| This paper studies the near infrared spectral analysis technology in fermentation product quality detection application.Application of PLS and RBFNN methods were established for the determination of Irpex lacteus mycelial polysaccharide content in NIR spectroscopy quantitative analysis model, using Irpex lacteus different mutant strains in the fermented mycelia of sample data, the samples are broadly representative, can meet the demand model. The present study are discussed the process of modeling, identification of abnormal samples using Monte Carlo partial least squares method, separately inspects convolution smoothing method, fast Fourier transform, a derivative method and derivation of two order number method for modeling the influence, was also studied the pretreatment method of pretreatment effect of window size, examines the partial least squares for modeling the effects of wavelength, screening variables using the movable window partial least squares method, the approximation degree as an index to choose the most suitable PLS model hidden variables, examines the radial basis function neural networkcalibration model construction method, established the RBFNN model to choose the suitable model of input nodes, number of hidden layer nodes and RBF width, finally, both undertake comparative choosing better model.Using the optimum models for predicting the contents of polysaccharide in the samples, the determination coefficient of calibration set (Rc) was 0.9490, which indicating that the fit of the model was fine. The root mean square error of calibration set (RMSEC) was 0.0067, the root mean square error of prediction set (RMSEP) was 0.0070, which indicating that the predictive ability of the model was satisfied. It can be used for quantitative analysis. This method can be used for the high throughput screening of mutant strains and the optimization of Irpex lacteus Fr. fermentation. It also can be extended to monitor the process of Irpex lacteus Fr. production. |
PDF Full Text Request