SPC And Abnormal Pattern Recognition In Glass Fiber Production And Its Implementation

Statistical process control(SPC)has made remarkable achievements in many fields such as automotive and aviation,but in the production process of glass fiber,the quality management is still mainly based on traditional sampling method,which has many defects.Therefore,exploring how to use SPC reasonably and effectively in the production process of glass fiber is of great significance to the development of glass fiber.In the implementation of glass fiber SPC,it is important to recognize and understand the abnormal patterns presented by control charts for the effective implementation of SPC,and by introducing pattern recognition into the control chart pattern analysis,the efficiency of control chart can be improved.This article is based on the actual SCADA project of a glass fiber factory,and the main work is as follows:(1)Under the current situation that there is nearly no SPC experience in the glass fiber industry,the SPC design is carried out according to the characteristics of the glass fiber production.The key quality characteristics of SPC in glass fiber were identified,the influencing factors were analyzed,and the performance analysis method of glass fiber measurement system based on MSA was determined.The SPC operating conditions of glass fiber are divided into normal operating condition,multiple specifications and small batches condition,and non-normal condition.Examples of different conditions are analyzed,the process of the SPC control chart is explained in detail,and the SPC parameters and optimization strategies are discussed.For non-normal special condition,a EWMA symbol control chart capable of adjusting sensitivity is proposed.This article applies SPC to all the key links of the factory's production line,and establishes a relatively complete SPC quality management system.(2)In order to identify the control chart abnormal patterns in glass fiber SPC,a recognition method based on model fusion is proposed,in which the light gradient boosting machine(light GBM)and the convolutional neural network(CNN)are combined through stacking.17 statistical and shape features were screened using correlation analysis and light GBM,and wavelet analysis was used to extract frequency-domain features.Finally,through comparative experiments,it is shown that the method in this paper has higher accuracy and generalization ability than other models,and the effectiveness of the model is proved by the effect of the model on actual data.(3)According to the characteristics of the factory's production line,the SCADA layout was reasonably planned,and the SCADA system was established through System Platform,which greatly improved the factory's digitalization level.On the basis of SCADA,an SPC online diagnostic system was further constructed to assist personnel in making decisions,which improved the information and intelligence level of the factory's SPC.This article details the key development process of the system,and the stenciled development method enables the system to be quickly deployed to other glass fiber production lines,greatly reducing the amount of engineering and shortening the project cycle.