Evaluation of High-Frequency Measurement Errors from Turned Surface Topography Data Using Machine Learning Methods.

Area of Science:

• Manufacturing Engineering
• Metrology
• Data Science

Background:

• Surface topography evaluation is crucial for industrial manufacturing control.
• Contactless measurement techniques offer speed but are susceptible to environmental noise, particularly high-frequency vibrations.
• Accurate surface roughness assessment is vital for product quality and performance.

Purpose of the Study:

• To investigate machine learning methods for reducing high-frequency noise in contactless surface topography measurements.
• To optimize digital filtration techniques for enhanced surface roughness assessment of turned surfaces.
• To compare the efficacy of various machine learning models in predicting optimal filtration parameters.

Main Methods:

• Application of digital filters (Gaussian regression, spline) to contactless roughness measurements of turned surfaces.
• Utilizing machine learning models including neural networks, support vector machines (SVMs), and decision trees.
• Analysis of surface topography data under specific machining conditions to identify noise reduction strategies.

Main Results:

• The Gaussian regression filter and spline filter were identified as most effective at a 22.5 µm cut-off for noise reduction.
• Support Vector Machines (SVMs) demonstrated superior performance in predicting optimal filtration methods.
• The study achieved high accuracy and sensitivity in identifying effective noise reduction techniques.

Conclusions:

• Machine learning, particularly SVMs, can significantly enhance the accuracy of contactless surface roughness measurements by optimizing filtration.
• The findings provide a robust framework for mitigating vibration-induced errors in industrial surface metrology.
• Optimized filtration using ML improves the reliability of surface topography data for manufacturing process control.