多元贝叶斯控制图在白车身焊接质量在线测量中的应用
汽车工艺与材料 唐镭 郭大川 汪继光 闫打虎
中图分类号：TB497 文献标志码：B DOI：10.19710/J.cnki.1003-8817.20250084
Abstract：To address the challenges of multi-model mixed-flow productionand insufficient sample sizes in BIW （body-in-white）welding workshops，which adverselyaffect qualitymonitoring，anonline quality monitoring method based ona multivariate Bayesian control chart is proposed.Thisapproach is compared with traditional Hoteling T2 and multivariateCUSUMcontrolcharts.Byintegrating historicaldatawithreal-timemeasurement information，themethod reduces theprobabilityoffalsealarms indynamicmulti-model switchingscenarios.Experimentalresultsdemonstrate that it achieves an average reduction of 34.6% in Type I error rates compared to Hotelling T2 control chart，and a 13.3% decrease in Type II error rates compared to multivariate CUSUM control chart.The method effectively responds to early warning requirements under complex conditions such as robotic trajectory drift.
Key words：Weldingqualityof BIW，Bayesiancontrol chart，Statistical processcontrol，Online quality monitoring
1前言
白车身的制造成本在整车制造中占比高达40%～60%（1） ，焊接质量直接决定车身结构强度与装配精度，因此，实现焊接过程的实时监控对保障整车安全性与可靠性至关重要。（剩余4481字）