Application of Algorithm of Adaptive Neural Network to Construction of Calibration Model in Quantitative Near IR Spectrometric Analysis

The feasibility of application of the algorithm of adaptive neural network,in combination with the internal model control,to the construction of calibration model in quantitative near IR-spectrometric analysis of petroleum and chemical products for their constituents was approached and discussed.In the testing,the hardware platform of dSPACE was taken as a basis,and samples of direct distilled diesel oil,hydrogenated refining diesel oil and catalytic-cracking diesel oil were taken as training samples,to test the validity of calibration model constructed by adaptive neural network.As shown by experimental results,the proposed method was proved to have quick response,small error and good robustness.In the IR spectral range of 800-2 300 nm,the values of mean square deviation of calibration samples and testing samples were found all less than 1×10-6.