Presentation by Ning Lu

Reliability analysis for aero engine gear: Multi-fidelity surrogate with active learning - April 10, 12:45 p.m., room 116

Reliability analysis for aero engine gear: Multi-fidelity surrogate with active learning

The limit state function (LSF) can be used to evaluate the performance of structures and characterize their reliability. Reliability analysis of complex structures usually involves computer simulations, and the LSF with simulations brings a large and costly number of calls for evaluation. Against this background, the surrogate model can approximate true LSF effectively to model complex LSF at lower computational costs, and active learning can be further introduced to achieve modeling accuracy with fewer calls to LSF by adaptively selecting learning samples through the machine learning algorithm. Nevertheless, the same analysis can correspond to multiple models with different paradigms, and different paradigms or even the same paradigm usually correspond to different fidelities. Detailed and coarse paradigms are generally considered high-fidelity (HF) models with low model uncertainty and low-fidelity (LF) models with low computational cost, respectively. At this point, a promising approach to further balance the prediction accuracy and computational cost is integrating the information from HF and LF simulations by constructing a multi-fidelity (MF) model. At first, following an introduction to the background of this series of studies, this presentation will provide a general methodology for a comprehensive solution to this type of problem: For the reliability analysis based on complex limit state functions, a method based on active learning multi-fidelity Gaussian process model is proposed by combining surrogate model with adaptive strategy under the MF framework, ensuring a balance between prediction accuracy and computational cost in terms of both surrogate modeling and active learning. After that, this presentation will give an application of this method to aero engine gear, including how to realize HF and LF modeling for the gear, and the results of the reliability analysis will demonstrate the efficiency and practicality of this method. At last, this presentation will sketch the visit plan and vision.

Mr. Ning Lu is a doctoral candidate from the Center for System Reliability and Safety at the University of Electronic Science and Technology of China (UESTC). He is on a one-year visit to the Institute for Risk and Reliability at the Leibniz University Hannover (LUH). His research interests lie in the theory of structural reliability and machine learning, along with its application to advanced equipment. Related studies have been published in international journals such as Reliability Engineering & System Safety and presented at international conferences like QR2MSE, ICRSE, ICRMS, PHM, and ICMR. He has obtained Honorable Mention in the Interdisciplinary Contest in Modeling (ICM) of the Consortium for Mathematics and its Applications (COMAP), First Prize in the Provincial Undergraduate Physics Competition, Samsung Scholarship, ICMR Young Researcher Award, and National Scholarship.

Presenter Information
First Name: Ning (Chinese: 宁)
Last Name: Lu (Chinese: 鲁)
Status: Doctoral Candidate
Major: Mechanical Engineering
Department: School of Mechanical and Electrical Engineering
Institution: University of Electronic Science and Technology of China
Address: No. 2006 Xiyuan Avenue, West Hi-Tech Zone, Chengdu 611731, China