Gearbox is an indispensable element allied with mechanical industries to fluctuate speed and load in accordance with the requirements. More improvements in its mechanical structure and procedure can increase the efficiency of the industries. For those reasons, the gearbox was designed and manufactured very carefully with the zero tolerance of defects. Gearbox failure can lead to an increase in financial loss and decrease in production strength. Vibration signal analysis is used to monitor the

more »

... arbox condition. But, non-stationary properties of vibration signal make this procedure very challenging. Therefore, this paper presents a unique gearbox fault investigation technique based on advanced computational intelligence. Firstly, Time Synchronous Averaging algorithm was used to study the nonstationary properties of the raw vibration signal. Then, nine types of time domain arithmetical features were calculated from the pre-processed TSA vibration signal. After that, the J48 algorithm was used for significant features selection of gearbox fault and Naïve Bayes algorithm for features classification of gearbox fault. A computer-aided fault simulator was used for experimental study. MATLAB and WEKA platforms were used for signal processing, features selection, and features classification of the gearbox faults. The performance of the proposed method shows meaningful results for gearbox fault diagnosis. Abstract Gearbox is an indispensable element allied with mechanical industries to fluctuate speed and load in accordance with the requirements. More improvements in its mechanical structure and procedure can increase the efficiency of the industries. For those reasons, the gearbox was designed and manufactured very carefully with the zero tolerance of defects. Gearbox failure can lead to an increase in financial loss and decrease in production strength. Vibration signal analysis is used to monitor the gearbox condition. But, non-stationary properties of vibration signal make this procedure very challenging. Therefore, this paper presents a unique gearbox fault investigation technique based on advanced computational intelligence. Firstly, Time Synchronous Averaging algorithm was used to study the nonstationary properties of the raw vibration signal. Then, nine types of time domain arithmetical features were calculated from the pre-processed TSA vibration signal. After that, the J48 algorithm was used for significant features selection of gearbox fault and Naïve Bayes algorithm for features classification of gearbox fault. A computer-aided fault simulator was used for experimental study. MATLAB and WEKA platforms were used for signal processing, features selection, and features classification of the gearbox faults. The performance of the proposed method shows meaningful results for gearbox fault diagnosis.