This paper presents a novel active noise cancellation (ANC) method to reduce the engine noise inside the cabin of a car. During the last three decades, many methods have been developed for the active control of a quasi-stationary narrowband sinusoidal signal. However, since the interior noise signal is non-stationary with a fast frequency variation when the car accelerates rapidly, these methods cannot stably reduce the interior noise. The proposed method can reduce the interior noise stably even if the speed of the car is changed quickly. The method uses an adaptive filter with an optimal weight vector for the active control of such an engine noise. The method of determining the optimal weight vector of an adaptive filter is demonstrated. In order to validate the advantages of the proposed method, a conventional method and the proposed method are simulated with three synthesized signals. Finally, the proposed method is applied to the cancellation of booming noise in a sport utility vehicle. We demonstrate that the performance of the ANC system with the proposed algorithm is excellent for the attenuation of engine noise inside the cabin of a car.
This paper proposes a condition monitoring method for the early defect detection in a chain sprocket drive (CSD) system and classification of fault types before a catastrophic failure occurs. In the operation of a CSD system, early defect detection is very useful in preventing system failure. In this work, eight fault types associated with the CSD system components, such as the gear tooth, bearings, and drive motor shaft, were arbitrarily damaged and incorporated into the CSD system. To detect the fault signals during the CSD system operation, the vibration was measured using an Internet of Things (IoT) device, which features a wireless MEMS accelerometer, Bluetooth function, Wi-Fi function, and battery. The IoT device was mounted on the gearbox housing. The measured one-dimensional vibration time-series was transformed into time-scale images using continuous wavelet transform (CWT). A convolution neural network (CNN) was employed to extract deep features embedded in the images, which are closely related to fault types. To update the learning parameters of the CNN, the RMSprop learning algorithm was applied, and the CNN was trained using 500 image samples. Multiple-classification performance of the trained network was tested using 100 image samples. Feature maps for different fault types were obtained from the final CNN convolution layer. For the visualization of fault types, t-stochastic neighbor embedding was employed and applied to the feature maps to convert high-dimensional data into two-dimensional data. Two-dimensional features enabled excellent classification of the eight fault types and one normal type.
This paper proposes a health monitoring method for the early detection of defects in a chain sprocket drive (CSD) system and classification of fault types before a catastrophic failure occurs. In the operation of a CSD system, the early detection of defects is very useful to prevent system failure. Specially, if the type of defect is known, it will be easy to find a method to fix it. In this work, eight fault types in the components of the CSD system, such as the gear tooth, bearings, and shaft of the drive motor, were arbitrarily made and assembled. To detect the fault signals during the CSD system operation, the vibration is measured by an Internet of Things (IoT) device, which features a wireless MEMS accelerometer, Bluetooth function, Wi-Fi function, and battery. The IoT device is mounted on the gearbox housing. The measured one-dimensional vibration time-series is transformed into time-scale images by continuous wavelet transform (CWT). A convolution neural network (CNN) is employed to extract deep features embedded in the images, which are closely related to fault types. To update the learning parameters of the CNN, the RMSprop learning algorithm is applied, and the CNN is trained using 500 image samples. Multiple classification performance of the trained network is tested using 100 image samples. Feature maps for different fault types are obtained from the final convolution layer of the CNN. For the visualization of fault types, t-stochastic neighbor embedding is employed and applied to the feature maps to convert high-dimensional data into two-dimensional data. Two-dimensional features enabled excellent classification of the eight fault types and one normal type.