2024伯明翰国际青年学者论坛-AI与数字化分论坛议程

Litao Zhu received his B.S. (2014) in Chemical Engineering from Dalian University of Technology, and his M.Eng. (2017) and Ph.D. (2021) in Chemical Engineering from Shanghai Jiao Tong University (SJTU). Dr. Zhu is now a postdoc at the Imperial College. His research interests include multiphase flows and reactors, multiscale CFD modeling, process design, data-driven modeling and optimization, and machine learning. He has published over 30 as the first or corresponding author in peer-reviewed journals such as AIChE J, and Chem Eng Sci. He delivered several invited talks and reviewed papers for many scientific journals. Dr. Zhu received several awards, including the Humboldt Fellowship, the Banting Fellowship, the CPCIF-Clariant Clean Tech Award, and the Chambroad Science and Technology Award, etc. Recently, he was invited to serve as a Co-Guest Editor for special issues in Ind Eng Chem Res and Phys Fluids.

报告内容：

Machine learning and data science in chemical engineering

Chemical engineering is a data-rich field. Chemical engineers and scientists collect and analyze data to understand flow patterns, develop empirical models, design and optimize chemical reactions, and monitor and control chemical processes and systems. Machine learning (ML) methods can be applied to approximate highly complex and nonlinear relationships between input and output variables. This type of problem is ubiquitous in chemical engineering, including but not limited to the prediction of molecular properties from structures, the development of closure models for flow dynamics and transport, the reconstruction and identification of flow patterns, the quantification of uncertainty in physical models, the prediction and optimization of reaction outcomes and reactor performance based on reaction conditions, assistance in building better catalyst models, understanding catalytic mechanisms, and predicting process responses to control actions. This talk will focus on presenting several examples of our recent work utilizing ML techniques in hybrid modeling and optimization of chemical engineering problems. These include the development of closure models for flow dynamics and transport, and an integrated method for multi-objective optimization of chemical reactor performance.

Dr Jiajun Zhang，拉夫堡大学计算机科学系Research Fellow。他的研究主要集中在深度学习、计算机视觉和人工智能。他将人工智能技术应用在风力涡轮机叶片缺陷检测上，对该领域做出了重要贡献。他已在知名会议、期刊（如《Journal of Imaging》）上发表了多篇文章。

报告内容

风力发电机上的AI - 缺陷检测、分析及解释

风力发电机通过其叶片的转动将风能转化成电能，为我们的生活提供了许多绿色可再生能源。但因为风力发电机的尺寸和运行环境，导致其维护工作相对来说比较困难并具有危险性。使用基于AI的技术（例如，机器/深度学习，计算机视觉等）可以帮助工程师更早地定位、检测并分析叶片表面的缺陷，辅助他们做出正确的决定。此外，对AI的输出提供合适的解释可以进一步的帮助工程师们对AI输出的结果进行正确判断。

Xu He，伯明翰大学机械工程专业在读博士，CASE车辆教育与研究中心成员。研究方向为混合专用发动机的数据驱动建模以及数字孪生优化。在Engineering Application of Artificial Intelligence，IEEE/CAA, Energy, International Journal of Powertrain等国际期刊发表论文。协助团队导师参与与国际知名企业的智能化开发项目。

报告内容：

基于条件对抗生成神经网络的发动机爆震迁移建模

本次报告将介绍一种新的数据增强交叉模型方法，用于提高内燃机爆震检测的效率和精确性。随着信息技术的发展，数据驱动技术被引入到爆震检测模型中，以加速模型开发过程，并能够高精度预测复杂的非线性物理现象。然而，数据驱动模型极度依赖训练数据，数据不足会显著限制模型的预测准确性，并且计算成本高，这限制了其实时应用。为了解决这些问题，本文提出了一种使用条件生成对抗网络（CGAN）进行数据增强的交叉模型方法，通过这种方法可以加速高质量数据的生成，并在保持计算成本和预测精度之间取得适当的平衡。

Xi He, 伯明翰大学计算机学院三年级博士研究生。目前研究方向主要包括组合优化，机器学习算法设计，基于范畴论的函数式编程。博士研究课题为精准机器学习算法的设计原则。

报告内容：

Recursive optimization formalism: Exact and efficient machine learning algorihm design principles.

Exact machine learning algorithms receive far less attention compared to approximate algorithms. This trend predominantly stems from the NP-hard nature of exact solutions for these machine learning problems. We present a generic algorithms design framework to construct effcient and exact machine learning algorithms. It subsumes the classical recursive optimization methods, such as greedy algorithms, dynamic programming, divide-and-conquer, and branch-and-bound algorithms. All algorithms designed in this framework are embarrassingly parallelizable.

Abdul Azis Abdillah received his M.Sc. in mathematics from the University of Indonesia in 2012. He is currently pursuing a Ph.D. in mechanical engineering at the University of Birmingham (UoB). His research interests include artificial intelligence, machine learning, and their applications in engineering field.

报告内容：

State of Health Estimation of Lithium-ion Battery for EV using Machine Learning Algorithms

The estimation of the State of Health (SOH) of Lithium-ion batteries is crucial for the reliable operation and longevity of Electric Vehicles (EVs). This study explores advanced machine learning algorithms to enhance the accuracy and efficiency of SOH predictions. We evaluate various models, including Random Forests (RF), XGBoost, and Neural Networks (NN), comparing their performance against traditional estimation techniques. Extensive datasets from battery cycling tests are utilized to train and validate these models. Key features such as voltage, current, temperature, and charge/discharge rates are extracted and analyzed. The findings indicate that machine learning models, particularly Neural Networks, achieve high performance in predicting battery SOH. This research underscores the potential of integrating machine learning algorithms in battery management systems, promising improvements in EV performance and safety. Future work will focus on real-time implementation and the integration of these models with onboard diagnostics for dynamic SOH assessment.

2024

University of

Birmingham