Abnormal Working Condition Detection System Of Finishing Mill Design And Implementation

With the development of big data,Artificial intelligence,Internet of Things and other technologies,steel industry are actively carrying out intelligent manufacturing to achieve industrial upgrading.As the core element of intelligent manufacturing,the purpose of equipment intelligent operation is to ensure the reliable operation of equipment and reduce the maintenance and repair investment reasonably.Hot-rolled line key equipment is high complex,high degree of automation and high cost,and the production is accompanied by high temperature,high speed and high dust,which brings great challenges to the operation and maintenance of equipment.Traditional equipment operation and maintenance mainly depends on spot inspection operation and maintenance and regular maintenance.point inspection operation and maintenance work can neither control the production conditions in real time,nor the safety and health of the poor environmental impact inspection personnel,and regular maintenance has the disadvantages of insufficient maintenance,excessive maintenance and wasting a lot of resources.Collect working condition parameters of finishing mill,monitor in real time,combine the process and fault diagnosis technology to analyze the production line anomaly,and transform traditional equipment operation and maintenance into data-driven operation and maintenance.Based on the stacked LSTM,this paper establishes the time series anomaly detection model,realizes automatic detection of working condition parameters of finishing mill,and hierarchically divide the system structure of the finishing mill,realizes the integration and modeling of the structure and state information of the finishing mill,and combines with the production process,the key equipment condition monitoring system is designed.The main research content of this article is as follows:(1)Research on anomaly detection method based on stacked LSTM.Based on the theory of LSTM neural network,this paper proposes a stack-based LSTM unsupervised anomaly detection method for a small number of abnormal samples in industrial production.This paper takes the vibration time series data generated by the main motor of the finishing mill as the research object,By learning from normal samples,extract the inherent distribution and regularity of time series data,construct anomaly detection indicators by predicting residuals,and verify the effectiveness of the algorithm through experiments(2)Research on abnormal processing method of finishing mill.Taking the finishing mill as the research object,this paper first analyzes the structure and process of the finishing mill,Layering of finishing mill system parameters,each parameter is used as a node,secondly integrates the position,attributes,status,abnormal propagation direction and abnormal degree of the node.finally,an abstract model between nodes is established,and node positioning and abnormal tracking are used to improve the efficiency of analyzing and handling abnormalities.(3)Design and implementation of the abnormal working condition detection system of the finishing mill.According to the demand and the hot rolling process,the overall architecture of the abnormal working condition detection system of the finishing mill system is designed.The system includes working condition monitoring,abnormal detection,fault diagnosis,root cause analysis and other modules,which can realize data collection and transmission,abnormal working condition detection,rotating equipment vibration monitoring and fault diagnosis,comprehensive evaluation of rolling mill status and other functions.