Visual Exploration and Analysis of Simulation and Testing Data in Motor Engineering

Area of Science:

• Electrical Engineering
• Data Science
• Machine Learning

Background:

• Automated defect detection in electric motors is crucial for reliability.
• Limited real-world failure data hinders data-driven defect detection methods.
• Simulated data lacks real-world complexity, impacting model performance.

Purpose of the Study:

• To introduce a visual analysis tool for comparing measured and simulated electric motor data.
• To identify domain-invariant features and assess simulation accuracy.
• To aid in selecting optimal training data for robust automated defect detection.

Main Methods:

• Development of a visual analysis tool for time-series data.
• Application of visual design principles tailored for electric motor professionals.
• Validation through a think-aloud study with specialized engineers.

Main Results:

• The tool facilitates the identification of discrepancies between real and simulated motor data.
• It aids in understanding simulation data limitations for defect detection.
• The visual design effectively supports engineers in data selection.

Conclusions:

• The proposed visual analysis tool enhances the reliability of automated defect detection systems for electric motors.
• It bridges the gap between simulated and real-world data challenges.
• The tool's design, validated by domain experts, meets practical engineering needs.