Accelerating 2d fault diagnosis of an induction motor using a graphics processing unit

Abstract

This paper presents a computationally efficient graphics processing unit (GPU) implementation of a reliable fault diagnosis method using two-dimensional (2D) representation of vibration signals. The fault diagnosis method first converts time-domain vibration signals into 2D gray-level images to exploit texture information from the converted images. Then, the global dominant neighborhood structure (GNS) map is utilized to extract texture features by averaging local neighborhood structure (LNS) maps of central pixels. In addition, the principle component analysis (PCA) algorithm is employed to select only the most dominant features. Finally, the selected features are used as inputs to a one-against-all multi-class support vector machine (OAA-MCSVM) to identify each fault of the induction motor. Despite the fact that the 2D fault diagnosis methodology shows satisfactory classification accuracy, its computational complexity limits its use in real-time applications. To accelerate the 2D fault diagnosis method, this paper utilizes an NVIDIA GeForce GTX 580 GPU, where all tasks are executed in parallel. The experimental results indicate that the proposed GPU-based approach achieves about 118.5 faster operation than the equivalent sequential CPU implementation while maintaining 100% classification accuracy.