Research On Automatic Evaluation Of DWTT Fracture Surface Using Visual Analysis

By calculating the Drop Weight Tear Test(DWTT) specimen cross section area percentage of brittle fracture can evaluate material brittle ductile characteristics. Mainly rely on manual visual judgment, subjective factors affecting accuracy, low detection efficiency, the urgent need for automation testing instrument. But the DWTT specimen fracture mode of the image is very complex, toughness, brittleness and fissure mixed characteristics distinguish is not obvious, and the entire fracture undulating, dramatic change slope in 0 ° ~ 90 °, the peak valley general up to 30 mm, human experts discriminant has difficulty, lighting in imaging, especially image automatic identifying a great technical challenge.In this paper, according to the characteristics of the fracture design the great depth of field 140 mm in magnification invariable telecentric image acquisition system, eliminate the error of measurement caused by the thing square cross-section. In order to highlight the brittle fracture character at the same time ensure that the section of the whole shining light evenness, oblique lighting hardware design combines the high window and multi-angle composite lighting system,. Brittle fracture zone identification algorithm based on image segmentation and pattern classification. Designed the background segmentation algorithm based on hysteresis threshold section, then the fracture of the segmented area breaks up a little fragile area using the dynamic threshold met hod, then study the segmentation algorithm based on graph theory, using diagram method will the rest of the characteristics of obvious fracture area divided into brittle ductile characteristics of single sub area, finally, the characteristics of regional p attern classification. Based on the Laws texture, one-dimensional Fourier power spectrum, and gray symbiotic matrix cross section area of combining feature extraction method. Pattern classification using machine learning methods, design and optimize the support vector machine(SVM) classification model based on RBF k ernel.Experimental results show that human experts to assess the results as the standard, this paper design the DWTT of cross section image automatic evaluation system to evaluate the shear percentage of absolute error within 1%, has realized the automatic evaluation of DWTT section.