| Process industry intelligent manufacturing is the integration of process informationization and automation.The core of informationization is process information perception technology.It is significant and challenging for industry process optimization and control to develop the process analytical technologies(PAT)of the key properties of the feeds,intermediates and products which are very complicated in composition and difficult to measure in severe site conditions.Near infrared spectrum is an ideal process analysis signal since it can reflect the composition and structure information of the material at the molecular level and is easy to measure.At present,online NIR analyzers have been widely used for transparent and mobile liquids,and homogeneous and fine powders at moderate conditions(such as constant temperature,pressure,etc.).However,for the natural sample such as cellulose material,biological samples and epoxy resin products,many factors,which include uneven distribution of composition,strong water absorbing ability,obvious size variation in individual samples or pellets,very short exposure time for solid samples transported by high speed convey belt and the phase transition phenomenon in the reaction process,etc.,will bring significant noise in the NIR spectrum collected in industrial process.The noise is so heavy that severely quenches the useful information used for analysis.It is difficult for using the common spectral treatment methods to eliminate their effects effectively.This problem becomes a technical bottleneck hindering the use of process NIR analyzers in the fields of natural products and polymerization process.This thesis aims to study the mechanisms and laws of the influence of these factors on the spectral information and to develop the theory and techniques for on-line near-infrared analysis of solid natural materials and polymerization process.The research contents and results are as follows.1.Analysis of the key properties of natural cellulose pulp.The ?-cellulose content and degree of polymerization(DP)are key properties of natural cellulose pulp.Being time-consuming,the existing analytical methods of them can't fit the control of the production process severely.Therefore,it has great significance to study the rapid analytical method of the properties of natural cellulose pulp.In virtue of fast,simultaneously determining multi-properties,and nondestructive,NIR analytical technology is potential to provide a solution.However,natural cellulose contains so large number of hydroxyl groups that the sample is prone to absorbing water.The change of environmental humidity will seriously affect the NIR analysis results.This becomes one of the technical problems in the NIR analysis of the properties of solid natural cellulose.In this paper,continuous wavelet transform and two-dimensional correlation spectroscopy were used to study the interaction mechanism between water molecules and cellulose molecules.Based on this research,an adaptive oblique projection algorithm(AOPA)was proposed to eliminate the affect of moisture change on NIR spectra.The method of competition adaptive weighted sampling(CARS)combined with the correlation coefficient was used to select the characteristic variables.Quantitative analysis models of a-cellulose content and DP of cellulose pulp for four kinds of raw materials(coniferous wood,hardwood,bamboo,pulp of modified paper)were established.The results show that the prediction accuracy of the model based the new methods is improved one order of magnitude compared with the existing common methods.The RSDs of reduplicative prediction results of the?-cellulose content and DP models are lower than 0.22 and 1.22,respectively.The SECs of their quantitative analysis models were 0.498 and 31.627,the SEPs 0.557 and 34.146,and the RPDs 5.22 and 3.00,respectively.The research fruit lie the development of real time analytical technology of key properties of pulp a theoretical basis.2.High-speed identification and sorting of living silkworm pupae.Crossbreeding is a key process in sericulture,and this process entails sorting silkworm chrysalises into male and female pupae.To date,sorting work has been performed manually through visual inspection of pupa organs due to the lack of automatic techniques for sex sorting.The automation of sex sorting is significant to the development of modern sericulture.This study aims to use NIR spectroscopy to realize the high-speed sexing living silkworm pupae.Not only the variety,location,shape,size,feeding,region,weather and seasons,etc.of silkworm pupae,have a great influence on the NIR spectra of living silkworm pupae,but also the economic feasibility imposes strict requirements on the sorting speed(the sorting time for single pupa is in the order of milliseconds).The spectra collected on the high speed moving pupae contain heavy noise which has a strong impact on the identification of male and female pupae and challenges the high-speed sorting.In this study,the infrared spectral difference between male and female silkworm pupae was analyzed.There are differences in the FT-IR spectra between male and female pupae,indicating that NIR can be used to identify male and female pupae.The static spectra of 200 932 varieties of silkworm pupae were collected by FT-NIR and grating array NIR,respectively.Using the conventional spectral preprocessing method such as derivative,SNV combined with SIMCA to establish the recognition model,it is found that the model using the second derivative combined with SNV has the highest correct rate of spectral preprocessing,which can reach 99.6%.A dynamic spectra acquisition system consisting of a grating array NIR spectrometer and a conveyor belt device was designed.Dynamic spectra were collected for four varieties(932,7532,Autumn abundance and White jade)of silkworm pupae.The noise of the dynamic spectra is higher one order than that of static spectra.The significant noise dominated by optical path difference(OPD)has a strong impact on the identification of male and female pupae.High-speed sex discrimination of living silkworm pupae was successfully realized by combining soft independent modeling of class analogy(SIMCA)and a preprocessing method involving angle spectra.The results indicated that this method achieved a correct identification rate of 98.0%.Finally,various varieties of living silkworm pupae were sorted with respect to sex using the high-speed sorting device.The sorting rate exceeded 7-10 pupae per second and the error rate could be controlled within 2.0%.The device has been applied successfully to sort living silkworm pupae about 1.2 tons per day in different seasons and regions,and can exert a significant effect on the development of modem sericulture.3.Development of online analytical method of gel point and gel time of epoxy resin curing and study of gel mechanismGel time and gel point are key indicators for controlling the quality and the cost of epoxy resin materials.Gel point is the conversation rate at gel time.Using conventional methods,the gel time and the gel point are measured by different methods and instruments.The accuracy of gel point result by those methods is poor since the sample volume and the heat transfer process are different for the different instruments.The methods are off line and lag behind the requirements of process quality control.This study is to develop an on line analytical method of simultaneously determining the gel time and gel point by NIR spectroscopy.In this study,the evolutionary NIR spectra of the epoxy curing process were collected.The spectral signal was decomposed by Discrete Wavelet according to frequency.The different component signals have different frequencies and the signals of chemical groups and light scattering-related are used to determine the conversion rate and gel time,respectively.The accuracy of the gel point result determined by the new method is much better than that of the existing methods.The online analytical method of simultaneously determining gel time and gel point were achieved successfully by NIR spectroscopy.In addition,the MCR-ALS method was applied to the evolutionary NIR spectra to analyze the gelation mechanism of epoxy resin,revealing the effect of temperature on the network growth during gelation of epoxy resin.The research results are as follow.At lower temperatures,a longer linear structure is formed.As the length of the linear chain increases and the chain motion is limited,a branched structure is formed,and finally a three-dimensional network structure is formed to a gel.When the temperature is moderate,the branched structure has more linear chains.Finally,the branched structures are connected to each other to form a large network structure.When the temperature is high,the reaction forms a lot of small three-dimensional network structure at initial,the numerous little network structures are interconnected finally to form a large three-dimensional network structure. |
PDF Full Text Request