| Additive manufacturing is an important method in the field of industrial intelligent manufacturing,and it is also one of the key research directions for the development of advanced,sophisticated and high quality manufacturing industry.Laser powder cladding technology is a kind of laser additive manufacturing,which has the characteristics of direct forming by additive manufacturing.In the coaxial powder feeding laser cladding technology,the shape of the powder flow field is one of the key factors affecting the forming quality.In order to realize the online monitoring and feedback control,the real-time detection of the powder flow field is very important.However,most of the existing detection methods take the overall powder flow as the research object to explore the powder cladding mechanism,and rarely detect and analyze the powder flow field morphology and associate with the cladding quality.Therefore,with the number,state and angular velocity,through high speed imaging,super resolution image enhancement and powder particle extraction algorithm,obtain the distribution of powder particles in the air,based on the powder particle tracking algorithm,obtain the dynamic characteristics of powder flow field,in order to explore the relationship between powder flow field morphological characteristics and forming quality.Therefore,the main contents are as follows:(1)Research on powder identification and distribution morphology detection based on improved super-resolution algorithm.According to the characteristics of high speed image frames collected during laser cladding,the image regions of the research object are divided by locating the relative positions of powder flow field and molten pool center.The SRCNN algorithm is improved by introducing atrous convolution and optimizing the loss function,which improves by nearly 6d B on PSNR and nearly 4%on SSIM for better information enhancement of powder flow field images.The grayscale and area range of powder particles are determined based on the grayscale histogram statistics,and the powder particles are identified and counted by double-threshold segmentation algorithm and area screening..(2)Research on target tracking algorithm and transport behavior detection for discontinuous powder particles..A two-stage image processing method is used to remove unclear particles and noise background.The results of DP index show that the performance of Da Siam RPN algorithm is higher than that of KCF algorithm in powder particle tracking.To solve the problem of powder loss and recurrence,the state equation of powder particle motion is established based on Kalman filter algorithm and combined with Da Siam RPN algorithm.The DP value is increased to 87.1%,which realizes the effective tracking of discontinuous powder particles.The velocity and angle distribution of the powder flow field are detected based on the powder particle tracking results.(3)Research on the relationship law of cladding quality reflected by powder flow field morphology based on visual inspection.The powder transfer ratio index is proposed according to the number of powder particles,and the changes of powder transfer ratio and powder utilization rate under different process parameters are studied.The powder transfer ratio and the powder utilization rate have a good linear fit,the R~2is 0.72.The results of particle movement velocity and movement angle in the powder splash zone are characterized and analyzed.Combined with the surface roughness results of the cladding layer,it is found that the surface roughness of the cladding layer is mainly affected by the movement velocity and movement angle of the splashed powder particles,and the powder particles with small speed and small angle are more likely to form sticky powder and increase the surface roughness of the cladding layer.The experimental results of studying the relationship between powder velocity,angle and roughness show that the surface roughness of the cladding layer can be reflected by the proportion of the low-velocity and low-angle splashing powder particles in all the splashing powder particles.The larger the proportion of low-velocity and low-angle splashing powder particles,the greater the roughness,and the smaller the proportion,the smaller the roughness.The conclusion shows that based on visual inspection of powder flow field combined with image processing technology,it is possible to reflect and even predict the changing trend of cladding quality. |
PDF Full Text Request