A method for transmitting radiation of the arc plasma with multimode fused quartz fiber onto the spectrograph has been studied. The plot of the Boltmann b c t i o n in emission spectral analysis is used for measuring temperature of the arc plasma.The measured temperature of the arc plasma is 5946.6K from least square linear regression of Z n ( W o ] and Ei for a number of the emission line intensities of the excited copper atom. Its regression coefficient and measured precision are -0.97% and 1.7%, respectively. The advantages of the method of the diagnostic temperature for the arc plasma are absolute measurements of the temperature, remote sensing, 244 WANG ET AL.precision and suitable for mal-environment, such as hgh temperature, toxic, explosion, strong magnetic or/and electrical fields.In addition, we have discussed the effect of the spectroscopic constants, such as transition probability, A , the statistical weight of the upper level, g , and the energy of the upper level, Ei , of copper lines on calculating temperature with a plot of the Boltzmann function in detail. The results show that the accurate measurement of the temperature for the arc plasma is obtained only when the spectroscopic constants are selected correctly.

The most serious problem in multivariate calibration analysis is that the prediction samples may contain the interferents whch are not modeled in the calibration step. Then how to detect the interferents and eliminate their influence is si&icantly important in order to obtain the correct compositional analysis results. The so-called residual spectra library search is supplied and an iterative loop regression algorithm to successfully correct the spectrum containing two interferents is developed in t h s paper. Three groups of whch each sample contains no, one or two interferents are supplied to amplify the above method. The mean relative standard deviation (RSD) for the three groups are 0.163, 0.375, 0.355%, respectively. There are no . 1452 GU ET AL.comparable difference among these RSDs, which proves the validation of the method provided in this paper. The results specify that PLS with FTIR spectroscopy is a powerful tool to resolve both the multicomponent simultaneous determination and the identification of interferents when combined with additional diagnostic and corrected procedure.

1998)Comparison of multivariate calibration methods for quantitative analysis of multicomponent mixture of air toxic organic compounds by FTIR, ABSTRACTThe quantitative prediction abilities of three competing multivariate calibration methods for concentration analysis of FTIR Spectra are compared. The calibration methods compared include classical least squares method (CLS), Kaiman filter method (KFM) and partial least squares method (PLS). The mixtures of seven air toxic organic compounds whose FTIR Spectra are known to seriously overlap were chosen to evaluate the preceding calibration methods.The concentrations of the seven air toxic organic compounds mixtures varied from 1 to 50 or 100 ppm. A relatively simple model involving the mean prediction error (MPE) and mean relative error (MRE) was developed for estimating each calibration method mentioned above.The results showed that PLS is the best calibration method among the three methods examined for a given real spectral data set while CLS and KFM had no obvious difference in the performance. Better predictable results were obtained when the measurement is taken at a series of equispaced wavenumbers of the absorption band for the desired component.

The application of Artificial Neural Networks (ANNsj for nonlinear multivariate calibration using simulated FTIR data was demonstrated in this paper. Neural networks consisting of three layers of nodes were trained by using the back-propagation iearning rule. Since parameters affect the performance of the network greatly, simulated data were used to train the network in order to get a satisfactory combination of all parameters. The mixtures of four air toxic organic compounds whose FTIR spectra are overlapped were chosen to evaluate the calibration and prediction ability of the network. The relative standard error (RSD%), the percent standard error of prediction samples (YoSEP) and the percent standard error of calibration samples (%SEC) are used for evaluating the ability of the neural network.

The multicomponent analysis with Partial Least-Squares Method (PLS) using FTIR spectroscopy has been studied in this paper. In order to ensure the reliability of the prediction results, a criterion SO and SA is created in the PLS algorithm to conclude the similarity between the prediction samples and the calibration ones. An example for the simultaneous determination of Ethylbenzene, Styrene, o-Xylene, m-Xylene and p-Xylene is supplied. The experimental designs of the GU AND WANG both the calibration and prediction samples are discussed as well as the sigtuficant factor number. With proper multivariate calibration conditions, the average mean relative error (MRE %) is 0.25 % and the average relative standard deviations (RSD%) for Ethylbenzene, Styrene, o-Xylene, m-Xylene and p-Xylene are 0.19, 0.15, 0.06, 0.44 and 0.13 %, respectively.