Abnormal Detection And Location Of Aluminum Profile Extrusion Energy Consumption Based On Bayesian Network

As the execution equipment for the extrusion of aluminum profiles,the aluminum extrusion press plays a vital role in the extrusion line.The operating state of the extruder plays a decisive role in the quality of the aluminum profile.Once the extrusion equipment is abnormal,it will have a serious impact on product quality and production.If the abnormality is not effectively processed for a long time,causing the abnormal state to cause a shutdown failure or even a safety accident,it will bring huge cost loss and reputation loss to the enterprise.With the integration of industrial manufacturing and information technology and the wide application of advanced technologies such as smart meters,DCS and fieldbus for monitoring process variables such as temperature,pressure and flow,the use of data mining to solve anomaly detection has become a research hotspot,but the existing When the mechanical equipment anomaly detection method is faced with serious problems such as incomplete data and complicated working conditions,the detection accuracy cannot meet the production requirements in practical applications.As a powerful uncertainty expression and reasoning model,Bayesian network has powerful uncertainty problem processing ability and can effectively carry out multi-source information expression and fusion.In recent years,it has been deeply studied by domestic and foreign researchers.Pay attention to it.To this end,this thesis proposes an anomaly detection and localization method for energy consumption in aluminum extrusion process based on Bayesian mesh.Firstly,the energy conversion mechanism of the extrusion process is studied,the energy consumption model of each module is established,and the influence of key factors such as extrusion force and extrusion speed on energy consumption is analyzed,and the Bayesian net work is constructed with key factors as the node;Using the production process data as sample data,the Bayesian network parameter learning and anomaly detection simulation experiments were carried out.Finally,the research content is integrated to realiz e the energy consumption abnormality detection module in the extrusion process.The specific research work of this thesis includes:(1)Based on the extrusion mechanism,the reasons for its anomaly were analyzed.At the same time,based on the energy transfer process and energy loss characteristics of the extrusion cycle,the energy consumption model of the hydraulic system of the extruder was constructed,and the influence of key process parameters on energy consumption was analyzed.(2)Combined with the extrusion energy consumption model,based on the causal transformation relationship between the influencing factors and energy consumption of key factors such as extrusion force and extrusion speed,a Bayesian network structure for extrusion anomaly detection was constructed.(3)Analysis of the abnormality detection Bayesian network structure,introducing a discretization algorithm for continuous nodes.Parametric learning of incomplete data is performed using the Gibbs Sampling algorithm.Innovatively transform the anomaly detection problem and the cause location problem into the posterior probability problem and the maximum possible interpretation problem,and use the Bayesian inference algorithm to perform anomaly detection and cause location.Through Ba yesian network inference experiments,it is verified that the Bayesian network model is effective in analyzing incomplete data sets,and has practical significance for anomaly detection in practical application scenarios.(4)Based on the above research content,the extruder abnormal alarm module based on HDFS cluster was developed in Java language,and the existing extruder energy management system was expanded and extended.It was initially applied in aluminum profile enterprises.