A technology was developed to recognize and predict the urgent degradation of the state of the rotating equipment installed in Hyundai-Kia factories. It is also being applied to activities to prevent equipment failures by establishing a monitoring system using this technology. Vibration data and artificial intelligence (AI) algorithms were used to predict conditions. It was developed and installed so that maintenance engineers could predict failures in advance. This is to improve preventive diagnosis of thousands of rotating equipment in the factory. And it has the advantage of allowing a small number of engineers to monitor exponentially increasing number of equipment. Vibration data including trends and alarms were collected along with the production schedule, and wavelet-based preprocessing DB9 (Daubechies 9) was performed to remove noise such as outliers. Two different AI algorithm models were developed to recognize and predict changes in equipment state. First, 1D-CNN (Convolutional Neural Network) was used as a model for initially recognizing rapid changes in vibration trend. The reliability of the model was evaluated by converting the difference between the inference result and the actual result data into numerical values based on the probability distribution. Second, a future vibration trend prediction within 7 days was developed by combining LSTM (Long-Short-Term Memory) and 1D-CNN algorithms. LSTM is well known for predicting time series data. The reliability of both models is demonstrated by RMSE (Root Mean Square Error) and MAPE (Mean Absolute Percentage Error, %). Each model of MAPE is 99.9% and 99.3%, respectively. Monitoring of equipment states has been established with two models that secured reliability, and rotating equipment at the factory are currently managed.
Developing Equipment Condition Prediction and Monitoring System Using Deep Learning Models in Automotive Production Factory
Sae Technical Papers
WCX SAE World Congress Experience ; 2023
2023-04-11
Conference paper
English
Road Traffic Condition Monitoring using Deep Learning
IEEE | 2020
|Deep Self-Supervised Learning Models for Automotive Systems
SAE Technical Papers | 2021
|Deep Self-Supervised Learning Models for Automotive Systems
British Library Conference Proceedings | 2021
|Deep Learning-based Road Condition Monitoring System and Method
European Patent Office | 2023