SIA OpenIR  > 数字工厂研究室
基于机器学习方法的汽车装配质量预测分析
Alternative TitleAutomobile Assembly Quality Prediction Based On Machine Learning
刘舒锐1,2
Department数字工厂研究室
Thesis Advisor彭慧
Keyword汽车装配制造 质量预测 质量分析 加权支持向量机 随机森林
Pages69页
Degree Discipline控制工程
Degree Name硕士
2019-05-17
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract目前利用机器学习方法,对工业生产过程产生的大量实时记录数据进行质量预测分析逐渐成为优化生产研究的主要方向。通过数据驱动,对汽车装配生产过程中的正常样本和缺陷样本进行学习从而生成质量预测模型,是现阶段汽车装配质量预测分析领域的热点研究内容之一。本文结合机器学习算法,对汽车装配过程质量数据进行预测分析,主要体现在以下三个方面:1)通过对MES采集的汽车装配数据的统计分析,指出了现有的MES系统数据采集和存储方面存在的弊端,对其并未考虑与其他系统的对接以及后续数据的集成、分析的不足之处提出了有效地建议;通过对数据的清洗和集成等方面的处理,完成了对镇江北汽和昌河铃木的汽车装配数据预处理的主要任务,提高了机器学习算法对装配质量数据分析处理效率及准确性。2)利用装配过程采集到的时间过点数据为基础,提出了一种基于数据和空间经验分布的r-z变换和分布概率密度的双向加权的SRZD-WSVM算法来实现装配过程质量数据的预测。并结合实际生产情况,通过对生产中可能产生缺陷样本的原因进行分析,得到实际装配过程对质量因素有影响的指导意见。3)利用装配过程记录的车辆选装件的具体数据信息,提出了一种利用基于粗糙集理论的进行填补的方法,并建立随机抽样的级联森林模型算法。进一步结合具体车型车系等特征信息分析装配生产选装件的质量预测以及质量相关因素,提出实际生产指导意见。本文根据汽车装配过程中人机料法环五大因素,从数据驱动的角度出发,对原先摒弃的时间过点数据和装配类型数据进行信息提取,利用改进后的机器学习方法对车间装配生产进行即时检测,对车辆生产的相关安排提出指导性建议,从而改善车间生产优化及装配质量。
Other AbstractThis paper combines machine learning algorithm to predict and analyze the quality data in the process of automobile assembly, which is mainly reflected in the following three aspects: 1) Through the statistical analysis of the automobile assembly data collected by MES, the shortcomings in data collection and storage of the existing MES system were pointed out, and effective suggestions were put forward for the drawback that it didn’t consider docking with other systems and the integration and analysis of subsequent data. Through the processing of data cleaning and integration, the main task of pre-processing automobile assembly data of Zhenjiang Beiqi and Suzuki of Changhe Automobile was completed, and the efficiency and accuracy of machine learning algorithm in analyzing and processing assembly quality data were improved. 2) Based on the time-lapse data collected in the assembly process, a bi-directional weighted SRZD-WSVM algorithm based on r-z transformation of data and spatial empirical distribution and distribution probability density was proposed to realize the prediction of quality data in the assembly process. According to the actual production situation, through the analysis of the possible causes of defective samples in production, the guiding opinions that has influence on the quality factors in the actual assembly process were obtained. 3) Based on the specific data information of automobile optional parts recorded in the assembly process, a filling method based on rough set theory was proposed, and a cascade forest model algorithm with random sampling was established. Furthermore, the quality prediction and quality-related factors of assembly and production optional parts were analyzed based on the characteristic information of specific automobile types and systems, and practical production guidance was put forward. According to the five factors of man, machine, material, method, environment in the process of automobile assembly, from the perspective of data drive, this paper extracted information from the previously abandoned time-lapse and assembly type data, and used the improved machine learning method to carry out real-time inspection of workshop assembly production, and put forward guiding suggestions on relevant arrangements of automobile production, thus improving workshop production optimization and assembly quality.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/25203
Collection数字工厂研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院大学
Recommended Citation
GB/T 7714
刘舒锐. 基于机器学习方法的汽车装配质量预测分析[D]. 沈阳. 中国科学院沈阳自动化研究所,2019.
Files in This Item:
File Name/Size DocType Version Access License
基于机器学习方法的汽车装配质量预测分析.(6499KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[刘舒锐]'s Articles
Baidu academic
Similar articles in Baidu academic
[刘舒锐]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[刘舒锐]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.