Fault detection in rotating elements by using fuzzy integrated improved local binary pattern method

An infrared thermography method is a promising tool for defect detection in rotating machines, as this approach is a non-intrusive and no-contact kind of approach. Although the performance of infrared thermography is limited by strong noise signals and the irrelevant information found in infrared images. This issue can be efficiently addressed by using an image segmentation process that can enhance feature extraction in infrared thermography image analysis. In this paper, an image segmentation process named fuzzy integrated improved local binary pattern (ILBP-Fuzzy) for fault detection is proposed to enhance feature extraction in the thermography images. In the ILBP-Fuzzy method, the thermal image is first converted into a grey-scale image and thereafter, a median filter is applied to make the image noise-free. Later on, the region of interest is identified and the fault is detected by the application of the suggested ILBP-Fuzzy approach. In this work, two cases are performed. The first one is on synthetic images and the other one is on thermography images. In case 1, a synthetic image (star image) is used to evaluate the effectiveness of the edge detection method. The outcomes from the suggested approach are compared with other methods and various parameters such as accuracy, jacquard similarity index, sensitivity, dice similarity index, and specificity are calculated. In case 2, the proposed method is tested on three types of industrial thermo-graphic images of bearings named as healthy, fault-initialized, and unhealthy for estimating the capability of the ILBP-Fuzzy approach. From the results, it is evident that the ILBP-Fuzzy approach provides superior performance to identify the conditions of the rotating elements in a machine as compared to other segmentation methods such as IHLBP, Sobel, Canny, Laplace of Gaussian, Otsu and LTIHLBP method.