Research Of Multi-component Dye Solution Concentration Detect Device Based On SVM And Spectrophotometric Method

With the improvement of the living standards, people put forward higher requirements onthe quality of textiles. As we all know, the color of textiles is an important factor of the textilequality in the production. Actually, The dye concentration is a key factor affecting the color oftextiles in the dying process, and the accurate detection of the dye solution concentratio n isimportant to ensure the color of textiles. However, the dye solution concentration detectiondevice is rarely applied in the current industrial production, especially in the detection ofmulti-component dye. To solve this problem,a multi-component dye solution concentrationdetect device is designed based on spectrophotometric method and support vector machine(SVM).Considering the miniaturization design requirements for most industrial monitoringsystems presently, the monochromatic optical system o f this paper is designed with theCzerny-Turner crossover optical structure, and the design has many advantages, such ascompact structure, small space size, easy installation and less effect of the stray light, etc.To improve the ability of full spectrum measurement for the ordinary photodiodes sensors,this paper select detector sensor with TCD1708D, which is a linear multi-channel CCD detectordevice. With advantages of small space size, high resolution, high sensitivity, wide dynamicrange,fast response, etc., TCD1708D can achieve the full range of visible spectrummeasurement. In this paper, the system is designed with modular electronic structure: CCD isdriven by CPLD device EPM240T100C; the control part of CCD signal acquisition andprocessing is based on ARM7kernel processor S3C2440.In fact, the industrial production often through mix a variety of dyes to obtain satisfactorycolor. However, the absorption spectra between the mixed dye components have a overlap orinterference problem, and that means using the Lambert-beer law directly is not suitable toanalysis dye solution concentration. To solve absorption spectra overlap or interference seriousproblem in the mixed dye system, a dye solution concentration detection model is established based on support vector machine regression. Compared to other chemical analysis models,support vector machine regression algorithm has better performance for the sample data of highdimension and small sample size. In this paper, the spectral calibration method is is adoptedto establish the correspongding relation of the spectral data between the serial number, thenanalysis support vector machine regression prediction model parameters and select theappropriate model parameters to establish a multi-component dye concentration detectionalgorithm.Based on spectrophotometric method and support vector machine regression method,wedevelop a set of multi-component dye solution concentration detection device,and predicted thedye solution concentration successfully. According to the experimental results, the predictionerror of dye solution concentration is within3%.