Chaobo Chen, Xi’an Technological University, China 陈超波，西安工业大学

Biography: Prof. Chaobo Chen was born in Ningbo, Zhejiang in July 1978. He holds a Ph.D. in Engineering and is a professor and doctoral supervisor. He is currently Dean of the School of Electronic Information Engineering at Xi’an Technological University and serves as Deputy Secretary-General of Shaanxi Provincial Association of Automation. His research interests include intelligent control, fractional-order systems, fault diagnosis and fault-tolerant control. He has led two national-level projects, including the National Natural Science Foundation of China and the National Key R&D Program, as well as five provincial and ministerial research projects. Dr. Chen has received three provincial/ministerial science and technology progress awards and two second prizes for provincial teaching achievements. He has published two textbooks, been granted eight national invention patents and eight software copyrights, and authored over 50 academic papers, with more than 20 indexed by SCI/EI.

Speech Title: Research on Open-Circuit Fault Diagnosis Technology for Inverters

Abstract: Voltage-source inverters are core components of motor drive systems, and open-circuit faults in power switching devices severely affect system safety and reliability. This paper systematically investigates open-circuit fault diagnosis techniques for inverters, focusing on NPC three-level inverters from three perspectives: model-driven, signal processing, and data-driven approaches. First, a hybrid system mathematical model based on mixed logical dynamic is established, and the current waveform characteristics and mechanisms of single-tube and dual-tube open-circuit faults are analyzed. Second, fault diagnosis methods based on interval sliding mode observers, residual performance evaluation, and Kalman filtering are proposed, enabling fast and low false-alarm fault detection and localization under dynamic operating conditions. Third, signal analysis methods including short-time Fourier transform combined with discrete Fourier coefficients, the Prony algorithm, and Park-FFT-K-means are integrated to effectively extract time-frequency domain fault features, enhancing diagnostic sensitivity and robustness. Finally, data-driven diagnosis schemes based on FNN-PCA-SVM and OCS-WPD-RBFNN are constructed, leveraging supervised and unsupervised learning to achieve high-dimensional fault feature matching and localization. Simulation and hardware-in-the-loop validation demonstrate that the proposed methods achieve fast diagnosis speed and high localization accuracy under complex conditions such as load transients and noise disturbances, with the maximum diagnosis time not exceeding one fundamental cycle, offering significant theoretical value and engineering application prospects.

Dongnian Jiang, Lanzhou University of Science and Technology, China 蒋栋年，兰州理工大学

Biography: Dongnian Jiang, Ph.D., Professor, Doctoral Supervisor, Member of the Committee on Fault Diagnosis and Safety of Technical Processes, Member of the Committee on Predictive Control and Intelligent Decision Making, and supported by Gansu Province “Longyuan Young Talents” Talent Program and Gansu Province Outstanding Youth Fund.
He received his undergraduate degree from Xiamen University in 2006, PhD degree in Control Theory and Control Engineering from Lanzhou University of Science and Technology in 2018, and postdoctoral research work in Gansu Electric Power Research Institute from 2019 to 2022. He is mainly engaged in the research of artificial intelligence, intelligent sensor design and fault diagnosis. In recent years, he has presided over two projects of the National Natural Science Foundation of China, and undertaken more than 20 projects including the National Key Research and Development Program, Gansu Outstanding Youth Fund Project, Gansu Provincial Key Research and Development Program, and Gansu Provincial Natural Science Foundation Key Projects. He has published more than 60 academic papers in IEEE Transactions on Reliability, ISA Transactions, Measurement and other journals.

Speech Title: A Unified Causal Framework for Complex Industrial Systems: Methodology and Applications

Abstract: Complex industrial systems are characterized by multivariable coupling and incomplete measurement information. Traditional correlation-based modeling methods have limitations in interpretability, transferability, and intervention analysis. Following the idea of moving from correlation-based modeling to causal modeling, this talk explores a unified causal framework for modeling complex industrial systems, with a focus on the integrated modeling of mechanistic knowledge, operational data, and expert experience.
The talk will focus on three aspects: causal structure construction, causal completeness evaluation, and causal-consistency-based diagnosis. In combination with application scenarios such as equipment condition monitoring and predictive maintenance, it will further explore the potential of this framework in multi-source information fusion and intelligent operation and maintenance. This study aims to promote the transformation of complex industrial system modeling from correlation analysis to causal mechanism understanding and trustworthy decision-making.

Zhengtian Wu, Suzhou University of Science and Technology, China 吴征天，苏州科技大学

Biography: Prof. Zhengtian Wu is a Professor and Vice Dean of the School of Electronic and Information Engineering at Suzhou University of Science and Technology, China. His research interests include medical image analysis, ultrasound image processing, deep learning-based medical image segmentation, three-dimensional muscle reconstruction, and computer-aided assessment of sarcopenia.

Speech Title: AI-Assisted Gastrocnemius Ultrasound Analysis for Sarcopenia Assessment: From Automatic Segmentation to 3D Volume Estimation

Abstract: Sarcopenia is closely associated with the loss of skeletal muscle mass, reduced muscle strength, and functional decline, especially in aging and chronic disease populations. The gastrocnemius is an important superficial lower-limb muscle that can be conveniently examined by ultrasound. Compared with CT and MRI, ultrasound is non-invasive, low-cost, portable, and suitable for repeated bedside assessment, making it a promising tool for sarcopenia screening, rehabilitation monitoring, and longitudinal muscle evaluation.
This talk will introduce our research on AI-assisted gastrocnemius ultrasound analysis for sarcopenia-oriented quantitative assessment. The overall goal is not limited to image recognition, but to build a quantitative analysis pipeline that includes automatic gastrocnemius segmentation, cross-slice structural modeling, three-dimensional reconstruction, and muscle volume estimation. Accurate segmentation is the foundation of this pipeline, because reliable muscle boundaries are required for cross-sectional area measurement, volume calculation, contour tracking, and follow-up comparison.
To address the challenges of gastrocnemius ultrasound images, including speckle noise, low local contrast, ambiguous fascial boundaries, large anatomical variation across adjacent slices, and false-positive background responses, our recent work develops a Transformer U-Net-based segmentation framework with cross-slice contextual fusion, boundary-aware multi-scale decoding, and anatomical plausibility constraints. Neighboring ultrasound slices are used to provide structural context for the center slice, while boundary-aware decoding improves contour recovery and anatomical constraints help generate more coherent muscle masks.
Based on stable segmentation results, the extracted gastrocnemius contours can be further used for three-dimensional muscle reconstruction and volume estimation. These quantitative indicators may provide objective support for evaluating muscle mass, identifying sarcopenia-related changes, monitoring rehabilitation progress, and assisting clinical decision-making. This research highlights the potential of combining deep learning, ultrasound imaging, and automated quantitative modeling to support accessible and repeatable musculoskeletal health assessment.