Vol. 10, No 1, 2012 pp. 1 - 12
DOI 10.2298/FUPCT1201001L
UDC 007.52 + 53.082.4 : 621.375.826
COMPUTATIONAL INTELLIGENCE BASED SIMULTANEOUS DETERMINATION OF THE SPATIAL PROFILE OF THE LASER BEAM AND VIBRATIONAL-TO-TRANSLATIONAL RELAXATION TIME BY PULSED PHOTOACOUSTICS
Mladena Lukić1*, Žarko Ćojbašić2, Mihailo D. Rabasović3, Dragan D. Markushev3, Dragan M. Todorović4
1Faculty of Occupational Safety, University of Niš, Čarnojevića 10a, 18000 Niš, Serbia
2Mechanical Engineering Faculty, University of Niš, Aleksandra Medvedeva 14, 18000 Niš, Serbia
3Institute of Physics, University of Belgrade, Pregrevica 118, 11080 Beograd-Zemun, Serbia
4Institute for Multidisciplinary Research, University of Belgrade, P.O. Box 33, 11030 Belgrade, Serbia
E-mail: mladena.lukic@znrfak.ni.ac.rs
Abstract.This paper is concerned with the possibilities of computational intelligence application for simultaneous determination of the laser beam spatial profile and vibrational-to-translational relaxation time of the polyatomic molecules in gases by pulsed photoacoustics. Results regarding the application of neural computing and genetic optimization are presented through the use of feed forward multilayer perception networks and real-coded genetic algorithms. Feed forward multilayer perception networks are trained in an offline batch training regime to estimate simultaneously, and in real-time, laser beam spatial profile R(r) (profile shape class) and vibrational-to-translational relaxation time VT from a given (theoretical) photoacoustic signals p(r,t). The proposed method significantly shortens the time required for the simultaneous determination of the laser beam spatial profile and relaxation time and has the advantage of accurately calculating the aforementioned quantities. Real coded genetic algorithms are used to calculate VT by fitting the p(r,t) with the theoretical one. The previously developed methods determine the laser beam profile and relaxation time with sufficient precision, but the methods based on the application of artificial intelligence are more suitable for practical applications, such as the real-time in-situ measurements of atmospheric pollutants.
Key words: photoacoustic spectroscopy, laser beam profile, vibrational-to-translational relaxation time, neural networks, genetic algorithms
PRIMENA VEŠTAČKE INTELIGENCIJE ZA ISTOVREMENO ODREĐIVANJE PROSTORNOG PROFILA LASERSKOG SNOPA I VIBRACIONO-TRANSLACIONOG RELAKSACIONOG VREMENA IMPULSNOM FOTOAKUSTIKOM
U ovom radu je istražena primena veštačke inteligencije na istovremeno određivanje prostornog profila laserskog zraka i vibraciono-translacionog relaksacionog vremena impulsnom fotoakustikom. Naime, rezultati dobijeni korišćenjem neuronskih mreža i genetske optimizacije su predstavljeni kroz primenu višeslojnih neuronskih mreža i realno-kodiranih genetskih algoritama. Višeslojna neuronska mreža sa prostiranjem signala unapred obučavana je u režimu nadgledanog obučavanja da istovremeno u realnom vremenu da procenu prostornog profila laserskog snopa R(r) (klasu oblika profila) i vibraciono-translacionog relaksacionog vremena VT iz datog fotoakustičkog signala. Realno kodirani genetski algoritmi (GA) su korišćeni za izračunavanje VT fitovanjem eksperimentalnog signala teorijskim. Cilj je nalaženje kombinacije parametara fotoakustičkog signala, poluprečnika laserskog snopa rL i VT, koji daju najbolje poklapanje sa datim fotoakustickim signalom. Predloženi metod značajno skraćuje vreme potrebno za istovremeno određivanje prostornog profila laserskog snopa i vremena relaksacije a ima i prednosti u tačnijem izracunavanju navedenih veličina. Iako do sada razvijeni metodi određuju laserski profil i relaksaciono vreme dovoljno precizno, metod zasnovan na primeni veštačke inteligencije je mnogo pogodniji za praktične primene, kao što je detekcija atmosferskih zagađivača u realnom vremenu.
Ključne reči: fotoakustička spektroskopija, profil laserskog snopa, vibraciono-translaciono relaksaciono vreme, neuronske mreže, genetski algoritmi