Analysis of Acceleration Data Using Low-Power Embedded Devices to Detect Gear Faults

Gear condition monitoring can prevent unexpected downtimes or sudden failure of machinery. Since gear damage usually results from tooth contact, data for reliable fault detection should be acquired as close as possible to this engagement to reduce other components' disturbances (such as vibrations). One typical gear damage mechanism is pitting. Although the detection of gear pitting using acceleration data is already covered in research, methods with integrated sensors and electronics into the gear (in-situ) are still in their infancy. Most fault detection approaches still rely on external high-performance measurement systems unsuitable for in-situ integration. Thus, this paper proposes an algorithm pipeline for detecting gear pitting using acceleration data suitable for low-power embedded devices, such as Microcontrollers (MCUs). Downsampling provides the minimum required acceleration data sample rate necessary for detection. It is the basis for future work on suitable sensor and hardware selection. Finally, implementing the algorithm pipeline on a PC and a low-power ARM-Cortex M0+ MCU shows its applicability.