Research On Infrared Image Processing Of Melt Pool In Selective Laser Melting And Its Application

Selective laser melting(SLM)is an advanced digital manufacturing technology.During SLM process,a laser beam is used to melt the powders on the building platform.Materials are made into objects in a layer-wise manner,according to the prescribed laser scan path.At present,SLM has been applied in many fields such as aerospace,medicine,military industry,etc.Each object is formed by stacked melt pools,and the melt pools affect the quality of the object directly.Therefore,it is of great significance to monitor the melt pools during SLM process.In this paper,a Lumasense near-infrared thermal imaging camera has been used to monitor the forming process,track the position of melt pools in real time,and acquire infrared images.Ti-6Al-4V powders were used to print 40 single-layer tracks and the process was recorded with different temperature ranges of the infrared camera.Then the acquired infrared images were pre-processed.The temperature gradient distribution characteristics of the melt pools were obtained,combined with theoretical analysis.The boundaries of the melt pools were determined,the melt pools were extracted and the widths of the melt pools were calculated.After that,the correspondence between the widths of melt pools in the infrared images and the widths of formed tracks was established.A best width divergency between the fitting and measurement around 5% is achieved,which indicates the reliability of the melt pool extraction method.The research on the extraction of melt pools shows that the infrared images contain rich information.During the forming process,a large amount information of the melt pools can be captured in a short time,due to the rapid action of the laser.To process the melt pool information efficiently,deep learning was used to extract the melt pools' features.A convolutional neural network model was established.The specific parameters of the model were determined,which was based on the self-built training set and validation set.Then the infrared images of test set were input to the model to predict the track width.The determination coefficient of the model is 0.974,which indicates that the model fits well.The relative error of the prediction result is not more than 10% among 82.55% of the test samples,and not more than 20% among 95.28% of the test samples,which further proves the good prediction ability of the model.In addition,the effects of laser power,scanning speed,and energy density on the track width were analyzed.This study obtains infrared images,extracts the information of the melt pools through in-situ monitoring,and uses artificial intelligence to predict the size of the tracks.It plays a guiding role for off-situ measurements(such as: microstructure,residual stress,density,etc.);And it provides important reference for the real-time adjustment of process parameters,which can make up for various defects during the manufacturing process and meet actual production needs.