Table 1 Comparison between DWT and other signal processing techniques.

Time-frequency resolution

High (adaptive, multiresolution)

Only frequency

Limited (fixed window size)

High (adaptive, non-linear)

High (depending on the model architecture)

R⁴¹ & R⁴²

Non-stationary signal handling

Excellent for transients

Poor

Moderate

Excellent

Good (if trained on non-stationary data)

R⁴³

Computational efficiency

Efficient for real-time

Efficient but lacks time information

Moderate (real-time possible)

Computationally expensive

Computationally expensive

(training)

R⁴¹ & R⁴²

Data requirements

Low (signal-driven)

Low

Low

Moderate (requires preprocessing)

High (requires large datasets)

R⁴³

Interpretability

High (clear decomposition)

High (but no time localization)

Moderate

Moderate

Low (black-box nature)

R⁴⁴& R⁴⁵

Real-time applicability

Excellent for real-time

Not suitable for real-time

Moderate

Not suitable for real-time

Limited (depends on optimization)

R⁴⁶

Noise robustness

High (denoising capabilities)

Moderate

Moderate

Low

Moderate to high (depends on preprocessing)

R⁴⁴

Adaptability to fault types

High (captures different faults)

Low (no time information)

Moderate

High (adapts well to dynamic signals)

High (High (if trained on multiple fault types))

R⁴⁶