CRE’s contributions to ICMD 2025 and ICCES 2025 conferences
From May 9–11 and May 25–29, 2025, Dr. Zhouzhou Song from the Chair for Reliability Engineering (CRE) participated in two international conferences: The 2025 International Conference on Mechanical Design (ICMD 2025) in Hangzhou, China, and The 31st International Conference on Computational & Experimental Engineering and Sciences (ICCES 2025) in Changsha, China. At both events, he represented the Chair and presented his research on high-dimensional uncertainty propagation.
Dr. Zhouzhou Song’s Research:
An active learning Kriging method based on improved sufficient dimension reduction for high-dimensional uncertainty propagation
Abstract:
Surrogate models are extensively employed for uncertainty propagation in complex, evaluation-expensive engineering problems. However, constructing high-accuracy surrogate models for high-dimensional uncertainty propagation with limited sampling resources remains a significant challenge. To address this, we propose an active learning Kriging method based on improved sufficient dimension reduction (AKISDR) tailored for high-dimensional uncertainty propagation. The martingale difference divergence is introduced into the sufficient dimension reduction framework to estimate a more accurate and stable projection matrix, which projects high-dimensional inputs into a low-dimensional latent space while preserving sufficient information for response prediction. The dimensionality of the latent space is determined using a ladle estimator, which considers variabilities in both eigenvalues and eigenvectors, enabling more accurate and efficient dimensionality estimation compared to existing methods. The projection matrix is embedded into the Kriging kernel function, reducing the number of hyperparameters and improving modeling accuracy and computational efficiency. Additionally, the analytic gradient of the likelihood function is derived to facilitate efficient hyperparameter estimation via gradient-based optimization. Finally, a novel adaptive sampling strategy is designed to identify the optimal next sampling point for improving the global accuracy of the surrogate model. The proposed AKISDR method is validated through numerical and engineering examples, demonstrating its capability to construct highly accurate surrogate models and efficiently perform uncertainty quantification for high-dimensional problems.
As CRE continues its mission to bridge theoretical innovation with practical impact, we are proud of our team’s contributions at ICMD 2025 and ICCES 2025. The conferences provided a valuable platform for showcasing our latest research, exchanging ideas, and fostering collaborations to advance resilient and reliable engineering systems globally.
Conferences details sourced from: ICMD2025 , ICCES2025