Industrial machine tool component surface defect dataset

Area of Science:

• Engineering
• Data Science
• Materials Science

Background:

• Machine learning (ML) and deep learning (DL) require substantial data, often scarce in technical fields.
• Manual inspection of industrial components and products is labor-intensive and ripe for automation.

Purpose of the Study:

• To provide a real-world dataset of ball screw drive spindle defects for ML model training.
• To facilitate the development of automated classification, wear prognostics, and predictive maintenance models.

Main Methods:

• The dataset comprises images detailing the progression of defects on ball screw drive spindles.
• An initial object detection model was used for preliminary analysis of the dataset.

Main Results:

• The dataset captures defect progression, enabling analysis of wear patterns.
• Initial analysis confirmed the utility of object detection for defect identification.

Conclusions:

• The dataset is valuable for creating robust ML-based failure detection and forecasting models.
• This resource supports condition monitoring and predictive maintenance strategies in industrial settings.