Research On Body Welding Quality Inspection Based On CUSUM Control Chart

Body welding quality has a derect influence upon the appearance and assemblyperformance of the vehicle, and largely determines the customers’ssatisfaction.Therefore how to control the body welding quality efficiently becomesthe popular research topic of the motor companies. Aiming at the situation in someChinese auto companies that the workload of body welding quality inspection is highand the data processing is extremely simple, some adaptive CUSUM control charts formean and variance monitoring are proposed in the paper.The main research and achievements are as as follows:(1) Conventional control methods, small shift process control methods wereanalyzed and the average run length is also researched. Furthermore, the performanceof the conventional control charts and small shift process control charts were analyzedbased on Markov chain, and the results show that small shift process control chartswere more sensitive to the smaller shifts.(2) In order to detect the shifts more quickly under the unknown in-controlprocess mean and variance, a new CUSUM control chart for monitoring process meanis proposed. The basic idea of this new chart is to first transform the observations tomake them under a normal distribution, adaptively update the reference value basedon an exponentially weighted moving average estimate and then to assign a weight onit using a certain type of weighting function. A comparison of run length of theproposed new chart and other control charts is shown and simulation studies show thatthe new chart performs better than the others for detecting tiny increases in mean shift.Its application to the body welding quality inspection also has got a good result.(3) In order to detect a broader range of mean shifts, an adaptive CUSUM controlchart for monitoring process mean is proposed and the basic idea of this adaptivechart is transform the parameter, adaptively update the reference value based on anexponentially weighted moving average estimate and link these two together. Atwo-dimensional Markov chain model was developed to analyze the performance ofthe adaptive CUSUM chart, simulation studies and result of the case of body weldingquality control show that this new chart has a better performance for detecting biggerincreases in mean shifts.(4) In order to detect a broader range of variance shifts, the idea of the adaptive CUSUM chart for monitoring process locations is extended to the case of monitoringprocess dispersion, the variance statistics ofS (~2)_t/σ~20is chosen as the variable of thenew chart. A Markov chain model is established to analyze and design the suggestedchart. Evaluation instance results show that, compared with traditional CUSUMcontrol chart and EWMA control chart, the proposed CUSUM control chart based onvariance monitoring is more sensitive to the abnormal variation fluctuation and candetect the abnormity of quality variation earlier.