The Method Of Spectral Data Analysis And Platform Realization

With the progress of society and the improvement of industrial level,the industrial production process has higher standards about the quality control.Meanwhile,it's vital to obtain the relevant parameters of production process information in order to ensure the quality.Based on the relevant technology of spectral analysis,first of all,it can have good application in many branches in the industrial process,secondly,it can use the spectral analysis to achieve quantitative testing of production process parameters,and lastly,it can achieve quality control process qualitative discrimination.Spectral analysis technology has many advantages,such as lossless,fast,and can be online,etc.At the same time,there are some disadvantages,such as weak signal strength and signal overlap,which lead to the degradation of accuracy and the limitation of application.As an indirect analysis technique,the core issue,on whether the method can be used in industrial processes or not,tremendously depends on the establish of an effective analytical model.Therefore,the quality of the spectrum analysis technology is inseparable from the modeling method.The data pre-processing method,the sample set selection method,the wavelength optimization method,the modeling method and the model transfer method are all core content in the field of spectrum analysis.In order to achieve good application in actual industrial production process,the enrichment and improvement of spectrum analysis technology method theory is extremely needed.Based on this background,this paper has completed the related work:(1)This paper mainly introduces the background and significance of the research on spectral analysis technology,expatiates the principle of near-infrared spectroscopy as the main spectrum analysis technology,briefly introduces the technical route of on-line detection based on the spectrum,and sketches the current research status and development trend on the spectrum analysis in domestic and abroad.(2)The research and improvement of data pre-processing,spectral wavelength selection,qualitative analysis,quantitative analysis and so on applied to the spectrum analysis technology are described in detail.Under the foundation on the technical research of algorithm,the software based on C#language and hardware platform is designed and implemented.(3)Based on the analysis of the fiber in the papermaking process,the contents of the components of water,methyl fiber,lignin and pentose were quantitatively analyzed on the basis of the spectrum analysis platform.The first-order differential calibration,standard normal variable transformation,multiple scattering calibration with first order differential in addition and other pre-treatment methods.Under the circumstance of ulti-scattering calibration plus first order differential for data pre-processing,a variety of methods,such as partial least squares regression analysis,neural network,is used as a comparison,then the analytical model established by the partial least squares regression has higher accuracy and reliability.(4)Using animal fiber wool and cashmere as a study object,however,there are plenty of similarities in the organizational structure that causes identification difficulties.In this part,we study the spectral pre-processing method and the wavelength selection method,and establish a variety of discriminant models.Finally,the correctness of competitive adaptive weighting sampling combined with PLS-DA method is 100%.