Manufacturing Condition Monitoring And Data Analysis System Based On Industrial Internet Of Things

Industrial production is in a period of transition to intelligent production.In China's massive manufacturing enterprises,most of them are still in the traditional production mode,such as slow information transmission,difficult production management and complicated process control,which leads to the disconnection between management decisions and actual production.As for this problem,this thesis considers the production of aluminum foil liner paper in packaging industry as our object,and proposes the design of a set of manufacturing condition monitoring system based on the IoT.By constructing a data acquisition platform,we collect and monitor the manufacturing condition data of each station equipment in the workshop,so as to timely respond to abnormal conditions in the production process and monitor the manufacturing condition more intuitively.At the same time,considering the quality prediction of products,we propose an analytical method with universal applicability.Also,we build a product quality prediction model through the XGBoost algorithm in integrated learning,and predict the product yield rate to improve the low efficiency and low accuracy of manual control in quality control.Firstly,the design scheme and overall structure of the manufacturing condition monitoring and data analysis system of the workshop equipment is put foward,including the monitoring system consisting of the sensor terminal equipment and the host computer monitoring configuration software,and the data analysis prediction model based on the XGBoost algorithm.Secondly,the thesis proposes the design of a workshop data acquisition and monitoring system based on the IoT.The study uses the multi-point sensor architecture to complete the condition data acquisition of the equipment,and combine the Modbus RTU,Modbus TCP and Ethernet protocols to realize the network communication function.At the same time,the interface and function of the monitoring software are designed in the host computer.In a word,we build a workshop monitoring network to provide a source of information for production management and data analysis and forecasting.As for data analysis,industrial production in age of big data will flood massive data,which is characterized by data clutter,variable variable types(numerical,text,image,audio,etc.)and unbalanced data distribution,which makes the data analysis process extremely difficult.In order to design efficient and accurate models to adapt to these difficulties,we introduce ensemble learning with strong adaptability to data in machine learning,and uses XGBoost to establish industrial product quality prediction model.By analyzing the effects of Extreme Gradient Boosting(XGBoost),Random Forest model and Gradient Boosting Decision Tree(GBDT)on quality prediction in the experiments,we can find that the former model performs better than the latter two models and has good performance for unbalanced data processing,which verifies the effectiveness of the model.