Research On Key Technology Of Ultrasonic Imaging Detection Of Large Rotary Parts

The large rotary parts have been widely applied in aviation,spaceflight,power fields,such as turbine disk,defense penetration cabin,projectile cabin,pressure vessel and so on.As an important part to assure the reliability and security of the equipment,the internal defect inspection for large rotary parts has been a hot spot.Due to the large size and curved surface of the rotary parts,the internal defect is detected by manual ultrasonic testing method,which depends on the experience of staff and has the shortage of high labor intensity,low speed,low efficiency and less test data.The test result depends on the experience of staff,therefore the traditional manual method cannot meet the modern industrial demands.With the improvement of the industrial requirements and the development of the power electronic technology,it is necessary to develop the ultrasonic imaging detection technology with features of high efficiency,display visual,quantitative analysis and so on.And it has not only theoretical but also practical significances to explore the ultrasonic imaging detection technology and image processing method.In order to solve the above problems,According to the imaging principle and the characteristic of large rotary structure,nondestructive imaging inspection system for large rotary parts is constructed in this paper.Based on the DAC correction technology,a new ultrasonic plane imaging method and a three-dimensional imaging detection method are proposed.The surface subdividing method and reverse technology are combined for scan path planning.And some efficient algorithm are proposed to extract defect and reduce the time-consumption by analyzing the characteristics of acquire image.Main research work can be summed up in the following aspects.(1)A mechanical scanning device with five freedoms is designed for large rotary parts inner defect inspection by analyzing the principle of the ultrasonic propagation and testing.Due to the introduction of the DAC,the data acquisition and imaging technology can be improved to avoid the impact of acoustic attenuation,near field region and rough surface.Apart from characteristics of conventional ultrasound imaging methods,the proposed DAC imaging detection method can show the defect equivalent diameter directly,which could improve the speed and the accuracy of the inspection.(2)On the basis of the kinematical model and structural analysis for large rotary parts scan path planning,the curved surface subdividing method and reverse technology are combined to improve the speed and automation level of the inspection.A fast isomap surface flattening method is proposed for solving the problems,such as the low rendering speed of large amount of determined 3D data,local observed scope and image distortion.The one to one mapping relationship between the unwrapped image and the original image is able to achieve the dual goals of global observability and high calculation accuracy.The defects distribution of the unwrapped image is also applied to image compressing.(3)In order to suppressing the speckle noise effectively,a new hybrid denoising method is proposed by combining NSCT with edge-preserving self-snake model.The method comprises of three key steps.Firstly,the noise model is established for the sake of the characteristics analyzing.Then the shrinkage function is developed to deal with the NSCT coefficients.Finally,the self-snake model is introduced to reduce the blurring which is caused by the second steps.The comparative experiments show that the proposed approach gets better results than other well-known methods in the area of the evaluation values and visual effect.(4)Aiming at the existing problems of Chan-Vese model for segmenting the detecting images with intensity inhomogeneity,low contrast and artifacts conditions,a novel artifacts detection and bias correction CV model is proposed based on the entropy and region information.The interior and exterior weight parameters are automatically regulated according to the entropy.The local and global region information are combined to improve the robust of region evolution automatically.An artifact indicator function is introduced to prevent intensity outliers and the Retinex theory is used to obtain the piecewise-constant structural part of image,which prevents the adverse effect of bias to segmentation.Then the proposed model terms only act on the structural image which only in not artifact regions.Experiment result shows that the proposed method has high segmentation accuracy and computational efficiency.(5)The differences of volume data resolution in three directions often lead to details missing and surface discontinuous in the industry fields.For above problems,a slice interpolation algorithm of multilevel modified curvature-based registration is proposed.Based on the multiresolution representation,the coarse resolution transformation is prolongated onto the next finer grid and serves as a typically excellent starting point for the finer resolution in the registration part.The modified curvature-based registration model is built to resolve the problem of transformation inconsistency.Experiment results show the proposed method produces lower MSD values,lower time consumption and less blurred edges.At the end of the article,three-dimensional DAC correction imaging detection technology is used to reconstruct the large conical ring parts.The proposed algorithms,such as hybrid denoising method,artifacts detection and bias correction CV model are used for image preprocessing.Experiment results verified the effectiveness of these algorithms in estimating and displaying three-dimensional defect information.