Study Of Deep Learning Method For Prediction The Mechanical Properties Of 3D Printing 316L Stainless Steel

Selective Laser Melting?SLM?is a technique for additive manufacturing by rapid lamination,which is playing an increasingly important role in medical treatment,aerospace and other fields.In order to fabricate samples with good mechanical properties,multiple optimization experiments are required to eliminate spheroidization,voids,cracks and other defects.Therefore,the convolutional neural network in deep learning is proposed in this thesis to predict the relationship between microstructure and mechanical properties of 316L stainless steel farbicated by Laser 3D Printing.Due to the highly abstract features of convolutional neural network,feature visualization methods such as activation images and clustering are adopted.In this thesis,the influences of laser power and scanning speed on the microstructure and mechanical properties of SLM have been systematically studied,providing strong experimental support for the process optimization of SLM forming 316L stainless steel,and providing data support for convolutional neural network training.The main work of this study is as follows:?1?The 99.97%densified workpieces have been fabricated,and the microstructure and grain formation mechanism of SLM forming parts have been studied.The XRD phase analysis shows that 316L specimen is austenitic phase.The austenite structure of columnar and cellular crystals is observed,the theory of formation is explained,and the typical microstructure of molten pool is proposed.?2?The effects of laser power of 140 W¹90 W and scanning speed of 600 mm/s⁹00 mm/s on the mechanical properties of 316L stainless steel by laser melting have been studied,which provid a preliminary technological basis for SLM manufacturing.The optimal process parameters are 190 W laser power,700 mm/s scanning speed,and the maximum tensile strength is 762.83 MPa and the maximum vickers hardness is 253.07 HV_0.2.?3?Different neural network models are used to predict the mechanical properties of 316L specimens fabricated by SLM and visualize the convolutional neural network.The best training results are obtained by deep convolutional neural network VGG16.The features of molten pool morphology are observed inside the convolutional neural network,which provides strong data support for the feasibility of artificial intelligence in predicting material properties in the future.