Time Series Prediction of E-nose Sensor Drift Based on Deep Recurrent Neural Network

Chemical sensors drift is a slow change in sensitivity that occurs in time which makes it difficult to construct an appropriate sensor drift treatment. The main purpose of this paper is to study a new methodology for time series prediction of chemical sensors drift based on LSTM (long short-term memory, LSTM) recurrent neural network, including data preprocessing and partition, network and training, network prediction. This technique can mine the deep information of sensors drift signals instead of manual extraction and match the complex nonlinearity much more exactly. It is proved that the proposed LSTM prediction model can make long-term, accurate prediction of chemical sensors baseline and drift.