Application of neural network to study of frequency range effect to photoacoustic measurement of thermoelastic properties of thin aluminum samples

• Published in: Measurement : journal of the International Measurement Confederation 2024-08, Vol.236, p.115043, Article 115043
• Main Authors: Djordjević, K.Lj, Stoisavljević, Z.Z., Dragaš, M.A., Stanimirović, I., Stanimirović, Z., Suljovrujic, E., Galović, S.P.
• Format: Article
• Language: English
• ISSN: 0263-2241; 1873-412X
• DOI: 10.1016/j.measurement.2024.115043

•Frequency range length effect on PA characterization of the sample is analyzed.•Three frequency ranges (20 Hz-20 kHz, 20 Hz-7 kHz, 20 Hz-3 kHz) are experimentally observed.•Measurement of the properties of thin Al plate is done by developing neural networks.•A separate neural network is formed for each frequency range and results are compared.•It is shown that the higher the frequency range the greater the error estimation. In this paper, the influence of the measurement range on the prediction of thermoelastic and geometric properties of a thin aluminum sample obtained by solving the inverse photoacoustic problem using neural networks was analyzed. Tests of neural networks on theoretical photoacoustic signals showed that the accuracy of solving the inverse problem increases with the increase of the measurement range, which is in accordance with the theory of solving inverse problems. Tests on experimentally measured signals in a transmission experimental setup with an open photoacoustic cell, from which the influence of the transmission characteristic of the detector was removed, show the advantages of using narrow frequency bands. This result indicates the need for further investigations of both: the influence of the measuring chain on the recorded signal and the influence of additional processes and sample properties on the thermoelastic behavior of thin metal films.