讲座时间：2025年12月18日(周四) 15：00
地点： 综合楼644会议室
讲座一：Condition-based maintenance strategy for multi-component systems with spatio-temporal degradation propagation
报告人简介：
赵秀杰，天津大学管理与经济学部工业工程系英才副教授。主要研究领域为可靠性与质量管理、实验设计、质保策略优化、系统维修优化与工业统计。曾在IISE Transactions，European Journal of Operational Research，Journal of Quality Technology等国际期刊上发表论文50余篇,主持1项自然科学基金面上项目，1项自然科学基金青年项目。
报告摘要：
For complex industrial systems, the degradation of a certain component is influenced not only by its intrinsic deterioration but also by health states of its neighboring components. We propose a maintenance optimization framework for systems where the degradation of components is simultaneously governed by the generation and propagation processes. The maintenance problem is formulated as a Markov decision process (MDP), which is then transformed into a factored Markov decision process (FMDP) by decomposing the factor structure of the system to reduce algorithmic complexity. Under the FMDP framework, we formulate the optimal maintenance policy to minimize the cumulative discounted cost and derive the structural properties which reveal a control-limit structure for the optimal policy. Numerical examples are provided to verify the effectiveness of the proposed model. The results demonstrate that the degradation propagation process exerts a significant impact on the optimal maintenance strategy, highlighting the importance of proactive preventive maintenance for components significantly affected by this propagation process. Compared to the proposed model, ignoring spatio-temporal degradation propagation results in a 65.88% increase in maintenance cost.
讲座二：智能制造系统中的设备可靠性与维修策略
报告人简介：
何曙光，天津大学管理与经济学部英才教授、博士生导师。主要研究方向包括质量管理、质量工程、质保管理及其商业应用。承担国家自然科学基金重点项目1项，面上项目多项，参与完成国家自然科学基金重点项目两项、国际合作项目两项。发表学术论文90余篇，研究成果发表于IISE Transactions, Journal of Quality Technology, Reliability Engineering and System Safety, International Journal of Production Economics、系统工程学报、系统工程理论与实践等。出版学术专著1部、研究生教材一部。获第11届天津青年科技奖，天津市科技进步二等奖和社会科学优秀成果奖等。兼任中国现场统计研究会可靠性分会副理事长、天津市工业工程学会副理事长、天津市现场统计研究会副理事长等。
报告摘要：
随着智能制造技术的发展和工厂无人化的不断推进，如何对生产设备进行运行维护，是智能制造工厂的稳定运行和低成本生产的重要保障。本报告将从智能制造的视角，结合课题组近年来的相关研究，从设备状态监控、设备退化与生产调度的协调、考虑设备退化对产出质量影响的设备维护策略等层面，介绍可靠性相关研究在智能制造工厂中的应用前景和存在的问题。
讲座三： Statistical Test-based Adversarial Client Detection in Federated Learning
报告人简介：
张忠良，博士毕业于东北大学，现任杭州电子科技大学管理自慰高潮 教授、博士生导师，副院长，兼任中国管理科学与工程学会智能决策分会副主任委员、浙江省信息资源管理学会副理事长等，入选杭州电子科技大学“优秀骨干教师支持计划”，是浙江省高校领军人才培养计划青年优秀人才，作为项目负责人主持省部级以上课题5项，包括国家自然科学基金面上项目2项、青年项目1项，浙江省自然科学基金重点项目1项和浙江省哲学社会科学规划课题1项，承担企业委托课题2项，作为核心成员参与包括国家自然科学基金重点项目、浙江省社科规划重大课题等省部级以上课题10余项，在相关课题资助下从事数据科学与商务智能等方面的教学与研究工作，已在管理科学学报、系统工程理论与实践、中国管理科学、PR、DSS、IPM、COR、KBS、INF等国内外知名期刊发表学术论文70余篇，其中SCI/SSCI检索45篇，Google他引1000余次，H-index为17，CSSCI检索6篇，3份咨政报告获省部级以上领导肯定性批示，授权中国发明专利30余件(其中9件已转让)，合作出版专著2部，获浙江省哲学社会科学优秀成果奖一等奖和二等奖各一次。
报告摘要：
Federated Learning (FL) is an innovative decentralized machine learning paradigm that enables multiple data owners to collaboratively train models while preserving data privacy. However, FL systems are susceptible to adversarial attacks that can significantly compromise the robustness and performance of the global model. Existing defense methods always utilize the similarity measure to detect adversarial clients which struggle with false positives and do not adapt to dynamic anomaly detection. To this end, we introduce Federated Learning Gaussian Mixture Model (FLGMM), a statistical defense method designed to detect adversarial clients and to enhance the robustness of FL against various poisoning attacks. In addition, we also develop a novel defense method Federated Robust Regression (FedRR) to effectively detect malicious clients in non-IID scenarios. Extensive experiments indicate that our proposed methods are powerful in identifying malicious clients, which lays the groundwork for future extensions that integrate advanced SPC techniques (e.g., self-starting control charts) to cope with complicated FL scenarios.
