EPF, Cachan, France. LMPS, Gif-sur-Yvette, France.

Benoit Delattre

Biography

Associate professor at EPF Engineering School Co-Head of the Sustainable Materials & Structures Major Researcher in data-driven mechanics at LMPS, Ens Paris-Saclay

Conferences

Room

Date

Hour

Subject

Room 6

19-11-2025

3:00 pm – 3:30 pm

34 Determination of adequate reference mission for reliable fatigue design using Machine Learning algorithms

Conferences Details

34 Determination of adequate reference mission for reliable fatigue design using Machine Learning algorithms

The complexity of service loads on industrial structures, such as agricultural machinery, arises from both the variability of real-world missions and their multidimensional nature. Additionally, the reliability of individual metallic components (frame, suspension, transmission, couplers, etc.) must be ensured despite the diversity of local assemblies and associated failure modes. Reference loads used in these validation processes are typically derived from industrial knowledge of service conditions. However, industrial constraints limit the scope of measurable and distributable variables, the duration of measurement campaigns, and, consequently, the amount of data available for analysis. In current paradigms, equivalent load theory is employed to maximize the utility of available data, as demonstrated by [Thomas 1998] and formalized by [Raoult 2020]. A key advantage of equivalent load approaches is that they simplify the specification of test loads for validating the reliability requirements of new projects. The nature of these test loads can vary, ranging from simple sinusoidal loads (as used in Wöhler curves) and theoretical Gaussian loading spectra (as in Gassner curves) to realistic loads derived from ad-hoc measurements, such as standard missions containing a representative number of severe events (schedules) or load ranges (spectra) [Berger 2002]. In this paper, we provide insights into developing a fatigue-motivated description of mission profiles by correlating them with realistic reference loads through equivalent load approaches. By applying a data augmentation process followed by dimensionality reduction via Principal Component Analysis (PCA) [Baroux 2022], we propose a simplified multidimensional representation of damage-inducing loads. The severities of specific usage scenarios are then simulated and distributed across various missions to determine an appropriate reference signal, ensuring a conservative validation of the project’s reliability. [Baroux 2022] Baroux, E., Delattre, B., Constantinescu, A., Pamphile, P., & Raoult, I. (2022). Analysis of real-life multi-input loading histories for the reliable design of vehicle chassis, Procedia Structural Integrity, Fatigue Design 2021, 38, 497-506. [Berger 2002] Berger, C., Eulitz, K. G., Heuler, P., Kotte, K. L., Naundorf, H., Schuetz, W., … & Zenner, H. (2002). Betriebsfestigkeit in Germany—an overview. International Journal of Fatigue, 24(6), 603-625. [Raoult 2020] Raoult, I., & Delattre, B. (2020). Equivalent fatigue load approach for fatigue design of uncertain structures. International Journal of Fatigue, 135, 105516. [Thomas 1998] Thomas, J. J., Perroud, G., Bignonnet, A., & Monnet, D. (1998). Fatigue design and reliability in the automotive industry. Technical Research Centre of Finland, Fatigue Design 1998., 1, 13-23.

An event made by Cetim