Fault Detection using Unsupervised Feature Learning on Big Data Based on Improved Weighted SoftMax Regression with Jaya Optimization

Abstract

Intelligent fault detection is promising to deal with big data due to its ability in rapidly and efficiently processing collected signals and providing accurate detection results. In traditional fault detection methods, however, the features are manually extracted depending on prior knowledge and diagnostic expertise, such processes take advantage of human ingenuity but are time-consuming. Inspired by the idea of unsupervised feature learning artificial intelligence techniques are used to learn features from the raw data.As the dimensionality increases, the accuracy of fault identification methods implemented on big data decreases significantly. For supervised learning, large volume of data is needed which leads to high cost and time consuming. In this paper, an unsupervised learning approach is proposed on the basis of weighted softmax regression for fault detection using the power signals. In the proposed approach, the features are extracted from the unlabelled data. The developed approach is based on squirrel search algorithm and voting based weighted softmax regression with jaya optimization algorithm (IWSRJO_SSA). The proposed approach is simple and easy to carry out, though it attains high accuracy when compared to that of more advanced techniques. The features from the signals are extracted normally and some of the best features are selected using the squirrel search algorithm (SSA). Because of selecting the best features and using the improved weighted softmax regression, the proposed method achieves high accuracy in fault detection. Experiment on the power signal dataset shows the applicability of the proposed approach in fault detection on big data. The proposed system will be applicable on medical field, industrial field, electrical field and so on. The experimental results prove that the proposed method attains high accuracy and is superior to the other advanced methods.