Research On Key Technologies Of On-line Quality Monitoring In Wire And Arc Additive Manufacturing Based On Molten Pool Vision And Temperature Field

Traditional metallurgical quality assurance methods can not fundamentally prevent defects,so that the hidden dangers are left.Therefore on-line metallurgical quality prediction is one of the hot and important research contents in wire and arc additive manufacturing.As one of the quality sensing media in metallurgical process,light is best for describing the quality characteristics of the process.The visual form of molten pool characterizes the stability of metallurgical processes and the distribution of the molten pool temperature field reflects the change of molten pool internal details which can help to judge the metallurgical quality.Firstly,in this dissertation,the hump forming trend is predicted based on the molten pool transient temperature field.Secondly,based on the molten pool image and the neural network,the reinforcement and the penetration depth of the weld can be monitored on-line along with the welding direction,as well as the reinforcement of the deposition.Finally,the prediction of the hump and unfusion trend is realized based on the molten pool image and the video prediction algorithm.The main research content are as follows:For the problem of the time delay in current hump bead monitoring technology,a hump bead monitoring method based on the transient temperature foeld of the molten pool is proposed,which can predict the hump forming trend and the accuracy is up to 90%.Based on the colorimetric temperature measurement of the CCD camera and the spectral distribution feature of gas metal arc welding(GMAW),a theoretical model of the transient temperature field measurement system for the molten pool is established,the characteristic characterization and extraction method of the molten pool surface temperature under the hump forming trend is studied and it lays a foundation for predicting the hump forming trend.For the problem that the existing technology can not monitor the change of weld geometry along the welding direction,an on-line collaborative and quantitative detection method for the weld reinforcement and penetration depth is proposed based on the frequency domain characteristic modeling of the molten pool.The prediction error of the weld reinforcement is less than 0.13 mm,and the the penetration depth is less than 0.09 mm.The accuracy and effectiveness of the method are verified under different welding parameters and the workpiece thickness,thus the variation of the weld reinforcement and penetration depth can be detected on-line in the typical welding process.For the problem that the exisiting technology can not accurately monitor the weld reinforcement due to the remelting area of the deposition layer,a multi-system coordinated deposition reinforcement measurement method is proposed.Meanwhile,an on-line quantitative prediction method of the deposition reinforcement is proposed based on the modeling of molten pool morphometric and temperature characteristics.The accurancy of this prediction method is up to 0.05 mm,which improves the prediction accuracy of the deposition reinforcement under the typical Cold Metal Transfer(CMT)process.Moreover,the prediction method has the genetration ability under different inter-layer cooling time.For the problem that the existing welding monitoring equipment can only provide the realtime metallurgical quality assessment but can not guarantee the on-line control,an on-line prediction model based on molten pool morphology and video prediction algorithm is proposed.Based on the temporal correlation of the molten pool morphology,a theoretical modle of the evolution process of the molten pool morphology during welding is established,which realizes on predicting the future molten pool morphology.Meanwhile,the loss function combined with the perceived loss function,the Mean Squred Error and the Structural SIMIiarity can play a positive role in obtaining better prediction results.Moreover,the predicted molten pool images are used to classify the unmelted penetration and the hump and the accuracy is more than 95%.It means that the purpose of predicting the metallurgical quality in advance is achieved.The on-line photoelectric detection technology based on the molten pool vision and temperature field coordination is proposed in this dissertation,which can effectively improve the accuracy of the metallurgical quality prediction and can evaluate welding quality in real time.