Discovery And Analysis Of Abnormal Production-events In Workshop Based On Deep Neural Network

In the manufacturing field,the new mode of "cloud manufacturing" is proposed with the development,integration and application support of cloud computing,Internet of things,big data and AI.It is driving the transformation of traditional Manufacturing Execution System(MES)into manufacturing execution system.Today,MES development becomes more rapid.Many problems of traditional MES is being addressed by new technologies merge into the manufacturing process.However,many problems have arisen with the application of new technologies to traditional MES.To this end,the following three aspects of the manufacturing execution system are further studied.Firstly,in order to adapt the MES to the big data environment,in this article,we improved the data storage aspect and proposes a storage scheme to adapt to the big data environment.Through the non-relational database and memory database combination,together as MES storage media.Making it highly data access capabilities and high scalability,and can store data with semi-structured and unstructured features.Use this optimized data storage method,can make the system management of the production process built on of a large number of manufacturing data.The system can quickly access data and improve processing speed and response time by storing real-time data efficiently.the storage of a large amount of historical production data also enables enterprises to manage their production from a more macro perspective,making statistics and analysis of the workshop data more meaningful.Secondly,in order to effectively solve the problem that manufacturing enterprises cannot effectively control production anomalies,this article presents a method of workshop production anomaly detection based on deep neural network(DNN).And an efficient and reasonable classification system is established for the factors that lead to abnormal production,and the corresponding quantitative methods are given according to the specific characteristics of each influencing factor.According to the analysis of various influencing factors and production anomalies,an production anomaly prediction model based on optimized data storage was established,which can make real-time prediction of quality and delivery period abnormality.The model uses DNN as the core of prediction,using the influencing factors and abnormal production informations in the database as the training data and the the real-time factor index as prediction data.This method can monitor and track product quality and the delivery period anomalies during the production process,providing data foundation for production management.Finally,in order to further reduce the impact of various abnormal factors on the processing and production,this article studies and analyzes the correlation between the influencing factors and the production anomalies and the strength of this relationship,and draws the importance of various anomalous influencing factors.And research on the related entities of equipment,processes,personnel and other major anomalies,and put forward the corresponding treatment methods of influencing factors.Taking the equipment as an example,this article starts with the key cutting equipment,and puts forward a set of methods for predicting the remaining useful life of the tool based on DNN.It reduces the influence of the equipment on the processing production from the point of efficient maintenance of the equipment and achieves the purpose of avoiding the abnormal occurrence.In summary,in this article,the existing MES system is improved from three aspects of data storage,anomaly detection and anomaly processing,and specific solutions are proposed for anomaly detection and processing.Using DNN and other methods,the AI technology is applied to the traditional MES,and the feasibility of this method is verified through the training and testing of various neural networks in the solutions,which also solves other manufacturing problems provided a reference.