Fault Characteristic Detection Method Of Rare Earth Extraction Production Line Based On Multi-information Fusion

In the rare earth production industry,extraction technology plays a decisive role.The extraction technology determines the level of production efficiency of the enterprise,and also seriously affects the quality of the product.The quality of the extraction equipment and the degree of automation of the equipment control also restrict the operation efficiency of the extraction production line.It is of great significance for improving the extraction conversion efficiency and production efficiency.At present,in the rare earth extraction production process,most enterprises rely on manpower relatively,and the degree of automation of the workshop is very low,which is also a factor restricting production efficiency,and also increases production costs.Status and production characteristics,this topic proposes a fault feature detection algorithm based on multi-information fusion rare earth deep processing production line to improve the production efficiency of rare earth extraction,the main work is as follows:Firstly,the quadratic discriminant analysis based on Bayes is proposed.In the actual production process,the amount of data is relatively large,and the fault information is relatively complex.First,the data is classified into general categories to determine which equipment has failed.Based on the original Bayesian algorithm,it is improved to form a new Bayesian classifier,comprehensively analyzes the fault data of the belt and voltage in the extraction device,and combines Bayesian secondary discrimination and regular secondary discrimination Construct a Bayesian classification model,and then through KL divergence,reduce the expected error,make the classification effect more excellent,and finally be able to accurately detect the fault category belongs to the belt or the motor.Secondly,on the basis of Bayesian secondary classification,try to detect which fault type belongs to the belt or the motor through the convolutional neural network,including belt deviation,belt breakage,belt slack,or motor overvoltage or undervoltage.Through theoretical analysis and comparison with expert systems,a more effective convolutional neural network fault detection model is proposed.When the fault information is transferred to the hidden layer through the input layer,after convolution,pooling,and activation function,it is transferred to the output at the layer,complete a forward propagation training.When the error is large,through the feedback function,compare layer by layer,complete the back propagation,and retrain until the correct result is output or the upper limit of the iteration is reached,and the output is terminated.After simulation analysis,the training network curve can converge very well,the error is small,and eventually a good fault detection model is formed.Finally,through a set of data from the production workshop,the Bayesian secondary classification algorithm and the detection algorithm of the convolutional neural network are verified,and the actual failure type can be detected more accurately and efficiently.