Classification of Fault Severity in Induction Machine Systems Based on Temporal Convolutions and Recurrent Networks

Figure 8

Feature visualization via t-SNE for the test dataset: t-SNE for ((b), (c)) the LSTM structure of the ALSTM-FCN architecture, ((d), (e)) the FCN structure, and (f) the softmax layer. (a) Raw signal. (b) LSTM layer. (c) Dropout layer. (d) FCN layer. (e) Concatenate layer. (f) Softmax layer.

(a)

(b)

(c)

(d)

(e)

(f)