An AI-Based Nonparametric Filter Approach for Gearbox Fault Diagnosis

The gearbox has wide application in Industry 4.0 due to its power or motion transmission flexibility. The most challenging task is to improve the accuracy of gearbox fault diagnostics with optimized usage of the Internet of Things (IoT) server. For that, the research is more focused on improving the existing technique or developing a new technique so that it can be easily compatible with IoT. This article presents an AI-based nonparametric filter technique for fault diagnosis of gearboxes that focuses on the current scenario issue. The proposed technique is a combination of energy operator (EO), genetic algorithm (GA), and support vector machine (SVM). The proposed technique is improved by adding proper features whose calculations are purely based on the properties of EO, which were lacking in the existing developed technique on EO. The proposed technique is tested on the dataset obtained from the bevel gearbox test rig under different localized fault conditions. The dataset is collected at an affordable sampling rate as per Nyquist's rate so that it may optimize the use of IoT servers to a considerable extent. At the end, a comparative analysis between different filter types and similar published work is presented to show the effectiveness of the proposed technique. In comparison, our proposed technique is quite simple in computation, more focused on optimizing IoT server use, and has the ability to give higher classification accuracy on those signals which are acquired at an affordable sampling rate with a comparatively smaller number of samples per fault condition.