Laser High-contrast Marking Experiment On Metal Surface And Simulation Prediction Of Temperature Field Uniformity

Direct part marking(DPM)processed on metal surfaces is a reliable way to provide lifetime traceability of parts that are exposed to harsh environments.DPM is already being used in the automotive,aerospace and electronics industries.Laser marking is a preferred method of performing DPM to attain permanent and highly contrasted superficial inscriptions on a wide variety of materials.However,the uniformity and stability of laser marking are still one of the reasons that affect the wide application of direct part marking.In addition,the metal surface generally has strong reflective characteristics.Many DPM codes are engraved on the reflective metal surface which can be very challenging for reliable identification.Therefore,the laser marking method was introduced in this paper to carry out experimental study on the processing high contrast marking on metal surfaces.A three-dimensional simulation model for laser marking process is developed,using the finite element method(FEM),aiming at the prediction of the final temperature field formed on workpiece and optimizing the process.The formation mechanism of high-contrast direct part marking on metal surfaces is also analyzed.At the same time,the prediction model based on machine learning is studied to predict the temperature field theoretically.The work of this paper is mainly reflected in the following aspects:Firstly,the form mechanism of high-contrast marking on high reflection aluminum alloy surface was studied by using picosecond laser.Laser marking experiments was carried out to obtain the making samples.A laser scanning electron microscope was used to acquire the surface roughness of the samples.The experimental results show that the smaller the laser marking speed and the line spacing,the greater the surface roughness and the higher the surface contrast.Then,the optical scattering model of rough surface is established based on electromagnetic wave theory.Through theoretical calculation,the mechanism of high-contrast of laser marking surface is explored.Secondly,based on the theory of laser-material interaction and numerical solution model,the laser scanning strategy to achieve uniform marking is explored.The non-uniformity laser marked samples have the same surface feature with that of the temperature field distribution obtained by simulation under the same process conditions.The temperature of the scanning area where the laser starts is obviously low.The middle temperature was high and the temperature of the starting and ending positions was low in each line of scanning trajectory.Based on the thermal accumulation effect,the temperature compensation of the area where the laser start is considered by changing the scanning direction during the marking process.Meanwhile,the temperature compensation of the low temperature region is realized by reducing the scanning speed of the low temperature region.Therefore,a multi-velocity laser scanning method based on local low temperature region is proposed.Based on the laser marking thermal simulation model and experimental verification,the feasibility is proved.Finally,in order to facilitate the selection of laser marking parameters,a prediction model of laser marking temperature field was established based on machine learning.The prediction of temperature field for unidirectional trajectory and reciprocating trajectory with different scanning intervals was realized.Firstly,a mathematical model of laser marking trajectory is established based on partial differential equation,which can realize the simulation solution of temperature field under any laser scanning trajectory.Based on LSTM network,the machine learning prediction model of laser marking temperature field is established.The temperature data obtained from the numerical model was used to build the data set.A data set structure conforming to the laser marking process was built.By comparing the prediction results with the finite element numerical calculation,the established prediction model has higher prediction accuracy and greatly saves the calculation time.