课程名称：
《机械工程国际前沿导论-工程系统预测与健康管理》
Prognostics and Health Management of Engineering Systems
课程学分：
2学分
课程简介：
工程系统预测与健康管理是现代工程师所需的一门新兴课程。它利用物理模型和人工智能来预测系统健康的未来行为和剩余使用寿命，以确定适当的维护计划。本课程以美国佛罗里达大学机械与航空航天工程系Nam-Ho Kim教授自己的教材《工程系统预测与健康管理》为基础进行讲解。该课程于2021年在佛罗里达大学开设。
本课程的目的是介绍预测工程系统未来运行状况和剩余使用寿命的方法，以确定适当的维护计划。本课程将不仅介绍各种预测算法，还从模型定义、模型参数估计以及处理数据噪声和偏差的能力方面，解释相关预测算法的属性和优缺点等。通过该门课程，学生还将深入学习并使用MATLAB编程。
授课对象：
面向机械工程学院大三及以上本科学生
授课时间：
2023年春学期每周二晚上6：30-9：30，具体授课内容见下表。
授课地点：
线上授课
主讲人简介：
Nam-Ho Kim教授是机械设计领域国际知名学者，目前是美国佛罗里达大学机械与航空航天工程系丹尼尔·C·德鲁克冠名教授。他于1999年毕业于美国爱荷华大学机械工程系，获得博士学位，并在计算机辅助设计中心担任博士后，直到2001年。他的研究兴趣包括不确定性设计、预测和健康管理、非线性结构力学和设计敏感性分析，著有七本专著和200多篇期刊与会议文章，是美国航空航天学会Associate Fellow，《Journal of Mechanical Design》、《Structural and Multidisciplinary Optimization》、《International Journal of Reliability and Safety》等机械设计领域国际著名期刊的编委。
合作教师：
胡伟飞研究员(weifeihu@zju.edu.cn)、张益鸣研究员(yimingzhang @zju.edu.cn)
课程考核方式：
课程作业+两次测试+课程项目(占比20%+50%+30%)
教材与参考资料：
[1] Nam-Ho Kim, Dawn An, Joo-Ho Choi, Prognostics and Health Management of Engineering Systems: An Introduction, 1st Ed., Springer, 2017.
预修知识：
理论力学、材料力学、工程测试、机械原理、微积分、线性代数等
《机械工程国际前沿导论-工程系统预测与健康管理》2023年教学内容
Unit | Topic | Content | Lecture Hours |
1 | Introduction to prognostics | The basic ideas of PHM are introduced along with historical backgrounds, industrial applications, reviews of algorithms, and the benefits and challenges of PHM. | 4 |
2 | Tutorials for prognostics | Before discussing individual prognostics algorithms in detail, prognostics tutorials with Matlab codes using simple examples are presented. Using simple polynomial models with the least-squares method, the most important attributes of various prognostics algorithms will be learned. | 4 |
3 | Bayesian statistics | Many prognostics algorithms utilize Bayes’ theorem to update information on unknown model parameters using measured data. For the purpose of prognostics, this course focuses on how to utilize prior information and likelihood functions from measured data in order to update the posterior probability density function (PDF) of model parameters. | 6 |
4 | Physics-based prognostics methods | Students will learn several algorithms, such as nonlinear least squares, Bayesian method, and particle filter. The major step in physics-based prognostics is to identify model parameters using measured data and to predict the remaining useful life using them. | 6 |
5 | Data-driven prognostics methods | Data-driven approaches use information from observed data to identify the patterns of the degradation progress and predict the future state without using a physical model. As representative algorithms, the Gaussian process regression and neural network models are explained. | 6 |
6 | Applications | Various applications of prognostics will be discussed, including fatigue crack growth, wear in a revolute joint, prognostics using accelerated life test data, and fatigue damage in bearings. | 6 |
| Homework | Homework is an essential part of this course. Various programming and formulation problems will be assigned. Late homework will not be accepted. |
|
| Exams | There will be two in-term exams but no final exam. Exams will be open-book and an engineering calculator is allowed. Exam questions are mostly problem-solving questions. |
|
| Project | There will be a term project. The project is related to computer implementation of prognostics algorithms using different prognostics methods. |
|