PSI - Issue 59

Oleh Yasniy et al. / Procedia Structural Integrity 59 (2024) 17–23 Oleh Yasniy et al. / Structural Integrity Procedia 00 (2019) 000 – 000

18 2

1. Introduction Shape memory alloys (SMA) are the basis for producing functional materials which possess the effect of shape memory and superelasticity by Andrea et al. (2018), Ma and Cho (2008), Onyshko (2018). It is known that SMA is better in some ways than traditional metals or alloys due to their properties and it shows high resistance to low-cycle fatigue. However, they have a moderate ability to withstand high-cycle loading. The regularities of mechanical behavior established for the conventional alloys can differ because of the influence of stress and strain state on the intensity of austenite and martensite transformation. The stress-strain diagram of SMA contains the large hysteresis loop while loading and unloading at a temperature higher than the finish of the austenite phase. This fact can be explained by the high dissipated energy density and strain energy level that leads to SMA's great productivity as a material for damping devices and energy storage by Kecik (2015), Iasnii and Yasniy (2019). These materials are essential for such areas of science and industry, as medicine, for instance, as the implants in the human body by Morgan (2004) and Majd et al. (2023), in dentistry due to their good mechanical properties, biocompatibility by Pandis and Bourauel (2010), corrosion resistance by Speck and Fraker (1980), also, they are widely used in machine-building and civil engineering by Auricchio et al. (2015). In particular, materials made of SMA are frequently employed in damping devices by Isalgue et al. (2006), Torra et al. (2012). While in operation, the critical structural elements are under cyclic loading, mainly of variable amplitude, which can lead to premature loss of functional properties and their failure. In turn, to assess the strength and lifetime of structural elements, it is necessary to model the material's functional properties. This task can be solved efficiently using machine learning methods, namely, by employing the methods of neural networks and random forests by Yasnii et al. (2018), Yasniy et al. (2020a,b, 2022). Machine learning is often used to find hidden or unknown relations and dependencies that increase productivity. It is known that the methods of machine learning are widely employed to solve the fracture mechanics problems, such as the prediction of fatigue crack growth rate by Yasnii et al. (2018), determination of stress intensity factor by Kim et al. (2004), estimation of stress-stain diagrams of aluminium alloys by Yasniy et al. (2020b), etc. Machine learning shows excellent results for various materials while predicting fatigue crack growth rate. For instance, in the study by Mohanty et al. (2009) there was modelled the fatigue crack growth rate in aluminium alloys for six various stress ratios R , where the input parameters were the stress intensity factor range, maximum stress intensity factor K max and stress ratio R . The paper by Dinda et al. (2004) deals with the comparison of machine learning results with the traditional approach. In the study by Ma et al. (2021), the algorithm was proposed based on neural networks that allow the prediction of the fatigue crack growth rate for 7B04 T6 aluminium alloy, as well as for TA15 titanium alloy. Whereas, the article by Vikas et al. (2021) focuses on the motion planning and control of an automated differential-driven two-wheeled E-puck robot using Generalized Regression Neural Network (GRNN). In particular, the study by Rodayna et al. (2022) highlighted the importance of artificial intelligence in SMA modelling and presented the deep connection between NNs and SMA in medicine, robotics, engineering, and automation. The paper by Trehern et al. (2022) deals with an artificial intelligence-enabled materials discovery framework that successfully identified both SMA chemistries and the associated thermo-mechanical processing steps that result in narrow transformation hysteresis and transformation range under applied stress.

Nomenclature y prediction predicted sample element y true

experimental value of the sample element

n

sample size

W d

dissipated energy

N

number of loading cycles

stress range

Δε