13. August 2024

Background

In today’s rapidly evolving manufacturing landscape, data-driven decision-making has become paramount. The production of ultrasonic sensors for automative industry requires precision manufacturing processes where even minor variations can significantly impact product quality and performance. Traditional machine learning approaches have been instrumental in predicting outcomes and identifying correlations in production data. However, they often fall short in distinguishing between correlation and causation, a critical distinction for effective process optimization. This is where causal machine learning comes into play. Causal machine learning represents a paradigm shift in data analysis. It goes beyond mere prediction to understand the underlying causal mechanisms in complex systems. In the context of ultrasonic sensor production, this approach promises to:

  1. Identify true cause-effect relationships in the manufacturing process.
  2. Enable more accurate predictions of how specific interventions will affect production outcomes.
  3. Provide insights for targeted process improvements that can enhance quality, reduce waste, and increase efficiency.
  4. Support decision-making by offering a clearer understanding of the consequences of potential changes in the production line.

Our goal is to harness the power of causal machine learning to drive meaningful improvements in ultrasonic sensor production. We aim to develop models that not only predict outcomes but also explain the causal pathways leading to these outcomes, thereby enabling more informed and effective decision-making in the production process.

Keywords:

Electronic Production, Data Science, Artificial Intelligence, Causal Machine Learning

Task Description

This thesis focuses on the development and application of causal machine learning methods to identify and leverage genuine cause-effect relationships in the automated manufacturing of ultrasonic sensors. The project will involve working with real production data and state-of-the-art causal inference techniques to drive tangible improvements in manufacturing processes:

Key objectives and tasks include:

    1. Development of Causal Models:
      • Create sophisticated causal models capable of simulating specific interventions in the production process.
      • Quantify the effects of these interventions on product quality, resource consumption, and overall efficiency.
      • Utilize frameworks such as DoWhy or CausalNex to implement structural causal models.
    2. Implementation of Causal Inference Methods:
      • Apply causal inference techniques to investigate the impacts of process changes.
      • Evaluate how these changes influence production efficiency and costs.
      • Explore methods like propensity score matching, instrumental variables, or difference-in-differences for causal estimation.
    3. Analysis of Real Production Data:
      • Work with complex datasets from actual ultrasonic sensor production lines.
      • Uncover hidden causal relationships that might be overlooked by conventional analysis methods.
      • Integrate data from multiple sources, including sensor readings, image data, and process parameters.
    4. Validation and Testing:
      • test the developed models and tools using historical data
      • Collaborate with production engineers to validate findings and ensure practical applicability

Benefits

  • Application-oriented research within an industrial project in the automotive sector
  • Opportunity to work in an advanced field of artificial intelligence
  • Possibility to solve real industrial problems with cutting-edge technologies
  • Flexible working conditions with the option to work remotely
  • Theoretically, 100% remote work is possible

Prerequisites

  • Excellent knowledge of German or English language
  • Very High self-motivation and a certain affinity for IT
  • Structured and independent work style
  • Prior knowledge of Python

Please submit applications with a CV and current transcript of records via email to Sven.Meier@faps.fau.de. For further information about the scope and specific focus of the work, I am available for a personal conversation.

 

Kontakt:

Sven Meier, M.Sc., M.Sc.

Department Maschinenbau (MB)
Lehrstuhl für Fertigungsautomatisierung und Produktionssystematik (FAPS, Prof. Franke)