【作者】 马祥宁；

【导师】 金隼；

【作者基本信息】 上海交通大学 ， 车辆工程， 2007， 硕士

【摘要】 本文遵从“数据驱动质量”的理念,围绕轿车白车身制造质量控制中“检测—评价—控制”的工作主线,对其中涉及到的若干关键问题进行了系统性的阐述,将Bootstrap方法引入车身质量控制体系,对白车身制造尺寸质量评价方法及其控制界限的设定方法做了一些创造性的研究,这在一定程度上丰富了轿车白车身制造质量控制的理论体系,为今后的车身制造质量控制工作提供了新的方法与思路,从而有助于更有效的提高轿车白车身的制造质量水平。本文的研究工作主要在以下几个方面:(1)对车身冲压件尺寸的质量控制进行了研究,发现受轮次、来料、修整模具、温度等因素影响的冲压件尺寸数据并不能用正态分布来描述,针对这个问题,将无须对分布进行假设的Bootstrap方法引入车身质量控制中。而且本文对Bootstrap方法进行了较深入研究,对车身质量控制中的Bootstrap方法进行了探讨,建立了车身焊装质量控制中的Bootstrap仿真过程,指出该方法可以用于处理车身检测数据采样的小样本问题。同时对Bootstrap方法的使用进行了进一步探讨,给出了该方法对于不同样本容量下的估计误差曲线,分析了其对所关注的参数的估计误差与样本容量的关系,发现该方法可以对车身质量控制中的分布未知的质量指标的分布做出估计,并且能够获得该统计参数的估计误差。(2)作为Bootstrap方法的应用,建立单因素条件下的车身焊装统计模型,用Bootstrap方法对指定的质量指标进行统计推断,并且在给更多还原

【Abstract】 This paper obeys the philosophy of“Data Driving Quality”, and follows the route of“Measurement-Evaluation-Control”in auto body manufacturing quality control. New quality evaluation method and control limits set-up method for auto body manufacturing dimension are proposed here, which enrich the theory on auto body manufacturing quality control and do favor to improve the auto body manufacturing quality effectively.The research work in this paper was focused on the two points as follows:(1)The quality control methods for dimensions of sheet parts are studied, the results indicate that the variation of the data of sheet parts doesn’t due to Normal-Distribution because of shift-to-shift, metal material, die form modifying, temperature and so on. So we apply Bootstrap Method to auto body quality control. Not so much, we do further research about the method in body quality control, it can process the problem of small sample in body measuring. Meanwhile, we analyze the relationship between error and sample number in Bootstrap Method. The simulating results indicate that the distribution of quality index can be gained by Bootstrap Method, and Bootstrap Method also gives the error for it.(2)With Bootstrap Method, body welding model under single factor condition is established. According to the problems for quality evaluation of body welding processes, small sample is translated to big sample using Bootstrap method. When the allowable error-maximums fixed and the probabilities of the two type errors are considered, the best volume of sample更多还原