Broken Core Diagnosis Model Of Iron Castings Of Engine Based On BP Neural Network And Its Application

Broken core is an important cause of scrapping of iron castings in automotive engines,and many complex factors such as man-made materials,equipment,and processes can affect the quality of the sand core and cause it to break.The current research on broken core defects of iron castings for automobile engines mainly adopts the “traditional experimental trial and error method”,which consumes time-consuming consumables and it is difficult to quickly and effectively obtain sand core quality control strategies.For this reason,this thesis analyzes the relationship between the relevant parameters of the process of sand core production to the whole process of pouring and the broken core,determines the main factors causing the core breakage,and uses “BP neural network method” to establish a set of broken core defects for iron castings of automotive engines.Diagnose the model,and based on this model,study the sensitivity degree of each influence factor to the defect;combine the fluctuation of the relevant parameters of the actual process to obtain the process control strategy to guide the actual production.First of all,this article analyzes the quality of production process of automotive engine iron castings,systematically describes the common defects,then conducts in-depth research on the status of broken core defects at home and abroad,and analyzes the production and application of sand cores in combination with the actual production of enterprises.The process of the whole process,to determine the main factors causing broken core defects.At the same time,with the company's ERP system,the data of the main factors of the broken core defects are excavated,and the data mining process is described in detail by taking typical iron castings as an example.Secondly,this thesis uses the method of BP neural network,network-based learning mechanism and the actual research of the broken core defect problem,and establishes the three-layer network model of input layer-single hidden layer-output layer to preprocess the defect factor data.Randomly divide it into training sample sets and test sample sets.MATLAB software is used to train the network model,and the functions,objective functions and training parameters among the layers are set.The parameters are constantly modified through forward propagation of data signals and reverse propagation of errors,and the network model of core-broken defects is finally optimized.Finally,Wuxi First Auto Foundry Co.,Ltd.was used as an applied research object to conduct on-the-spot investigations into the processes of manufacturing cores,cores,cores,moldings,smelting and casting,and post-processing of castings in the workshop.The relevant production quality data were excavated,and qualitative analysis was conducted.The advantage of "BP neural network method" compared to "traditional experimental trial and error method" in the multi-process,multi-influence factor of the broken core research,and quantitatively verifies the accuracy of the network model in the prediction of broken core defects;and based on This model studies the sensitivity of the influence factors on the quality of the sand core,and combines the fluctuations of the actual process parameters to obtain a process control strategy to guide the actual production.