Smart Factory Setting Awareness by Noise-Pattern Analysis

Recently, to build a smart factory, research is being conducted to perform fault diagnosis and defect detection based on vibration and noise signals generated when a mechanical system is driven using deep-learning technology, a field of artificial intelligence. Most of the related studies apply various audio-feature extraction techniques to one-dimensional raw data to extract sound-specific features and then classify the sound by using the derived spectral image as a training dataset. The image data–based training model requires a lot of data, so it takes a lot of time to build the training dataset. In this paper, we propose a two-step data pre-processing method to efficiently detect anomalies based on numerical data, which is raw data. It was confirmed that the time required to build the training dataset was reduced through two steps of data pre-processing: a process for analyzing sound-signal information and feature extraction and a process for applying the proposed algorithm. The process was found to take about 9 times longer when using an image dataset than a numerical dataset.