차량 DTC 데이터 기반 고장 상태 예지 방안에 대한 사례 연구

Abstract

Sudden vehicle problems while driving cause great damage to the driver. In this context, it is necessary to monitor important vehicle parts’ condition and take appropriate actions in advance based on condition analysis. This paper implements a model for predicting the occurrence of a certain failure code before 24 hours based on gathered DTC (Diagnostic Trouble Code) data with LSTM (Long Short-Term Memory)-Autoencoder. LSTM is a type of RNN (Recurrent Neural Network) that can solve data long-term dependency problems and is suitable for learning many time-series data to create classification and regression models. In particular, the model is a stacked autoencoder structure consisting of several LSTMs, showing higher accuracy than normal LSTM. The case study shows that the proposed method gives a reasonable performance on predicting the failure code.