PENG9650 Sannsynlighetsbasert maskinlæring for prediktivt vedlikehold av energisystemer Emneplan

In today's fast-paced energy industry, predictive maintenance plays a crucial role in ensuring the reliability and efficiency of energy systems. This course is designed to equip students with advanced knowledge and skills in applying probabilistic machine learning techniques to optimize maintenance strategies for energy systems.

The course features a structured progression, starting with foundational concepts in systems engineering and energy systems, transitioning to maintenance engineering, and advancing to probabilistic machine learning techniques such as Gaussian processes, hidden Markov models, probabilistic graphical models, and deep belief networks. Real-world case studies provide hands-on experience, enabling doctoral students to bridge theoretical knowledge with practical applications in predictive maintenance.

A course in Statistical Learning similar to ACIT4510

A course in Python Programming similar to MEK1300

Students who complete the course are expected to have the following learning outcomes, defined in terms of knowledge, skills and general competence:

Knowledge:

• Understand the principles of systems engineering and their relevance to energy systems and predictive maintenance.
• Gain a comprehensive overview of energy systems, from fossil fuels to renewables, and the unique challenges they present.
• Acquire advanced knowledge of probabilistic machine learning techniques, including Gaussian processes, hidden Markov models, probabilistic graphical models, and deep belief networks.
• Understand the integration of machine learning techniques into predictive maintenance frameworks and their impact on system reliability.

Skills:

• Apply systems engineering principles to design and analyze energy systems, ensuring efficient integration of predictive maintenance strategies.
• Develop and implement predictive maintenance frameworks tailored to energy systems, transitioning from reactive to proactive maintenance approaches.
• Employ probabilistic machine learning techniques, such as Gaussian processes and graphical models, to model system behavior, predict failures, and optimize performance.

Competence:

• Collaborate effectively across disciplines to design solutions for predictive maintenance in diverse energy system contexts.
• Evaluate uncertainties in predictions and make informed decisions to improve the reliability and efficiency of energy systems through predictive maintenance strategies.
• Demonstrate proficiency in implementing probabilistic machine learning algorithms using Python and relevant libraries for energy system predictive maintenance tasks.

Teaching methods will include lectures, group work and guest lectures from industry personnel.

The following coursework requirements must be approved in order for the student to take the exam:

One assignment project. The assignment project should include a detailed project report (1500-2000 words) and a Python-based implementation of the model.

Students must analyze real-world or simulated datasets to identify failure patterns, apply probabilistic machine learning techniques, and propose a maintenance strategy for a selected energy system (such as wind turbine).

An individual project report approximately 4000 - 6000 words, excluding appendices.

If a project report is graded fail or if a medically certified illness prevents you from submitting the exam within the appointed deadline, the candidate has one opportunity to resubmit a revised report within a given time-period.

The exam can be appealed.

All aids are permitted.

Pass or Fail

Two internal examiners. External examiners are used periodically.

Arvind Keprate

The course overlaps 2.5 ECTS with PENG9560 - Topics in Artificial Intelligence and Machine Learning. Completing both courses will result in a total of 17,5 ECTS.