Research On Modeling Method Of Laser Cladding Temperature Field Prediction Model Based On Deep Learning

Laser cladding technology is widely used in the field of surface treatment of vehicle mechanical parts.Especially in cylinder,piston,crankshaft,drive shaft and camshaft.It is of great significance to study the heat conduction-melting process of molten pool in laser cladding for improving the performance of laser cladding.The physical process in the molten pool is complex and the temperature gradient is great.It is a great challenge to establish the temperature field model of molten pool and explore the heat conduction-melting process of laser cladding by physical field coupling method.Therefore,it is difficult to measure and simulate the temperature field of molten pool during laser cladding.This paper draws on the application of deep learning in model construction,and takes the temperature field model of laser cladding molten pool as the research object to study the coupling mechanism of each physical field of laser cladding temperature field.In this paper,the simulation modeling method of molten pool temperature field based on deep learning is deeply studied to realize the prediction and expectation of molten pool temperature field.The operation mechanism of deep capsule neural network in simulation model analysis is explored,and the theoretical framework of deep capsule neural network in simulation model analysis is improved.Based on the actual laser cladding and process experiments,the feasibility verification method of simulation model modeling method based on deep learning is proposed.Specific research contents include:(1)A modeling method of laser cladding molten pool temperature distribution prediction model based on deep learning is proposed.Firstly,the core idea of modeling method of laser cladding molten pool temperature distribution prediction model based on deep learning is introduced.Secondly,the prediction model framework of laser cladding molten pool temperature distribution based on generative adversarial capsule neural network is constructed.Finally,each part of the overall idea of the above prediction model modeling method is briefly described.(2)The temperature distribution simulation model of laser cladding molten pool was established by using the finite element analysis method.A deep learning model dataset is obtained.The coupling mechanism of each physical field in the temperature field of laser cladding molten pool was studied.The COMSOL Multiphysics software is used to model the temperature field of molten pool based on the coupling model of temperature field,radiation field and convection field based on Gaussian heat source in molten pool.(3)The temperature distribution characteristics of laser cladding pool are analyzed,and samples and test data sets of deep learning models are established.The reasonable setting of process data and the method of obtaining sample images of temperature distribution of laser cladding pool are studied.A large amount of simulation experimental data were obtained,and image samples of the temperature distribution of the laser cladding molten pool at different power and different times were extracted.A sample set for deep neural network prediction model training is established.(4)A prediction model of melt pool temperature distribution based on generating a network of anti-capsules is constructed.The mechanism of deep capsule neural network in simulation model analysis is studied.The capsule neural network is combined with the generation adversarial network,and the deep generation anti-network model architecture based on the capsule network is constructed.A prediction model of melting pool temperature distribution based on generating a network of anti-capsules is designed and trained.(5)On the basis of laser cladding and process experiments,the experimental verification of the predictive model is carried out.Laser cladding and process experiments were carried out.Through the tissue hardness test of the cladding layer,the temperature distribution of the melting pool is analyzed from the side and compared with the results of the prediction model of the temperature distribution of the melting pool.Finally,the feasibility of modeling the modeling method of the temperature distribution prediction model of laser cladding pool based on deep learning is realized.