Prediction of creep failure time using machine learning

Abstract

A subcritical load on a disordered material can induce creep damage. The creep rate in this case exhibits three temporal regimes viz. an initial decelerating regime followed by a steady-state regime and a stage of accelerating creep that ultimately leads to catastrophic breakdown. Due to the statistical regularities in the creep rate, the time evolution of creep rate has often been used to predict residual lifetime until catastrophic breakdown. However, in disordered samples, these efforts met with limited success. Nevertheless, it is clear that as the failure is approached, the damage become increasingly spatially correlated, and the spatio-temporal patterns of acoustic emission, which serve as a proxy for damage accumulation activity, are likely to mirror such correlations. However, due to the high dimensionality of the data and the complex nature of the correlations it is not straightforward to identify the said correlations and thereby the precursory signals of failure. Here we use supervised machine learning to estimate the remaining time to failure of samples of disordered materials. The machine learning algorithm uses as input the temporal signal provided by a mesoscale elastoplastic model for the evolution of creep damage in disordered solids. Machine learning algorithms are well-suited for assessing the proximity to failure from the time series of the acoustic emissions of sheared samples. We show that materials are relatively more predictable for higher disorder while are relatively less predictable for larger system sizes. We find that machine learning predictions, in the vast majority of cases, perform substantially better than other prediction approaches proposed in the literature.

Size Distribution of Emitted Energies in Local Load Sharing Fiber Bundles

“…Such behavior indicates critical dynamics, particularly selforganized critical dynamics for the system, where the universality hypothesis is still applicable, without having to fine-tune a driving parameter [7]. Such a phenomenon is therefore open for analysis with the tools of critical phase transitions, universality and therefore is an important step toward predictability of imminent failure [8][9][10].…”

Machine learning predictions of COVID-19 second wave end-times in Indian states

“…In that case, the underlying lattice structure approaches a fractal and the fluctuations in the cluster size diverges with system size [32]. It is seen before that the predictability using ML approaches increases with the increase in the disorder in the system (see e.g., [17]). Indeed, the fluctuations in the time series of the various attributes used for the ML algorithm have richer characteristics (carrying more information), and consequently, the training of the algorithm is better.…”

“…As a simple way to introduce disorder, we remove a fraction q of the sites i.e., there are no individuals occupying that fraction of the sites This modification, of course, introduces a fluctuation that diverges near the critical point q c % 0:4 of site percolation [32] (see also [33] for percolation threshold with longer than nearest neighbor connections). It is generally known that a system with higher disorder is relatively more predictable through machine learning, compared to the systems having less disorder [17]. It is also known that the distribution of population in a city follow a fractal character [34], which will happen here near the percolation threshold.…”

“…However, a multidimensional set of data is often used in statistical learning approaches for making predictions [12,13]. Indeed, such predictions have been attempted in a wide range of cases, such as financial time series, weather data, medical applications and many other physical systems [14][15][16][17]. There have been multiple earlier attempts in using machine learning approaches for predicting epidemic spreading in the context of COVID-19 (see e.g., [21,22]) as well.…”

Machine learning reveals strain-rate-dependent predictability of discrete dislocation plasticity

“…Recent years have witnessed a surge in application of artificial intelligence (AI) in general and machine learning (ML) in particular to gain novel insights on properties of materials and related problems in physics [9][10][11][12][13][14][15][16][17][18][19]. Broadly speaking, such developments fall under the umbrella of the emerging research field of materials informatics [20], where informatics methods -including ML -are used to search for novel materials [21,22], establish novel structure-property relations [13,23], etc.…”