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

Area of Science:

• Engineering
• Computer Science
• Materials Science

Background:

• Current maintenance schedules are predetermined, leading to inefficiencies and potential safety risks.
• Actual vehicle usage data can predict structural stresses for more accurate maintenance planning.

Purpose of the Study:

• To develop a machine learning (ML) and finite element analysis (FEA) approach for predicting structural stress.
• To create a surrogate finite element model using ML for real-time structural response estimation.
• To improve the safety and efficiency of maintenance scheduling.

Main Methods:

• A review of current ML applications in finite element analysis was conducted.
• A surrogate finite element approach using ML algorithms was developed to estimate the time-varying response of a one-dimensional beam.
• Various ML regression models, including decision trees and artificial neural networks, were implemented and compared.

Main Results:

• ML-based surrogate finite element models accurately estimate beam structure responses.
• Artificial neural networks demonstrated superior accuracy in predicting stress distribution.
• The FEA-based ML approach effectively estimates stress distribution during operation.

Conclusions:

• ML-driven FEA provides a powerful tool for condition-based maintenance.
• This methodology enhances the ability to define ad-hoc, safe, and efficient maintenance procedures.
• The proposed approach optimizes maintenance scheduling by utilizing real-time structural stress predictions.