近期，我院本科生林子轩（第一作者），教师范伟（通讯作者）等的研究成果Dual-attention informed probabilistic sparse identification of nonlinear dynamics for multimode dynamic process monitoring在《Measurement》（IF=5.6）上发表
论文简介如下：
System dynamics in modern industrial processes are often nonlinear, time-varying, and influenced by frequent mode transitions. These complexities present substantial challenges for accurate modeling and process monitoring. To address these issues, this paper proposes a Dual-Attention Informed Probabilistic Sparse Identification of Nonlinear Dynamics (DAPSINDy) method, which reformulates dynamic modeling as a sparse optimization problem in a latent variable space. By integrating clustering, attention mechanisms, and probabilistic modeling, DAPSINDy effectively captures multimode dynamics and variable interactions. A dual-attention structure is introduced to enhance the relevance of mode classification and the interpretability of nonlinear basis function selection. Moreover, a probabilistic Expectation-Maximization framework combined with particle filtering is used for parameter estimation under uncertainty. Two industrial case studies are conducted on a three-phase flow facility and marine engine system. Results demonstrate that DAPSINDy achieves superior performance in fault detection, false alarm reduction, and dynamic adaptability compared with existing methods.
