NEURAL NETWORK BASED MODELING OF NOX DETECTION WITH A SENSOR – POLYMER SYSTEM

  • CIPRIAN PIULEAC
    “GH. ASACHI” TECHNICAL UNIVERSITY, DEPARTMENT OF CHEMICAL ENGINEERING, BD. D. MANGERON NO. 71A, 700050, IAŞI, ROMANIA
  • SILVIA CURTEANU
    “GH. ASACHI” TECHNICAL UNIVERSITY, DEPARTMENT OF CHEMICAL ENGINEERING, BD. D. MANGERON NO. 71A, 700050, IAŞI, ROMANIA
  • GABRIELA TELIPAN
    RESEARCH AND DEVELOPMENT NATIONAL INSTITUTE FOR ELECTRICAL ENGINEERING ICPE CA, SPL. UNIRII 313, SECT. 3, BUCHAREST, ROMANIA
  • MARIA CAZACU

Abstract

Some types of synthetic polymers have been tested in nitric oxide sensing. Feed-forward neural networks with different types and topologies are used in mathematical modeling of the system, to predict the voltage of the sensor as function of polymer type, gas concentration and time. In this way, the efficiency of the sensor can be appreciated. The relative errors, extremely low, obtaining in the validation phase, prove the validity of the neural models. The optimization problem performed by inverse neural network modeling answers the question what are the initial conditions that lead to an imposed value for output signal of the sensor.

Cuvinte cheie

neural networks mathematical modeling polymer-based sensors direct and inverse modeling nitric oxide