Flower pollination algorithm for convolutional neural network training in vibration classification

Abstract

A convolutional neural network (CNN) is among the branches in deep learning (DL) which is a new field of research in machine learning. This model artificially mimics the visual cortex to learn the representation of visuals from low to high levels of abstraction and representation to mining data patterns such as image, sound, text, and signal. Despite the proven CNN advantages in various applications, training this model is challenging especially in complex scenarios. Some methods to optimize CNN training have been proposed, such as stochastic and meta-heuristic algorithms. In this paper, we propose a flower pollination algorithm (FPA) to train CNN. A CWRU bearing dataset is used to ensure the accuracy and efficiency of the proposed method. Moreover, we also compare our proposed method with the original CNN. The results show that the proposed method needs to be refined to achieve the required performance of CNN.