Lectures

Machine Learning Applications

Event type Lecture (master´s degree)
Lecturer Prof. Dr.-Ing. J. Metternich, Prof. Dr.-Ing. M. Weigold, Prof. Dr. K. Kersting (Department of Computer Science) und Prof. Dr.-Ing. U. Klingauf
Mentoring Organization Heiko Ranzau M. Sc.
Co-Mentoring Tobias Biegel M. Sc. Bastian Dietrich M. Sc. Stefan Seyfried M. Sc. M. Sc. Amina Ziegenbein M. Sc. M. Sc.
Workload
Self-study
Module duration
Offer cycle
Credit points
Language
180 hours
142 hours
1 semester
Winter semester
6 CP
German
Contents Progressive digitization is opening up new opportunities in the value chain. With the increasing integration of sensors, relevant parameters of systems can be continuously recorded and stored in constantly growing databases. It is thus possible, for example, to monitor the health of technical systems. In addition to conventional approaches based on physical modeling, data-based methods in the field of machine learning have come to the fore in recent years. Corresponding algorithms can train models on the available database, with which new data sets can be interpreted and, for example, converted into a health status.

These approaches are not only interesting from an information technology perspective, but can benefit in particular from the expert knowledge that engineers contribute. Thus, mechanical engineers, for example, are asked to select meaningful metrics that can be transformed into data-based models.

In this lecture, students will get an application-oriented insight into the basics of machine learning based on examples from current research of the participating institutes. The lecture introduces relevant areas of statistics, data mining and algorithm development. Based on the presented use cases, the students learn to collect interesting data from an engineering point of view, to filter the data, to extract relevant features and to build models for diagnosis and prognosis using machine learning methods. Common process models are taught as well as the final evaluation and assessment of the developed methods and models.
Courses of the module vl: Machine Learning Applications (Lecture) 34 h (3 SWS)
pr: Machine Learning Applications (practical part / group work) 4 h (0.3 SWS)
Teaching content / Syllabus Theory
Application-oriented basics of Machine Learning and related areas of statistics (descriptive, explorative, inductive), Advanced Analytics, Data Mining, Data Science and Big Data; basics of Machine Learning procedures, functionalities and algorithms; development processes; basic Data Science principles and techniques: Discussion of business scenarios; Collection, sifting and quality assessment of data; Data preparation, feature engineering; Application of procedures and development environments using examples in Matlab and Python; Pointing out and evaluating possible solutions; Model selection, optimization, performance evaluation; Essential ideas for model integration in decision-making processes, recommended actions, system of systems; Examples from current research, eg. Predictive maintenance in aviation and manufacturing;

Practical group work
Application of basic principles of a software development methodology (e.g. Scrum); implementation of theoretical knowledge in a cooperative development task; practical solution development of an industrial challenge by programming and data evaluation (implementation); documentation and presentation of results;
Learning outcomes After students have successfully completed the course unit, they should be able to
  • evaluate basic developments and possible applications of artificial intelligence (machine learning) in engineering applications (e.g. mechanical engineering).
  • differentiate and explain essential concepts and (mathematical) methods in Machine Learning.
  • evaluate selected algorithms and models (e.g. from the field of diagnostics/forecasting) with regard to their performance, robustness and quality in engineering terms
  • apply learned skills in the areas of data acquisition and processing, data-based modeling (diagnostics and prognostics), and prescriptive analysis.
  • independently structure simple and medium analysis tasks using process models (CRISP/OSA-CBM), implement them on the basis of data and estimate their economic value
Participation requirements Programming knowledge in Matlab and/or Python is required
Form of examination Special form50 % written exam (60 min) and 50 % practical group work (during the semester) of a cooperative development task (“Hackathon”) incl. implementation, documentation and presentation (extra date)
Prerequisite for the award of credit points Passing of both exams (written exam & group work)
Grading Standard (numerical grade)
Applicability of the module WPB Master MPE III (elective subjects from natural and engineering science)
WPB Master PST III (Subjects from natural and engineering science for paper technology)
Literature Lecture materials will be made available on Moodle throughout the semester.