Using probabilistic neural networks for modeling metal fatigue and random vibration in process pipework

Abstract

Many experiments are usually needed to quantify probabilistic fatigue behavior in metals. Previous attempts used separate artificial neural network (ANN) to calculate different probabilistic ranges which can be computationally demanding for building probabilistic fatigue constant life diagram (CLD). Alternatively, we propose using probabilistic neural network (PNNs) which can capture data distribution parameters. The resulted model is generative and can quantify aleatoric uncertainty using a single network. Two tests are presented. The first captures the fatigue life aleatoric uncertainty for P355NL1 steel and successfully builds a probabilistic fatigue CLD. The resulted network is not only more efficient but also provides higher accuracy compared with ANN. To assess fatigue, the second test examines vibrations of a pipework assembly. The proposed methodology quantifies the nonlinear relation between the vibration velocity and the equivalent stress and successfully reflects measurements uncertainties in fatigue assessment. The proposed methodology is published in opensource format (https://github.com/MShadiNashed/probabilistic-machine-learning-for-fatigue-data).

Effect of the porosity on the fatigue strength of metals

“…12-18), and biaxial loading (test No. [19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36]. Note that, for tests performed at the same stress level (for instance, test No.…”

“…12-18), and biaxial loading (test No. [19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36]. Moreover, also the experimental values of fracture plane orientation, θ exp , 9,10 (see Section 4) are reported in Table 5.…”

“…It is worth noting that the great potential of the present procedure is that the defect content analysis (performed by means of a statistical method deriving from EVT) can be easily performed by using machine learning techniques. It deserves to be mentioned that, in the last years, a great effort has been made by researchers in order to extend the use of the machine learning techniques to the fatigue field 18–20 . As a matter of fact, it has been proven that fatigue problems are appropriate for the machine learning analysis, mainly due to the large amount, complexity, and uncertainty of the available data 21 …”

“…It deserves to be mentioned that, in the last years, a great effort has been made by researchers in order to extend the use of the machine learning techniques to the fatigue field. [18][19][20] As a matter of fact, it has been proven that fatigue problems are appropriate for the machine learning analysis, mainly due to the large amount, complexity, and uncertainty of the available data. 21 The present work is structured as follows.…”

Modelling fatigue uncertainty by means of nonconstant variance neural networks

“…Nashed et al 35 used PNNs with constant variance to model fatigue. PNNs with constant variance are suitable when the uncertainty in the data set is linear (or weekly nonlinear).…”