Improvement of accuracy under dynamic conditions for NO_x sensors using artificial neural network

ABSTRACT

The aim of this study is to eliminate the impact of inconsistency in probe production through artificial neural network model, to reduce the defect rate in the production process of NO_x sensors. Different from existing research, the voltage and current signals inside the sensor probe were the inputs of the artificial neural network model. The parameters of the artificial neural network model were optimized through the data obtained under transient engine operation conditions, so as to ensure that the model works in real time and with high accuracy within the NO_x sensor controller. It was found that the current of the main pump and the current of the measuring pump had high impact on the NO_x signal, with a contribution rate of more than 80%. The voltage of the auxiliary pump in the sensor probe had the lowest influence on the NO_x signal, with a contribution rate of less than 2.36%. Experimentally, the results showed that the artificial neural network model can eliminate the adverse effects caused by hardware differences of the NO_x probes. The accuracy of the probes was improved from 0.7048 to 0.9748. An increase of 38.3% on average in dynamic conditions.

KEYWORDS:

• Diesel
• NO_x sensor
• artificial neural networks
• consistency of probes
• dynamic response

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the National Key Research Development Program of China(2017YFB0103402)