A Piezoresistive Pressure Sensor Modeling by Artificial Neural Networks
Behzad Hadi
1
(
Mechanical Engineering Department, Faculty of Engineering, Razi University, Kermanshah, Iran
)
Farzin Shama
2
(
Department of Electrical Engineering, Kermanshah Branch, Islamic Azad University,
Kermanshah, Iran
)
Hamid Sherafat Vaziri
3
(
Electrical Engineering Department, Kermanshah Branch, Islamic Azad University, Kermanshah, Iran
)
Mohsen Eghbalkhah
4
(
Electrical Engineering Department, Faculty of Engineering, Razi University, Kermanshah, Iran
)
کلید واژه: ANN, Modelling, Piezoresistive sensor,
چکیده مقاله :
An efficient artificial neural network (ANN) model for a pressure sensor is presented in this paper. This pressure sensor has a piezoresistive structure and its experimental data is used in this work to train and test the proposed ANN model. The selected network is a multi-layer perceptions ANN type, which has one hidden layer with five neurons inside. Each neuron uses an activation function of tangent sigmoid. The proposed MLP modelled system has been trained to recognize the output of the pressure sensor, which is a generated voltage in millivolts, according to the temperature and pressure levels. The designed model has been tested using 30 percent of the data, which have been separated before for validation. The obtained mean square relative error (MSRE) error of this method is only 0.137 for test data set that marks good and strong accord between the proposed MLP modelled network and experimental data set.
چکیده انگلیسی :
An efficient artificial neural network (ANN) model for a pressure sensor is presented in this paper. This pressure sensor has a piezoresistive structure and its experimental data is used in this work to train and test the proposed ANN model. The selected network is a multi-layer perceptions ANN type, which has one hidden layer with five neurons inside. Each neuron uses an activation function of tangent sigmoid. The proposed MLP modelled system has been trained to recognize the output of the pressure sensor, which is a generated voltage in millivolts, according to the temperature and pressure levels. The designed model has been tested using 30 percent of the data, which have been separated before for validation. The obtained mean square relative error (MSRE) error of this method is only 0.137 for test data set that marks good and strong accord between the proposed MLP modelled network and experimental data set.
