Modelling fatigue uncertainty by means of nonconstant variance neural networks

Abstract

The modelling of fatigue using machine learning (ML) has been gaining traction in the engineering community. Among ML techniques, the use of probabilistic neural networks (PNNs) has recently emerged as a candidate for modelling fatigue applications. In this paper, we use PNNs with nonconstant variance to model fatigue. We present two case studies to demonstrate the developed approach. First, we model the fatigue life of cover-plated beams under constant amplitude loading, and then we model the relationship between random vibration velocity and equivalent stress in process pipework. The two case studies demonstrate that PNNs with nonconstant variance can model the distribution of the data while also considering the variability of both distribution parameters (mean and standard deviation). This shows the potential of PNNs with nonconstant variance in modelling fatigue applications. All the data and code used in this paper are openly available.

A State‐of‐Art Review on Prediction Model for Fatigue Performance of Welded Joints via Data‐Driven Method

“…[6] As shown in Figure 34, a probabilistic neural network with nonconstant variance was developed to describe the reliability of fatigue behavior, which exhibited good accuracy and robustness. [189] In addition, the random variable method and MC scheme were integrated to realize the fatigue reliability prediction of welded components. [113] Based on the equivalent stress range, a probability model for the stress response of welded joints was constructed via the random fatigue load and corrosion models (Figure 35).…”

“…Figure 34. Prediction for S-N curve of welded cover plate beams with the consideration of uncertainty: a) training and b) validation [189]. …”

A physics‐informed neural network framework based on fatigue indicator parameters for very high cycle fatigue life prediction of an additively manufactured titanium alloy

“…27,28 Mohamad employs probabilistic neural networks (PNNs) to model fatigue uncertainty, outperforming conventional Bayesian methods. 29,30 Lei et al utilize ensemble learning for accurate, stable models across 13 metallic materials, optimizing processes with domain knowledge. 31 The size, position, and morphology of defects as well as loading, were usually selected as input parameters to predict the fatigue life of AM metals.…”

Application of machine learning methods in multiaxial fatigue life prediction

“…Zhu et al 17 used an ML‐based method combined with a physics‐based parameter to evaluate the fatigue life of four materials in the very high‐cycle fatigue regime. Nashed et al 18 developed a probabilistic neural network (PNN) and evaluated it using case studies available in the literature. Lyu et al 19 applied a deep neural network (DNN) to segment fatigue fracture surface images.…”