无引脚陶瓷片式载体(Leadless Ceramic Chip Carrier，LCCC)器件焊点、四方扁平无引脚(Quad Flat No-lead，QFN)器件焊点和塑料球栅阵列(Plastic Ball Grid Array，PBGA)器件焊点等典型SMT焊点三维形态预测模型，实现了上述焊点三维形态预测并分析了钎料量、焊盘长度、焊盘宽度和间隙高度等相关工艺参数对焊点形态的影响。结果表明所分析的工艺参数对焊点形态均有影响，任一参数的变化都会对焊点形态产生相应的影响。对采用无铅焊料的片式元器件焊点以及采用有铅焊料的PBGA焊点进行了焊点形态预测结果的试验验证，结果表明焊点形态预测结果与实际焊点形态吻合良好。研究结果为SMT焊点形态的设计和控制奠定了基础。采用统一粘塑性Anand模型描述了SMT钎料合金的本构关系，建立了基于焊点三维形态的LCCC焊点、片式元器件焊点、QFN焊点和PBGA焊点的有限元分析模型，对上述焊点在热循环条件下的力学行为进行了有限元分析与热疲劳寿命预测。对多工艺参数(样件规格、芯片配重、焊盘直径和钢网厚度)组合下的PBGA器件焊点可靠性进行了研究，获得了多工艺参数对PBGA焊点可靠性的影响顺序及其显著性结论。研究了空间域的SMT焊点图像处理算法，在对各种算法的处理效果进行分析比较的基础上，确定了适用于SMT焊点计算机视觉信息获取与处理系统的算法。建立了基于激光三角法的SMT焊点测量系统的数学模型，完成了SMT焊点测量系统硬件设计与搭建，对SMT焊点测量系统参数进行了标定；研究并应用了基于激光三角法的SMT焊点三维表面形状重构方法；运用实体模型切片算法对重构的焊点模型进行切片，获取了焊点关键切面并提取润湿角、焊点高度及切面面积等焊点形态参数，实现了焊点三维质量信息提取。通过三维重构精度检测试验及误差分析得到了重构焊点的准确度，结果验证了本文所研究的焊点三维表面形状重构方法的有效性和准确性。利用最小二乘法进行相关性分析确定了表征焊点缺陷的特征信息参数，提取出用于焊点缺陷智能鉴别和缺陷原因智能分析的样本数据信息；对标准反向传播(Back Propagation，BP)算法进行了改进研究，通过算法测试证明所提出的改进措施能够满足使用要求；利用虚拟成形的表征片式元件焊点缺陷特征信息的数据样本对改进的BP算法模型进行了成功的训练，对片式元件焊点缺陷进行了智能鉴别，所获得的结果与实际结果基本吻合，满足焊点质量智能鉴别精度的基本要求。建立了基于模糊神经网络的焊点缺陷原因智能分析评价模型，利用智能鉴别中训练样本的输出变量作为模糊神经网络的训练样本的输入变量，完成对模糊神经网络的训练，利用智能鉴别中测试样本的输出变量作为模糊神经网络的训练样本的输入变量，完成对模糊神经网络的测试，智能分析结果较为合理。基于本文所研究的SMT焊点图像处理技术、焊点三维形状重构技术、焊点质量信息提取技术、焊点缺陷智能鉴别和缺陷原因智能分析技术，开发了SMT焊点质量评估软件和SMT焊点质量智能鉴别与分析软件。通过在实际SMT生产线上进行实际组装焊点的图像采集、图像预处理、焊点三维形状重构、焊点重构模型形态参数提取、焊点缺陷检测与焊点缺陷原因鉴别试验验证了本软件的实用性。更多还原

【Abstract】 The reliability of the solder joints formed by the surface mount technology (SMT) is the core of the SMT productions. The inspection of the soldered joints defects and the control of the assembly quality are the key technology to ensure the quality and the reliability of SMT products. In order to get high production yield, the goals we pursued in SMT manufacture are to inspect the assembly defect, to feedback the defects, to adjust the technology parameters in time according to the feedback information and to achieve the inspection and the feedback control of the SMT products quality.In this paper, the solder joints quality inspection and intelligent discrimination are all-around studied based on the virtual evolving technology. The key technology including solder joint three-dimensional surface reconstruction, solder joint three-dimensional information measurement, solder joint defects intelligent discrimination and solder joint defects causation intelligent analysis are solved. The system model and software are developed. The achievements in our research have the advantages of remarkable real-time characteristic in the assembly quality information measurement, feedback and control of the solder joints, which applies to the small batch SMT manufacturing that requires the technology parameters being adjusted and the mound production being carried through rapidly to get the high primary production yield. The research, which meets the engineer and reality demands, is of great practical, economic and technical values.Based on the minimum energy principle the three-dimensional shape predictive models of such typical SMT solder joints as chip components, leadless ceramic chip carrier (LCCC) components, quad flat no-lead (QFN) components and plastic ball grid array (PBGA) components are set up and the three-dimensional shapes of above solder joints are predicted. The influence of the correlative technology parameters on the solder joint volume, the pad length and the pad width on the joint shapes is analyzed. Results reveal that all the parameters affect the joint shape and each of them impacts the joint shape in varying degrees,。The predictive results for the lead-free solder joints of chip components and lead solder joints of PBGA components are demonstration tested and the experimental results show that the shapes of predictive the solder joints fit those of the real joints very well. The research results might be widely used in the design and control of the solder joint shape.The constitutive relation to SMT solder is described through the unified viscoplastic Anand model. Based on the solder joints three-dimensional shape, the finite element (FEM) model is built to analyze the mechanical characteristic and to get the prediction of the solder joints thermal fatigue life of LCCC component, chip component, QFN component and PBGA component. The reliability of PBGA solder joints under the multi-technology parameters combination (the specification size, the chip weight, the pad diameter and the stencil thickness) is studied. The influence order and significance of the technology parameters that affect the PBGA component solder joints reliability is achieved.The image processing algorithm of SMT solder joints in three-dimensional area are studied. By the processing effect comparison of different algorithm, the appropriate algorithm is selected to obtain and process the computer optical information.Based on the laser triangulation method, the mathematic model for SMT solder joint measurement system is built, the hardware to measure the solder joint shape is designed and implemented, the solder joints system test parameters are calibrated and the three-dimensional surface reconstruction method is studied and applied to practice. The algorithm is adopted to fillet the physical model, to locate the key section of the solider joints and to measure the solder joint shape parameters as wetting angle, solder joint height and the area of the key section. In this way, the three-dimensional quality information of the solder joints is obtained. The accuracy of the reconstructed solder joints is gained through the three-dimensional reconstruction precision testes and the error analysis. The test result approved that the three-dimensional reconstruction method we adopted in the paper is effective and accurate in the study of the solder joints shape.The characteristic information parameters to describe solder joints defects is determined by using the least square method. The sample data, which is used in the solder joints defects intelligent discrimination and defects causation intelligent analysis is measured. The standard BP algorithm is improved in this paper. The test results of the algorithm proved that the improved measures might satisfy the service demands. The improved BP algorithm is trained successfully by using the virtually formed data samples that are used to describe the chip component solder joints defects information. The defects in the chip components solder joints are discriminated intelligently. The discriminated results are identical with the actual defects. This work meets the precision requirement of the solder joint intelligent discrimination. The intelligent analysis and estimate model for the solder joints defect causations is built based on the fuzzy neural network. With the output variables of the training samples in the intelligent discrimination being selected as the input variables of the training samples in the fuzzy neural network to train the fuzzy neural network and the output variables of the testing samples in the intelligent discrimination being selected as the input variables of testing samples in the fuzzy neural network to test the fuzzy neural network, the fuzzy neural network is tested and the relatively rational intelligent analysis results are obtained.The SMT solder joints quality estimate, the SMT solder joint defects intelligent discrimination and defects causation intelligent analysis software is developed based on the research achievements in this paper which include the technology of the image process, three-dimensional shape reconstruction, quality information measurement, defects intelligent discrimination and the defect causations intelligent analysis of the SMT solder joints. This software is proved to be practical through the experiments of the image collection and process of the actual solder joints, the three-dimensional reconstruction of the solder joint shape, the measurement of the reconstructed solder joint shape parameters, the defects test and defect causation discrimination of the solder joints.更多还原