Frequency Domain Analysis of Sensor Data for Event Classification in Real-Time Robot Assisted Deburring.

Area of Science:

• Engineering
• Signal Processing
• Manufacturing

Background:

• Real-time process monitoring often utilizes indirect sensing methods, leading to challenges in managing and analyzing large datasets.
• High sampling rates are employed to capture infrequent process events, necessitating advanced signal processing techniques.
• Distinguishing between different vibration sources, such as spindle and finishing processes, is crucial for accurate monitoring.

Purpose of the Study:

• To investigate the correlation between power spectrum density (PSD) and the number of completed cycles/passes in a process.
• To evaluate Welch's estimate method for calculating PSD and classifying process events.
• To differentiate spindle-generated vibrations from those occurring during the finishing process.

Main Methods:

• Acquisition of sensor data at high sampling rates (40-51.2 kHz).
• Calculation of power spectrum density (PSD) using Welch's estimate method.
• Comparison of Welch's estimate with traditional statistical methods for signal analysis.

Main Results:

• A clear correlation was observed between PSD calculated via Welch's estimate and the number of completed cycles/passes.
• Welch's estimate proved effective in classifying the progress of cycles/passes.
• The study successfully distinguished vibration signals originating from the spindle versus those from the finishing process.

Conclusions:

• Welch's estimate method is a viable tool for analyzing sensor data in real-time process monitoring.
• PSD analysis using Welch's method enables accurate classification of process events like cycle completion.
• This approach enhances the ability to identify and differentiate specific vibration sources within a manufacturing process.