Investigation of the Rupture Surface of the Titanium Alloy Using Convolutional Neural Networks

Abstract

The research of fractographic images of metals is an important method that allows obtaining valuable information about the physical and mechanical properties of a metallic specimen, determining the causes of its fracture, and developing models for optimizing its properties. One of the main lines of research in this case is studying the characteristics of the dimples of viscous detachment, which are formed on the metal surface in the process of its fracture. This paper proposes a method for detecting dimples of viscous detachment on a fractographic image, which is based on using a convolutional neural network. Compared to classical image processing algorithms, the use of the neural network significantly reduces the number of parameters to be adjusted manually. In addition, when being trained, the neural network can reveal a lot more characteristic features that affect the quality of recognition in a positive way. This makes the method more versatile and accurate. We investigated 17 models of convolutional neural networks with different structures and selected the optimal variant in terms of accuracy and speed. The proposed neural network classifies image pixels into two categories: “dimple” and “edge”. A transition from a probabilistic result at the output of the neural network to an unambiguously clear classification is proposed. The results obtained using the neural network were compared to the results obtained using a previously developed algorithm based on a set of filters. It has been found that the results are very similar (more than 90% similarity), but the neural network reveals the necessary features more accurately than the previous method.

Effect of Shear Strain Rate on Microstructure and Properties of Austenitic Steel Processed by Cyclic Forward/Reverse Torsion

“…Macroscopically, the cross-sectional shrinkage of the L100 sample is significantly higher than that of the H100 sample, indicating that the L100 sample has better ductility. Generally, on the micro level, the fracture surfaces of ductile materials exhibit dimples generated by dislocation activities at the final fracture stage [39,40], while the brittle fracture surface often has cleavage planes [41], quasi-cleavage planes, or even rock-candy patterns [42]. Both the low-speed and high-speed twisted samples show dimples in the core.…”

Morphological Characteristics of Dimples of Ductile Fracture of VT23M Titanium Alloy and Identification of Dimples on Fractograms of Different Scale

“…The investigated fracture surface has a dimple structure formed as a result of coalescence of a set of micropores. The localization and identification of dimples in the image show that the appearance of pores was accompanied by plastic deformation of the intersections between them [18,19]. However, they are thick enough to identify each individual dimple.…”

“…It is established empirically that the dimple shape coefficient is informative for their classification. In order to detect changes in the shape of dimples, the roundness coefficient K c was used, which is equal to the percentage of dimple pixels that fall into a circle with an equivalent diameter di, the center of which is aligned with the center of mass of dimple C i (x ci , y ci ) [18,19,20]:Kc=∑m=1fig(truerm→,di)fi⋅100% where g(truerm→,Di) is the indicative function that shows whether the m-th pixel falls into the circle with equivalent diameter d i ; g(truerm→,di)={1, at |truerm→|≤di/20, at |truerm→|>di/2 where truerm…”

“…Also, we compared the results of dimples recognition by the proposed network with results obtained by a known algorithm to analyze the fractograms. It has been found that the results are very similar (more than 90% similarity), but the neural network reveals the necessary features more accurately than the previous method [19,20,21].…”

“…It should be noted that in the previous works, the authors proposed approaches that formed the basis for the development of the method of fractographic recognition, control and calculation of parameters of the dimples of tearing based on the use of neural networks [18,19,20]. In previous work, 17 models of neural networks with various sets of hyperparameters have been developed and investigated.…”