Correction of Dynamic Errors of a Gas Sensor Based on a Parametric Method and a Neural Network Technique

Abstract

The paper presents two methods of dynamic error correction applied to transducers used for the measurement of gas concentration. One of them is based on a parametric model of the transducer dynamics, and the second one uses the artificial neural network (ANN) technique. This article describes research of the dynamic properties of the gas concentration measuring transducer with a typical sensor based on tin dioxide. Its response time is about 8 min, which may be not acceptable in many applications. On the basis of these studies, a parametric model of the transducer dynamics and an adequate correction algorithm has been developed. The results obtained in the research of the transducer were also used for learning and testing ANN, which were implemented in the dynamic correction task. Despite the simplicity of the used models, both methods allowed a significant reduction of the transducer’s response time. For the algorithm based on the parametric model the response time was shorter by approximately eight-fold (reduced up to 40–80 s, i.e., about 2–4 sample periods), whereas with the use of an ANN the output signal was practically fixed after a time equal to one sampling period, i.e., 20 s. In addition, the use of ANN has allowed reducing the impact of the transducer dynamic non-linearity on the correction effectiveness.

Correction of Dynamical Properties of Data Acquisition Systems

“…The extensive literature search resulted in some publications correlated with the subject of this paper. The correction of dynamical properties of data acquisition systems and sensors is obtained there by inverse modelling [ 2 , 3 , 4 , 5 , 6 , 7 , 8 ], or inverse modelling aided by a feedback control system [ 9 ], or one-step forward specialised prediction [ 10 , 11 , 12 ], or joint input and state estimation based on Kalman filtering [ 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 ]. All these correction methods use linear discrete-time dynamic models of data acquisition systems.…”

Projeto e Experimentação de um Filtro Acelerador para um Sensor Dinâmico de Temperatura Sujeito a Parâmetros Incertos

“…Dentre os trabalhos que procuraram melhorar o desempenho dinâmico de sensores tem-se: a comparação de um método baseado em modelo discreto com outro, usando redes neurais, para a correção de erros em sensores de gases foi feita por Roj (2016); Güther et al (2013) aplicaram os dados lidos de um sensor de temperatura veicular a um filtro de Kalman, o qual realizou uma predição de estado. No entanto, o filtro foi implementado em Matlab R o que, para muitas aplicações, nãoé prático.…”

Precision Improvement for the Detection of TGC via RBF Network

“…BP (back propagation) neural network model has been brought into infrared temperature and humidity compensation [3,4]. RBF (Radial Basis Function) neural network is applied to precision motion system [5] and neural network is used in the pressure analysis [6] and gas concentration measurement [7] in industrial environment; and a new method of Correction of Dynamic Errors of a Gas Sensor Based on Neural Network has been presented [8], etc.…”