Listing translated from German by Talent Club.
About the role:
Background and scientific classification
Modern production sites have extensive building automation and historian systems in which operational, status and measurement data of technical systems are continuously recorded. Particularly in regulated production environments, the early detection of persistent deviations, inefficient operating modes and critical conditions is highly relevant, as these can influence both technical availability and energy efficiency, operational stability and quality-relevant ambient conditions.
This master's thesis therefore focuses on the development and evaluation of an AI-supported method for identifying chronically critical operating states using building automation time series, in order to subsequently also develop optimization approaches for improving ecological and economic parameters.
What you will achieve:
Objective of the master's thesis
The aim of the work is the development, prototypical implementation and systematic evaluation of an AI-supported method for detecting chronically critical operating states in multi-year building automation time series. The focus is on the question of how permanent or recurring deviations from expected system behavior can be identified, characterized and translated into technically interpretable findings using data-driven methods.
Research questions
Which characteristics of chronically critical operating states can be identified in multi-year building automation time series, and how can these be distinguished from short-term outliers or operationally expected fluctuations?
Which data-driven anomaly detection methods are suitable for analyzing historical building automation data with regard to detection performance, robustness and interpretability?
How can identified anomalies be evaluated and presented in a structured manner with regard to affected systems, control loops, room conditions or external influencing factors?
What prerequisites and limits arise for a practical transfer of the developed approach into an operational monitoring or decision support system?
Methodological framework
The work comprises the structured preparation and exploratory analysis of multi-year time series data from building automation, the selection and implementation of suitable anomaly detection methods as well as the systematic evaluation of the results based on professionally defined criteria. In particular, a comprehensible methodological derivation of the chosen approaches is expected, for example on the basis of statistical methods, rule-based baselines or machine learning methods for modeling normal operation and detecting persistent deviations.
A particular focus is on the evaluation of the developed method with regard to detection quality, robustness against seasonal and operational fluctuations, traceability of the findings as well as practical applicability in an industrial environment. The work should show to what extent chronically critical states can be identified automatically and what form of result preparation is useful for later operational use.
Expected results
Scientifically sound problem definition and delimitation of the application case of anomaly detection in building automation data
Preparation and analysis of a multi-year data set from system operation
Development and prototypical implementation of a method for identifying chronically critical operating states
Systematic assessment of the method with regard to detection performance, robustness, interpretability and practical feasibility
Derivation of professionally justified recommendations for the future use of the approach in technical monitoring and operational optimization
Scientific added value
The master's thesis contributes to the question of how data-based anomaly detection can be used in complex technical building systems under real industrial boundary conditions. The focus is not only on the development of a prototypical method, but also on its scientifically sound assessment, the reflection on methodological limits as well as the transferability of the results to comparable system contexts.
What you bring:
Ongoing master's degree in automation engineering, computer science, data science, technical mathematics, process engineering or a comparable field
Interest in time series analysis, machine learning and big data analysis in technical systems
Experience in the preparation and evaluation of larger databases as well as in handling programming languages or analysis tools for data processing
Structured and scientific way of working as well as ability to document methods and results in a comprehensible manner
Interest in combining scientific methodology and practical application in an industrial environment
Good German and English skills in spoken and written form
Advantageous
Basic knowledge in building automation, HVAC systems, industrial automation or process data analysis
Understanding of sensor technology, measurement data quality, control loops and operational boundary conditions of technical systems
What we offer you:
For this important and responsible position, the salary is EUR 2,010.75 gross/month (full-time, chemical industry collective agreement).
Awards as a top employer and certified family-friendly company
Comprehensive further education and training offers as well as personal development and mentoring program
Diverse development programs (talent, trainee, apprentice program)
Employee referral and recognition programs, employee stock purchase plan
Active participation in various network groups (e.g. diversity, equity & inclusion; sustainability)
Diverse health offers (e.g. free vaccinations, psychological counseling, massages)
Fitness offers
Company events & celebrations
Company restaurant with subsidized prices
Company childcare / bilingual company kindergarten
Good public transport connections