Segmentation And Spatial Locating Of Defects In Precision Weldments With Complex Structure

Titanium alloys have better over-all properties, laser welding holds great predominance in the aspect of sheet joint of titanium alloy, so laser welding of titanium alloys is widely used in important structure of aerospace craft. Since the inherent characteristic of laser welding, it is very difficult to avoid pores and other defects in weld. If the pores of continuous distribution lie in the key position of structure, the failure will take place in the action of load when aerospace craft flies with high velocity, and this will bring safety problem. Therefore it has important theoretical and actual significance developing corresponding research of nondestructive testing such as defect locating and identification and so on.Aimed to the nondestructive tesing demand of one skin and sternum welding structure in important parts of aerospace craft, the detection scheme was demonstrated, relevant researches such as image denoising and enhancement, defects segmentation, defects spatial locating and defects visualization were studied in this paper.According to the structure characerstic of precision weldments, detection images which contain small defects were obtained by using X-ray real time imaging system and multiple view method. Aimed to the problem that the obtained images have low contrast, and most of the small defects are submerged by noises, the combination processing method was carried out to remove noises based on image superimposing and wavelet transformation, sym8 wavelet, hard threshold and minimax criterion were used to get rid of noises. The image enhancement method based on unsharp masking and automated adjustment contrast was adopted to solve the problem that defects extraction can not be achieved easily for high intensity of weld zones, and the contrast between defects and their background was improved greatly, this can provide better basis for subsequent defect segmentation and other processing.The segmentation method based on weld zones extraction was provided. Twice automated threshold were adopted to extract the weld zones, then sharpening and smoothing processing were carried out in the extracted weld zones individually, finally the small defects were extracted by combining the background subtraction method. But it it easy to emerge shadow in the defects edge for many times sharpening, and the segmentation morphology of defects is changed also.Segmentation of small defects was achieved by using morphology simulation background in the basis of wavelet denoising, and the segmented defects can keep inherent morphology. The AOI (aera of interests) was extracted firstly, then wavelet denoising and morphology simulation background were carried out individually, finally the small defects were segmented by combining the background subtraction method. In order to obtain better segmentation results, the determination method of optimal structure element size was provided by using the fitting and subtraction method of line gray distribution curve, the size can be obtained by searching the maximum value of the above subtraction results. The method is not sensitive to the size of AOI, and it has stronger adaptability, this can provide better basis for the subsequent automated extraction of defect locating feature point.For the research object in this paper, defects position in weld has great influences on structure reliability. According to the features of weldments structure and detection image, mathematic models of defect depth and deviation were established individually. The models verification was carried out by the destructive tesing of actual weldments, the results show that the established models are correct, positional error of defects depth is not more than 5.1 percent in actual weldments. In order to increase the defects locaing accurary and detection efficiency, automated extraction of projection distance which can influence the locating accurary was carried out, and the better results of automated extraction can be obtained. At the same time in order to achieve the automated extraction and corresponding of defect data in multiple view images, the automated extraction algorithm which can extract the defects size and longitudinal distribution was provided, and the automated corresponding of defects data can be achieved through corresponding criterion, and the batch input of extracted data can be realized in visualization program of defect spatial position.Visual display of defects spatial distribution in weld has important significance on evaluation of joint reliability. Therefore defects visualization program based on Matlab software was developed in this paper, and the GUI interface was designed also, the automated input of bulk defects data and simultaneous display can be obtained. In the visualization program the following functions can be achieved, such as modification of weldments size, automated generation of three-dimensional model of weldments, adjustment of rotation angle, transparentizing processing, three views display and inquiry of defects information and so on. This can provide a new solution for three-dimensional display of defect positions in X-ray detection image of precision weldments with complex structure.