Algorithm 1: Hyperparametric optimization algorithm of PSO-CNN.

Input: MAX_Iteration, boundary of the hyperparametric space θ, number of particles n

Output: Optimal super parameters and maximum fault feature recognition accuracy

1 Randomly generate n particles to form the initial population:

${\theta }_j=\left\{cn{n}_L,ci,k,d,r\right\}, j=1,2,\dots ,n$

Set an optimal super parameter (${\theta }_{best}$) and maximum accuracy (accuracy_best)

2 for I = 1:MAX_Iteration

3   for j = 1: n

4    train CNN model and output accuracy

5    accuracy $\leftarrow$ execute (CNN model)

6    if accuracy > accuracy_max

7      ${\theta }_{best}\leftarrow {\theta }_j$

8      accuracy_best $\leftarrow$ accuracy

9 Update the speed and position of each particle