Font Size: a A A

Application Of Wavelet Transform And Artificial Neural Network In The Fluorescence Temperature-measurement Signal Processing

Posted on:2016-02-22Degree:MasterType:Thesis
Country:ChinaCandidate:Y X ShaoFull Text:PDF
GTID:2308330461489339Subject:Optical Engineering
Abstract/Summary:PDF Full Text Request
With the temperature measurement becoming more and more important in scientific research and industry control process, the scope of application of temperature measuring instrument, measuring accuracy requirements are also getting higher and higher, so the fluorescence temperature measurement technology emerge as the times require, quickly get the favour of people and made great development.Compared with other types of optical fiber temperature sensor, fluorescence optical fiber temperature sensor has many advantages, which can avoid the cross sensitivity of fiber loss, environment, radiation, emission wave bandwidth caused by factors such as temperature measurement accuracy, but also has good stability, high reliability,long life characteristics, simple production, low cost.This paper firstly expounds the mechanism of the fluorescence temperature measurement, choose Ca2 Mg Si2O7:Au+as the fluorescent material, build the fluorescence optical fiber temperature measuring system based on fluorescence lifetime. In the fluorescence lifetime measurement, because temperature only a direct relationship with the fluorescence lifetime, therefore only need to acquire the fluorescence lifetime can be, and the noise is the key problem of calculating impact analysis of fluorescence lifetime.This paper proposes an improved wavelet threshold denoising, not only to retain the advantages of the traditional soft and hard threshold function, but also can deficiencies improvement of the two, so as to achieve the optimal signal noise separation effect; through MATLAB simulation analysis, contrast the denoised signal to noise ratio, this paper expounds the importance of the wavelet basis, number of layers, the selection of threshold the.In the actual measurement, the fluorescence signal after denoising exhibits non exponential form, the need to establish the model obtained by data fitting of fluorescence lifetime. By fitting the data to the disadvantage of the traditional method of comparison, this paper uses artificial neural network to fit, and the use of wavelet functions instead of sigmod activation function, combined with the improved BP algorithm and genetic algorithm for network learning, not only improves the fitting precision and convergence speed, and can avoid the local large error.The new threshold wavelet de-noising application in fluorescence signals, wavelet,threshold, screened layers right, go to the noised signal through MATLAB simulation comparison, we can see that the signal can avoid the emergence of shock, reduce the loss of useful information, become more smooth, and improve the signal to noise ratio;fluorescence the service life of several fitting methods of comparison, illustrates the advantages of artificial neural network, the wavelet neural network fitting data, notonly improves the fitting accuracy of the signal curve, decreases the fluorescence lifetime measurement error, but also improve the accuracy of temperature measurement. The experimental results show the effectiveness of the proposed method.
Keywords/Search Tags:fluorescence lifetime, wavelet transform, denoise, wavelet neural network, data fitting
PDF Full Text Request
Related items