Gearbox Fault Detection Using Exponentially Weighted Moving Average Control Charts

Gearbox is commonly used in many manufacturing and engineering applications. An accurate condition monitoring and health assessment is necessary and important to detect faults in gearbox to monitor machinery performance and avoid machine breakdown. This paper aims to demonstrate an approach for the condition monitoring of gearbox to recognize the fault at an initial stage by applying control chart as a tool by analyzing the acceleration-time domain signals. In this paper, time series model is used to extract the features characterizing faults and the exponentially weighted moving average control (EWMA) charts are obtained to monitor the deviations in the feature values. The variation of these features is monitored using EWMA control charts. The gear fault can be simulated in gearbox by different techniques. The most commonly observed fault in the gearbox is gear tooth failure due to scoring, wear, pitting and tooth fracture and it can be simulated by partial tooth removal during the experiment. In this study experiment, the fault with increasing severity is simulated on the spur gear tooth by filing operation. Experimental data for healthy gear and gear with three stages gear tooth removal, i.e. 37.5, 75 and 100% is acquired. EWMA charts were plotted using the data collected for the above severity of faults and it is observed that the control chart is an effective tool in identifying the deviation from the normal condition.