Quality Automatic Diagnosis Research In The Laser Additive Manufacturing For Automotive Engine

Nowadays,with the increasingly fierce competition in the automotive market,the frequency of new product research and development of automotive enterprises is getting higher and higher,which puts forward higher requirements for the speed of new engine research and development.Under the background of lightweight automotive parts,the application of laser additive manufacturing technology can assist the rapid development of automotive engines,and reduce the cost of research and development.However,there are obvious gaps in dimension accuracy and surface quality of additive manufacturing parts.Therefore,it is of great theoretical significance and application value to study the quality diagnosis of engine in the additive manufacturing process.Based on the quality assurance of the engine in the laser additive manufacturing process,this paper explored the influence of different process parameters on the information of molten pool and plasma.Then the process parameters were determined,which were used in the subsequent quality diagnosis experiments.And statistical process monitoring was used in the engine for laser additive manufacturing in order to judge its process stability.At the same time,the existing quality problems were judged and analyzed,so as to realize the quality diagnosis of engine in the laser additive manufacturing process.Firstly,the image processing algorithm of laser molten pool was developed.Through ROI extraction,mean filter,gray transformation,threshold segmentation and edge detection,the features of molten pool area and temperature were extracted.A single-tractmufti-layers deposition experiment was designed and carried out under the condition of single variable.At the same time,the molten pool images and spectral signals were collected in the additive manufacturing process.The plasma temperature and electron density were calculated by the relevant theoretical formulas,Then the relationship between monitoring information and process parameters was established.Secondly,the defect experiments of Al Si12 and 316 L materials were designed and carried out.Based on the analysis for the change of both the molten pool area and temperature,the defects was distinguished in the additive materials manufacturing.And the statistical variables used in the control chart were determined.On the other hand,the forming parts experiments under different controlled conditions were designed and carried out.The T-square statistics and upper control limit were calculated,and and the ability of the control chart method to judge the process state of additive manufacturing was verified.Further,the control chart method was used to analyze the typical structural additive manufacturing process in the engine,and the judgment of the process out of control caused by the heat accumulation during the material addition was realized.Finally,for the aluminum alloy parts and exhaust system of engine,different singletract-mufti-layers deposition defect experiments were carried out.Ten melting pool features were extracted by image processing,and three plasma features were obtained by the relevant theoretical formulas.The dimension of features was effectively reduced by principal component analysis method.Further,the transformed principal components were used as input data,and three models of support vector machine,decision tree and RBF neural network were established respectively.According to evaluation indexes,the support vector machine method and decision tree method were selected for the quality diagnosis of the additive manufacturing process of aluminum alloy parts and exhaust system,respectively.In the end,the defect recognition ability of the selected models were verified by experiments.