Design And Implementation Of On-line Monitoring And Abnormal Diagnosis System For Micro Casting Forging And Milling

The topic of this paper is a subproject of the "Digital Twin System Research on Arc Additive Manufacturing Process",a collaborative project of the School of Mechanical Engineering.The existing arc additive manufacturing system has insufficient utilization of forming process data,and lacks effective data interaction between manufacturing and monitoring,and lacks a set of online monitoring system to link processing data and diagnose and analyze real-time data.In response to the above needs,the thesis develops modules for the acquisition end,transmission system,real-time data processing,data storage and visualization from a modular perspective,realizes the integration of key information of the machining process,forms a complete system,and performs system set-up and experimental verification in the field of machine tools.The main work of the theory is summarized as follows.Based on the current widely used big data technology,the thesis uses Flink’s streaming processing engine to clean,process and diagnose the real-time machining data of machine tools,thus building a complete set of online monitoring of data flow for machine tool machining process.The project involves the use of various technologies,and the whole system covers real-time transmission,real-time data processing,data abnormality diagnosis and data visualization services.Real-time transmission uses Kafka,a high-performance messaging middleware,to interface between the collection side and the server side,ensuring efficient transmission performance while guaranteeing sequential message transmission.Real-time data processing is performed through Flink’s state management mechanism to capture real-time processing data that exceeds a threshold value(the threshold value is obtained through experimental analysis).Anomaly diagnosis is performed by creating a one-dimensional neural network CNN for temporal data to identify temporal data,and by classifying and predicting print data over time with steps such as multi-layer convolutional pooling feature extraction.This model is also combined with the windowing mechanism of Flink framework to complete the system’s anomaly diagnosis function for real-time data.Finally,the large-screen real-time monitoring is performed according to the user-defined monitoring requirements.The system solves the problem of low data utilization in industrial processing by using big data technology,and enables real-time abnormal alarms for processing data,so that factories can monitor the operation status of machine tools online.Finally,after a series of tests,the system’s availability,stability and scalability were verified.