| Nickel-based superalloys are widely used in key parts of aerospace equipment because of their excellent mechanical properties such as creep resistance,fatigue resistance,corrosion resistance and high strength at high temperatures.At present,casting and forging as the typical representative of the traditional material reduction manufacturing process is still widely used in the preparation and processing of nickel base superalloy,its processing cycle is long and difficult to meet the increasingly complex high temperature parts processing task,the development of a new nickel base superalloy processing process is imminent.The selective laser melting(SLM)technology developed in recent years overcomes the above problems and has the advantages of high performance,short process and high degree of freedom,providing a reliable solution to the difficult machining problem of complex superalloy parts.The macroscopic mechanical properties and microstructure of laser selective melting workpiece depend on the evolution of temperature field during laser selective melting process.It is very important to elucidate the temporal and spatial evolution of temperature field and its influence on the microstructure/properties of workpiece during selective laser melting.On the other hand,the forming quality of workpiece is affected by the coupling of laser selective melting parameters.It is necessary to realize the cooperative optimization of laser selective melting multi-machining parameters to determine the optimal machining parameters.In this paper,the finite-element Method(FEM),Molecular dynamic(MD)simulation and machine learning methods are combined to study the evolution of temperature field under different laser selective melting parameters,and the scanning rate and laser power parameters are optimized.It provides theoretical guidance for precision machining of complex superalloy parts by laser selective melting technology.The main research work of this paper is as follows:(1)Combined with macro finite element simulation and micro molecular dynamics simulation,the effects of laser output power and scanning speed on the thermal behavior,molten pool size and microstructure evolution of selective laser melting nickel-base alloy were revealed at multiple scales.Macroscopically,the laser output power dominates the size of the pool,the maximum cooling rate and the temperature gradient.Increasing the laser output power and decreasing the scanning speed can obtain larger molten pool size,higher temperature gradient and faster cooling rate.It is found that the aspect ratio of molten pool size is dependent on laser output power but insensitive to scanning rate.At the microscopic level,the dynamic process of selective laser melting of In718 superalloy accompanied by nano-scale crystallization was obtained,and it was found that the formation of nano-scale Cr-rich clusters during the condensation process of the alloy,and the segregation of Cr atoms mainly occurred or formed in the molten pool region,while there was no obvious agglomeration and segregation of Fe and Ni atoms in the molten pool.(2)The machine learning technology based on artificial neural network as the core algorithm,taking the flatness of laser selective melting monolayer as the evaluation index of processing quality,coordinated optimization of laser scanning rate and output power.Based on the sample data set obtained from literature survey and finite element simulation,the artificial neural network prediction model of "machining parameters(scanning rate,output power)-molten pool width(flatness of single layer surface)" was established.The results show that the artificial neural network model can accurately predict the molten pool size,the linear regression coefficient R=0.99987,the mean square error MSE=0.37471 between the sample data and the predicted data.Furthermore,the composition range of selective laser melting parameters with good lap ratio is further predicted,which provides the scientific basis for obtaining the optimal laser selective melting parameters,good surface flatness and dimensional accuracy. |