Artificial Intelligence (AI) has been a popular topic in recent days not only in academia/industries but also in general public. Most of the time, they refer machine learning when AI is discussed. This paper gives introduction to machine learning. First, the overview of the machine learning is given. Then, three modes in the machine learning such as supervised learning, unsupervised learning, and reinforcement learning are described. In particular, supervised learning is explained in detail. In order to give a step-by-step introduction to the supervised learning, a linear function and a perceptron are explained and develop the discussion to three-layered neural networks. Furthermore, deep neural networks are explained as the extention to the three-layered neural networks. Finally, the application of the deep neural networks are shown with some examples.

This study focusses on computer vision to be applied to fractography. Automatic image segmentation by Hough transform and average brightness in region of interest was performed to find an origin of brittle fracture in martensitic steels. In addition, it is emphasized that morphology of microstructure or fracture surface should be quantified as much as possible to extract its features.

Metals have elastic anisotropy depending on the crystal orientation, so that Young’s modulus and Poisson’s ratio have different values depending on the crystal orientation. Diffraction analysis in poly-crystalline metals yields another type of Young’s modulus termed as “Diffraction Young’s modulus and diffraction Poisson’s ratio”. Generally, the elastic anisotropy is reduced in poly-crystal due to the interaction among crystal grains. This means that diffraction Young’s modulus and diffraction Poisson’s ratio reflect the elastic deformation behavior of each crystal grain in poly-crystalline metals. In the present investigation, an ideal poly-crystal model with isotropic crystal orientations is proposed and then the average values of diffraction Young’s modulus and diffraction Poisson’s ratio are estimated for poly-crystalline metals: Al, Cu, Ni, Fe(fcc), Fe(bcc), V, Mo and Cr. In ideal poly-crystalline metals, Poisson’s ratio and Young’s modulus were theoretically estimated as follows: Al(0.345, 71.0 GPa), Cu(0.340, 131.1 GPa), Ni(0.329, 196.1 GPa), Fe(fcc)(0.292, 199.3 GPa), Fe(bcc)(0.292, 204.7 GPa), V(0.355, 136.4 GPa), Mo(0.305, 316.5 GPa), Cr(0.214, 282.9 GPa).

In the dislocation characterization using radiation diffraction, the modified Williamson-Hall (mWH) method gives the parameter α, ϕ and q, in which the information about crystallite size, dislocation density and the character of dislocation are contained respectively. The following mWH equation is applied to determine these parameters in the mWH method.

(ΔK－α)²/K ²＝ϕ²C_h00(1－qΓ )

In the mWH method, the optimal α–value is selected to give the best linear fitting on the above mWH equation. Putting the optimal α-value into the mWH equation, the other parameters ϕ and q are automatically determined in the relation between the left side term and the orientation parameter Γ. The original mWH equation is expressed by the following equation, hence the values of ΔK can be calculated on each {hkl} crystal plane applying the obtained values for parameter α, ϕ and q.

ΔK＝α＋ϕK√C, here C＝C_h00(1－qΓ )

As a result, the Williamson-Hall (WH) plots; K vs. ΔK are constructed by the calculation and they are compared with those obtained by experimentally. After that, the values of full width at half maximum are corrected to minimize the error between experimental and calculated WH plots. This process is termed as the modified Williamson-Hall feed-back (mWHFB) method. It was confirmed that the accuracy of dislocation analysis is greatly improved by applying the mWHFB method in comparison with the conventional mWH method.

The effectiveness of damage evaluation methods was surveyed among the master curve method, the parametrical statistics method and machine learning method by applying those methods to damage evaluation using KAM(Kernel Average Misorientation) parameters obtained from EBSD(Electron BackScateer Diffraction pattern) observations for the interrupted creep and creep fatigue tests of SUS304HTB equivalent heat resistant steel for boiler tube use. As for the parametric statistics method, the log-normal distribution was judged as the best fit distribution type among normal, log-normal and Weibul distributions. Being the algorithm of machine learning effective for pattern recognition, neural network was adopted for the analysis. As a result, it was found that the accuracy was higher in the order of neural network method, parametic statistics method, and master curve method. The reason why the neural network method was more accurate than the parametric statistics method was the latter method could not approximate the frequency distribution shape of KAM accurately. If the frequency distribution profile is unknown as in this case, the method like neural network independent on the distribution profile is considered to be very effective.

In order to investigate the correlation between fatigue limit and static strength of extruded Mg-Al-Zn alloys, rotating bending fatigue tests with three types of Mg-Al-Zn alloy, AZ31, AZ61 and AZ80, were carried out. And static tensile test and compression test with these alloys were also performed. On the other hand, in order to clarify the correlation between them, more than 200 past literature researches on fatigue property of various Mg alloys were surveyed. As a result of past literature research, it was found that there was a proportional relationship between fatigue limit and static strength. The fatigue limit and the static strength obtained by the experiment of this study agree well with the proportional relation obtained as a result of the literature survey. In particular, considering the super long life fatigue range up to 10⁸ cycles or more, the fatigue limit of AZ31 and AZ61 was found to be almost equal to their compressive 0.2% proof stress. However, it was found that the fatigue limit of AZ80 is lower than the compressive 0.2% proof stress. The best method to predict the fatigue limit of all materials was through the use of tensile strength.

Long-term creep rupture behavior and structural stability of a nickel-based Alloy 740H were investigated. Creep tests were conducted at 650°C, 700°C, 750°C and 800°C using smooth bar specimens, and the creep deformation and fracture were discussed in relation to the changing microstructure and hardness. It was found that primary and secondary creep stages were considerably shorter than the tertiary stage irrespective of applied stress and temperature. Creep exponent in Norton law significantly decreased with decreasing the stress from 11 at high stress region to about 1 at low stress region. Times to creep rupture were expressed fairly well by using Larson-Miller parameter with the quadratic equation of stress. The size of γ' precipitates continuously increased as aging time increased, and shape of γ' remained spherical at 650°C and 700°C but tended to change to cubic above 750°C. A higher growth rate of γ' precipitates was observed during creep, and it seems to connect in a specific direction in grain. In addition, neither η phase in grain nor G phase at grain boundaries were found. Room temperature hardness on the head portion of specimen increased initially with the aging time but decreased inversely after reaching the maximum value. The hardness on the gage portion of specimen continued to be larger than that on the head portion due to strain hardening during creep.

One of the joining methods of carbon fiber reinforced thermoplastics (CFRTP) is a direct resistance heating method and we have reported that the tensile lap-shear strength and flexural strength were improved by using carbon nanotube (CNT) grafted carbon fibers as the heating element for welding. In order to reduce the weight of CFRTP structural parts, further improvement of welding strength is needed to reduce the welding area where materials overlap. CNT/polyamide 6 (PA6) composite nanofibers in which CNTs are dispersed in PA6 nanofibers had successfully been fabricated. By using this CNT/PA6 composite nanofibers as the matrix resin of the heating element for welding, further improvement of welding strength can be achieved. In this study, the tensile lap-shear test and flexural test were carried out by using carbon fibers and CNT grafted carbon fibers as the heating element and CNT/PA6 composite nanofibers as the matrix resin for welding; in order to clarify the effect of CNT on the welding strength of resistance welded CFRTP. The welding strength was improved by using CNT/PA6 composite nanofibers with the decrease of void content in the welded zone of resistance welded CFRTP. Moreover, the lower void content and the higher welding strength were obtained when CNT grafted carbon fibers and CNT/PA6 composite nanofibers were used for the heating element and the matrix resin for welding. Due to the reinforced resin by CNTs in CNT/PA6 composite nanofibers and the improvement of the fiber/matrix interfacial shear strength by CNT grafting on the surface of the carbon fibers, the welding strength of resistance welded CFRTP was improved.