Many real-world vibration signals are non-stationary, meaning their frequency content changes over time. Capturing these dynamics is essential for extracting meaningful information and building robust machine learning models.
In Time–Frequency Analysis Methods for Feature Engineering, I review the main techniques used to analyze signals jointly in time and frequency, including spectrograms, wavelets, and other time–frequency representations. The paper explores how these methods can be used to extract informative features for condition monitoring, fault detection, and predictive modeling.
This work supports my research at the intersection of vibration analysis and machine learning, where effective feature engineering is often the key to turning raw sensor data into actionable insights.