Course - Data Driven Prognostics and Predictive Maintenance - TPK4450
TPK4450 - Data Driven Prognostics and Predictive Maintenance
Technical systems embed more and more sensors and calculators making it possible to monitor their health. We are moving from reactive maintenance strategies to condition-based and predictive ones. In addition, installations, infrastructures and critical infrastructures are ageing in Norway. The problem of their maintenance and their lifetime extension is arising. These issues can be reasonably addressed by a joint use of relevant methodologies related to condition monitoring, data driven prognostics and maintenance optimization.
This course is organized around the three following main blocks of competence.
1) Diagnosis: concepts, methods and tools to detect that a system is experiencing a deviation from the nominal state and to isolate the root cause of the deviation/degradation. Statistical methods and data driven ones (Machine Learning) will be presented.
2) Prognosis: concepts, methods and tools to built a prediction model of the system degradation with historical data. Model based methods and data driven ones (Machine Learning) will be presented.
3) Decision optimization: concepts, methods and tools to define and to optimize preventive decisions (avoid failure/production loss/accident with minimal intervention costs). Analytical formalism and Monte Carlo simulation will be used.
A special focus will be put on the links between these three blocks and how they can operate altogether.
Knowledge: what is the consistency principle and overview of different approaches for diagnosis (statistical, data-based and machine learning), what is the concept of remaining useful lifetime, overview of models for degradation and/or prediction (trends models, physic based models, time series models, stochastic processes and data driven approaches based on machine learning concepts), understanding of the concept of preventive decision rule, condition-based ones and predictive ones, overview of relevant assessment tools including analytical formalism and Monte Carlo simulation. Some inputs regarding optimization techniques will be gienve (optimization with constraints, multi-objectif optimization, robust optimization).
Skills: in a given situation, be able to make a status, to identify what are the lacks and/or how to manage with the given inputs in terms of diagnosis, prognosis and maintenance optimization. Be able to have a global vision of the whole modeling frameworks and technical issues from data processing to decision making in operation.
General competence (attitudes): be part of or lead a team for system/infrastructure/assets health management or maintenance management with a good background in quantitative analyses. This course is relevant for future manager, for future engineers in research, development and for future researchers in the academia.
Learning methods and activities
The course will be led as seminars
The students will have to implement concepts and methods on use cases and will develop Matlab or Scilab codes (or any equivalent language) or use Matlab toolboxes for this purpose. Several use cases will exemplify the three main blocks of the course.
Lectures, project work and exercises. A mandatory project shall be carried out, and will count 30% in the evaluation. The lectures, exercises and exam are in English. Students are free to choose Norwegian or English for written assessments.
Further on evaluation
Portfolio assessment is the basis for the grade in the course. The portfolio includes a semester paper counting 30 % and a written exam counting 70 %. The results for the parts are given in %-scores, while the entire portfolio is assigned a letter grade according to the grading scale using percentage points. Mandatory work from previous semester can be accepted by the Department by re-take of an examination if there haven't been any significant changes later. By a re-take of an examination all assessments during the course, that counts in the final grade, have to be re-taken.
If there is a re-sit examination, the examination form may be changed from written to oral.
Exam registration requires that class registration is approved in the same semester. Compulsory activities from previous semester may be approved by the department.
Recommended previous knowledge
Basic course in statistics and probability theory.
TPK4120 - Safety and Reliability Analysis.
Some skills in using scripting languages and toolboxes (e.g. Matlab, Python).
Required previous knowledge
TPK4120 Safety and Reliability Analysis.
One dedicated compendium and selected chapters in several books.
Credits: 7.5 SP
Study level: Second degree level
Term no.: 1
Teaching semester: AUTUMN 2020
No.of lecture hours: 3
Lab hours: 2
No.of specialization hours: 7
Language of instruction: English
- Production and Quality Engineering
Department with academic responsibility
Department of Mechanical and Industrial Engineering
Examination arrangement: Portfolio assessment
- Term Status code Evaluation form Weighting Examination aids Date Time Digital exam Room *
- Autumn ORD Semester assignment 30/100 A
Room Building Number of candidates
- Autumn ORD Written examination 70/100 A
Room Building Number of candidates