報(bào)告題目：控制理論輔助機(jī)器學(xué)習(xí)的復(fù)雜動態(tài)系統(tǒng)故障診斷
報(bào)告時(shí)間：2025年4月30日（星期三）下午15:30開始
報(bào)告地點(diǎn)：中南大學(xué)岳麓山校區(qū)（校本部）民主樓210會議室
報(bào)告人：李琳琳教授，北京科技大學(xué)
報(bào)告摘要：
Data-driven and machine learning (ML) methods build the mainstream in engineering diagnosis research in recent years. ML-based algorithms focus on feature generation in the data space and on system modelling by means of, technically speaking, identification and optimisation algorithms, known as training and learning as well. In particular, integration of various metrics known in statistics and optimisation theory in the training/learning process makes data-driven and ML-based algorithms considerably capable for dealing with fault diagnosis issues as classification problems. On the other hand, the model-based diagnosis framework is established on the basis of system input-output models and rigorous application of control-theoretic methods. They are efficient and reliable for engineering diagnosis in dynamic control systems. In this talk, the recent attempt to establish a uniform control-theoretic framework is introduced for fault diagnosis in dynamic control systems. The objective of establishing such a framework is twofold. On the one hand, it enables the application of ML-methods to the design of model-based diagnosis systems. On the other hand, application of data-driven and ML-algorithms to diagnosis in industrial automatic control systems will be control-theoretically guided and explained.
報(bào)告人簡介：
李琳琳，北京科技大學(xué)自動化學(xué)院教授，博導(dǎo)。2015年獲德國杜伊斯堡-埃森大學(xué)博士學(xué)位。主要研究方向包括：模型與數(shù)據(jù)驅(qū)動的故障診斷與容錯(cuò)控制、面向性能的過程監(jiān)測、控制理論輔助機(jī)器學(xué)習(xí)的智能診斷與控制等。在Springer發(fā)表學(xué)術(shù)專著一部，在IEEE系列匯刊、Automatica等控制領(lǐng)域權(quán)威期刊發(fā)表SCI檢索論文70余篇。申請或授權(quán)發(fā)明專利10項(xiàng)，獲批鋼鐵工業(yè)協(xié)會團(tuán)體標(biāo)準(zhǔn)1項(xiàng)。主持包括國家自然科學(xué)基金優(yōu)青項(xiàng)目在內(nèi)的國家級項(xiàng)目4項(xiàng)、其他省部級項(xiàng)目5項(xiàng)。擔(dān)任ISA Transactions等SCI期刊的編委。
