Research On State Identification Of Cladding Layer In Metal Laser Cladding Process

Metal laser cladding is a process of rapid melting and solidification of alloy powder and substrate surface by using high energy density laser beam,so that a layer of composite material with excellent comprehensive properties can be obtained on the substrate surface.However,due to the interaction and restriction of numerous physical fields in laser cladding process,it is difficult to guarantee the processing quality of metal laser cladding process.The forming state of cladding layer plays a key role in the quality of the final workpiece.There are many factors to measure the forming state of cladding layer.Three factors which have a greater impact on the forming state are selected as the research object of this paper,which are surface crack density,surface flatness and dilution ratio.In this paper,a method based on neural network is proposed to identify the forming state of laser cladding layer:(1)The prediction model of surface crack density and surface flatness of laser cladding layer based on BP neural network is proposed.The generation of cracks has a great influence on laser cladding forming,and limits the popularization of this technology.Aiming at the problem that it is difficult to measure the surface crack of laser cladding layer,the concept of crack density is introduced,and the shape of crack is described by CAD software,so as to calculate the total length of crack and the total area of cladding area,and finally obtain the crack density under different process parameters.Aiming at the problem that it is difficult to predict the crack density and surface roughness of laser cladding layer,a prediction method based on BP neural network is proposed.The input parameters are laser power,powder feeding rate and scanning speed,and the output parameters are surface crack density and surface roughness.The training data of single-layer and multi-channel cladding test are substituted into the training prediction model to obtain the model which can be used for prediction and has high prediction accuracy.(2)A classification and recognition method of laser cladding dilution ratio based on multi-channel convolution neural network is proposed.Aiming at the problem that the measurement process of dilution ratio is very complex,and considering that dilution ratio is very important to the forming state of cladding layer,the image data of molten pool is used to realize the classification of dilution ratio.According to the literature,the workpieces with different dilution rates are divided into three categories,which are normal dilution rate,high dilution rate and small dilution rate.For the problem of the commonly used convolutional neural network,the feature extraction capability is insufficient and the accuracy is low.The method of threshold segmentation and edge detection is proposed to extract the interested area of molten pool image as input of neural network.Then,the multi-channel convolution neural network model is constructed by using parallel small convolution kernel,which can increase the depth of the model,enhance the ability of feature extraction,and improve the accuracy of model classification;Aiming at the problem that the model is prone to over fitting,this paper puts forward that adding dropout layer to the model can effectively improve the problem of over fitting.Finally,through comparative experiments and data validation,the classification accuracy of the proposed model is the highest,with an average of 92.83%,which proves the effectiveness of the proposed model.Finally,a new method of laser cladding forming state recognition based on neural network is proposed.The method uses BP neural network to predict the surface crack density and surface flatness of cladding layer,and then classifies the dilution rate by using multi-channel CNN model.The accuracy of the method is verified by the data measured by the single layer multi-channel cladding experiment.Through the research of this project,the identification of the above three forming state parameters can be realized,which provides theoretical and technical support for improving the quality of laser cladding workpiece.