A Machine Learning Approach as a Surrogate for a Finite Element Analysis: Status of Research and Application to One Dimensional Systems

Abstract

Current maintenance intervals of mechanical systems are scheduled a priori based on the life of the system, resulting in expensive maintenance scheduling, and often undermining the safety of passengers. Going forward, the actual usage of a vehicle will be used to predict stresses in its structure, and therefore, to define a specific maintenance scheduling. Machine learning (ML) algorithms can be used to map a reduced set of data coming from real-time measurements of a structure into a detailed/high-fidelity finite element analysis (FEA) model of the same system. As a result, the FEA-based ML approach will directly estimate the stress distribution over the entire system during operations, thus improving the ability to define ad-hoc, safe, and efficient maintenance procedures. The paper initially presents a review of the current state-of-the-art of ML methods applied to finite elements. A surrogate finite element approach based on ML algorithms is also proposed to estimate the time-varying response of a one-dimensional beam. Several ML regression models, such as decision trees and artificial neural networks, have been developed, and their performance is compared for direct estimation of the stress distribution over a beam structure. The surrogate finite element models based on ML algorithms are able to estimate the response of the beam accurately, with artificial neural networks providing more accurate results.

From data to insight, enhancing structural health monitoring using physics-informed machine learning and advanced data collection methods

“…As discussed above, PINN uses physics as a soft penalty constraint on a lower-dimensional manifold and thus can be trained with a small amount of data while preserving vital physics properties described by the governing equations. PINN eliminates any non-realistic solutions that appear in the noisy and sparse dataset that contradicts physical laws [152]. This is the reason why PINN works exceptionally well in interpolation settings.…”

“…Moreover, DT gives insight into the system behaviour where sensors or instruments cannot be implemented easily [151]. For example, [152] explains a detailed review of applying ML algorithms to approximate the results obtained by the fidelity FE model in which the stress distribution over the entire system can be estimated and hence predict the general behaviour of a time-varying mechanical system. To develop the ML algorithms, the author used eight attributes and 1.6 million instances obtained from the FE model.…”

Numerical and Experimental Mechanical Analysis of Additively Manufactured Ankle–Foot Orthoses

“…An entirely discrete series of parameters is used to properly describe and regulate the design of a computer-aided design (CAD) model. An FEA of CAD models allows for the prediction of in situ stresses and strains by using boundary conditions that replicate real-world conditions [ 32 , 33 ]. FEA has been used to analyze and develop a range of devices in the prosthetics and orthotics field.…”

Machine Learning-Based Surrogate Model for Press Hardening Process of 22MnB5 Sheet Steel Simulation in Industry 4.0

“…The aim of surrogate models is to give an approximation of a function that relates input variables with output target variables to offer a faster response than the one that a complete simulation model provides. The surrogate models can be generated using real-world data or simulation data [ 32 , 33 , 34 ]. Despite the simplicity of the model, the response is helpful for the understanding and the optimization of the process.…”