Development of AI-based Diagnosis Technology using Noise/Vibration Data of Automobile Powertrain

Typically, indexes for diagnosis are developed by extracting the characteristics of noise and vibration data through traditional processing. However, finding an appropriate signal processing method and diagnostic index is difficult and time-consuming. Fortunately, the use of artificial intelligence in analyzing and judging data has increased in recent years, and a lot of research related to this topic has progressed. This study focuses on the development of AI-based diagnostic technology using noise and vibration data measured from an automobile powertrain. The purpose of this technology is to reduce quality cost and improve service efficiency. The first case involves a technology for diagnosing parts that cause abnormal noises in the powertrain, by training artificial intelligence using the collected noise data. The second involves a technology that uses engine vibration data to find cylinders with abnormal injectors in the engine. The deep learning methods used here were RNN and DNN. The developed diagnostic technologies have been applied to the equipment used for mechanics in our service centers. Therefore, they can be used to verify the diagnosis results within seconds when noise or vibration data is input to the equipment.