Table 2 Calculations of time domain and frequency domain features.

From: A hybrid approach combining deep learning and signal processing for bearing fault diagnosis under imbalanced samples and multiple operating conditions

Time domain features

Frequency domain features

\(TF_{1} = \frac{1}{L}\sum _{i=1}^{L}|x_{i}|\)

\(FF_{1} = \frac{\sum _{j=1}^{n}s_{j}}{K}\)

\(TF_2 = \sqrt{\frac{1}{L}\sum _{i=1}^{L}\left( x_{i}-\bar{x}\right) ^{2}}\)

\(FF_2 = \frac{\sum _{j=1}^{K}\left( s_{j}-FF_{1}\right) ^{2}}{K}\)

\(TF_3 = \sqrt{\frac{1}{L}\sum _{i=1}^{L}\left( x_{i}\right) ^{2}}\)

\(FF_3 = \frac{\sum _{j=1}^{K} f_{j} \cdot s_{j}}{\sum _{j=1}^{K} s_{j}}\)

\(TF_4 = \frac{TF_{3}}{TF_{1}}\)

\(FF_4 = \sqrt{\frac{\sum _{j=1}^{K}\left( f_{j}-FF_{3}\right) ^{2} \cdot s_{j}}{K}}\)

\(TF_5 = |\operatorname {min}\left\{ x_{i} \mid i=1,2,...,L\right\} |\)

\(FF_5 = \frac{FF_{4}}{FF_{3}}\)

\(TF_6 = |\operatorname {max}\left\{ x_{i} \mid i=1,2,...,L\right\} |\)

\(FF_6 = \sqrt{\frac{\sum _{j=1}^{K}s_{j}^{2}}{K}}\)

\(TF_7 = TF_6 - TF_5\)

\(FF_7 = \frac{\frac{1}{K}\sum _{j=1}^{K}\left( s_{j}-FF_{1}\right) ^{4}}{\left[ \frac{1}{K}\sum _{j=1}^{K}\left( s_{j}-FF_{1}\right) ^{2}\right] ^{2}}\)