• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    An integrated spectroscopic strategy to trace the geographical origins of emblic medicines:Application for the quality assessment of natural medicines

    2020-09-04 09:33:20LumingQiFurongZhongYangChenShengnanMaoZhuyunYanYuntongMa
    Journal of Pharmaceutical Analysis 2020年4期
    關(guān)鍵詞:花生殼碳纖維理論

    Luming Qi,Furong Zhong,Yang Chen,Shengnan Mao,Zhuyun Yan,*,Yuntong Ma,**

    aState Key Laboratory of Characteristic Chinese Medicine Resources in Southwest China,Chengdu University of Traditional Chinese Medicine,Chengdu,611137,China

    bSchool of Pharmacy,Chengdu University of Traditional Chinese Medicine,Chengdu,611137,China

    Keywords:

    Emblic medicine

    Quality assessment

    Geographical traceability

    Spectroscopic analysis process

    ABSTRACT

    Emblic medicine is a popular natural source in the world due to its outstanding healthcare and therapeutic functions.Our preliminary results indicated that the quality of emblic medicines might have an apparent regional variation.A rapid and effective geographical traceability system has not been designed yet.To trace the geographical origins so that their quality can be controlled,an integrated spectroscopic strategy including spectral pretreatment,outlier diagnosis,feature selection,data fusion,and machine learning algorithm was proposed.A featured data matrix(245×220)was successfully generated,and a carefully adjusted RF machine learning algorithm was utilized to develop the geographical traceability model.The results demonstrate that the proposed strategy is effective and can be generalized.Sensitivity(SEN),specificity(SPE)and accuracy(ACC)of 97.65% ,99.85% and 97.63% for the calibrated set,as well as 100.00% predictive efficiency,were obtained using this spectroscopic analysis strategy.Our study has created an integrated analysis process for multiple spectral data,which can achieve a rapid,nondestructive and green quality detection for emblic medicines originating from seventeen geographical origins.

    1.Introduction

    The development of a generalized geographical traceability system for natural medicines remains a significant challenge because the growing environment always has a noticeable influence on their quality[1].This interference is multidimensional and unpredictable.Primary and secondary metabolite compounds,which are mainly responsible for the healthcare and therapeutic functions of natural medicines,always vary significantly because of their different geographical origins[2,3].Effective analytical methods and instruments for obtaining more insights into the metabolic characterizations and regional variation of natural medicines are essential because these variations affect both producers and consumers.The well-identified to geographical origin for a natural product is a prerequisite to its optimal application.

    At present,many strategies such as molecular,chromatographic and spectroscopic methods have been applied to identify the origins of natural products based on their respective advantages[4—8].Especially,spectroscopic analyticalinstruments have attracted more and more attention to characterizing natural products originating from different geographical origins.These techniques are worth to be recommending because they are rapid,simple and environment-friendly.These advantages can further promote the efficiency and safety of the quality control process for natural medicines.However,natural products are always an especially complex mixture with diversified metabolic ingredients.The descriptive information generated from different spectroscopic sensors is always sizable,so there are still a large number of irrelevant and redundant attributes.A large number of data optimization algorithms have been developed to enhance the availability of spectral data.

    For example,feature selection is one of these requisite algorithms regarding a geographical traceability task.It can produce a clean and informative sub-dataset,which is necessary to improve the accuracy of analysis and to decrease the computation cost.Generally,feature selection algorithms can be classified into three types of filter,wrapper and embedded models,which have different efficiency for feature selection with respective criteria[9].Another effective strategy for measuring geographical traceability is data fusion.It is utilized to integrate multi-source descriptive information when two or more instruments are implemented simultaneously[10].It can provide a complementary approach to constructing a more effective geographical traceability model regarding the regional variation of natural products.These data optimization algorithms further enlarge the application of spectroscopic techniques.

    Generally,a complete spectral analysis process for a geographical traceability model should contain several key steps,including spectral pretreatment,outlier diagnosis,feature selection,and machine learning algorithm.Each step needs to be strictly optimized.So far,many spectroscopic geographical traceability studies of natural medicines have been conducted[11—16].A limited number of studies can use a complete spectroscopic analysis process for geographical traceability,and the universality of developed models is insufficient especially for the quality assessment of natural medicines.

    The fruit of emblic(Phyllanthus Emblica L.)belonging to Euphorbiaceae is a popular natural medicine for treating cough and indigestion in China.It has been recorded in“Chinese Pharmacopoeia”.The World Health Organization has designated this species as a plant worth of extensive cultivation in the world because of its outstanding healthcare and medicinal functions.Phytochemical and pharmacological researchers have demonstrated that this product has a broad range of metabolic ingredients such as phenolic,flavonoids and terpenoids.These compounds are capable of producing many biological benefits such as antidiabetic,antioxidant,anticancer,and other additional benefits[17—21].This fruit,especially rich in vitamin C,has greater than 100 times the vitamin C contained in an apple.To the best of our knowledge,it is extensively distributed in many countries,including China,India and the American Continent.There is a considerable variance in the quality of embolic medicines that comes from these different regions.A rapid and effective spectroscopic quality aseessment strategy concerning different geographical origins is still lacking.Such a quality control strategy is required for the consistent supply of top-quality original materials.

    Fig.2.The visualization of regional variation of emblic products constructed by the levels of six metabolites in these medicines.

    Table 1 The levels of active compositions of emblic materials from different geographical origins.

    With these ideas in mind,this study aimed to design a rapid and effective spectroscopic geographical traceability model for natural emblic medicines.Our research team collected different emblic materials(cultivated and wild)from seventeen geographical origins in six provinces of China during 2017.The main bioactive compounds(gallic acid,corilagin,chebulagic acid,ellagic acid,quercetin,and vitamin C)were first determined using a highperformance liquid chromatography-ultraviolet detection(HPLCUV)method.These ingredients largely determine the healthcare and medicinal properties of these materials so that the result can revealthequalityvariationsofthem abouttheir different geographical origins.An integrated spectroscopic analysis process was proposed using two high-throughput spectroscopic techniques of Fourier transform near-infrared(FT-NIR)and Fourier transform mid-infrared(FT-MIR).This work flow included spectral pretreatment,outlier diagnosis,feature selection,data fusion,and machine learning algorithm.Especially,twelve feature selection models including filter,wrapper and embedded were applied to collect informative spectral variables comparatively.Data fusion theory was further used to combine the information learned from two spectroscopic techniques.We hope this study can provide a universal geographical traceability strategy for emblic medicines and also promote the application of spectroscopic techniques for the quality assessment of multi-source natural medication.

    2.Materials and methods

    2.1.Reagents

    Methanol(chromatographicgrade)waspurchased from Thermo Fisher Scientific(Shanghai,China).Deionized water used for chromatographic analysis was produced using an ultrapure water system(Millipore,USA).Chemical standards of gallic acid,corilagin,chebulagic acid,ellagic acid,quercetin,and vitamin C were provided by Chroma-Biotechnology Co.,Ltd.(Chengdu,China).Other analytical grade reagents were supplied by Chron Chemicals Co.,Ltd.(Chengdu,China).

    2.2.Sample preparation

    The detailed information of collected emblic materials from seventeen geographical origins in six provinces of Chinais shownin Table S1.Their fresh fruits and medicinal materials are shown in Fig.S1.The fruits of these plants were collected from September to December 2017.After removing the dirt from the surface,these samples were put into a drying oven for 24 h at 60°C.Then they were labelled according to their geographical origins and smashed using a powder machine.Power filtered with an 80 mesh sieve was used for final chromatographic and spectral analysis.Professor Yuntong Ma of Chengdu University of Traditional Chinese Medicine authenticated all the plants of P.emblica.

    Fig.3.FT-NIR and FT-MIR spectra after spectral pretreatment optimization.

    2.3.Chromatographic and spectral analysis

    A Shimadzu system(Shimadzu,Japan)equipped with an LC-20AT quaternary pump,a SIL-20A XR autosampler,a CTO-20AC column oven,and an SPD-20A UV/Vis detector was utilized to determine bioactive compounds of emblic fruits.An Agilent ZORBAX Eclipse XDB-C18(4.6 mm × 250 mm,5μm)column was applied to separate objective compounds.

    For the determination of gallic acid,corilagin,chebulagic acid,ellagic acid and quercetin,each sample of 0.100 g was first weighed.The powder was ultrasonically extracted in 10 mL methanol solution for 60 min.Other HPLC-UV conditions arelistedbelow:column temperature:30°C;mobile phase:methanol(A)and 0.1% phosphoric acid(B); flow rate:1 mL/min;elution gradient:0—15 min,5% A;15—35 min,5% —37% A;35—39 min,37% —47% A;39—60 min,47% —60% A;injection volume:5μL;detection wavelength:273 nm.

    For the determination of vitamin C,each sample of 0.100 g was exactly weighed and then ultrasonically extracted in 10 mL of 0.5% oxalic acid for 30 min.Other HPLC-UV conditions are as follows:column temperature:30°C;mobile phase:0.1% phosphoric acid;flow rate:1 mL/min;isocratic elution:15 min;injection volume:10μL;detection wavelength:254 nm.All test solutions were filtered using a 0.45μm membrane before HPLC-UV analysis.

    Two spectroscopic sensors of FT-NIR and FT-MIR spectrometers(PerkinElmer,USA)were used to directly record the spectral signals of sample powder without an extraction pretreatment.Their scan ranges were set as 10000—4000 and 4000—500 cm-1,respectively.The accumulated scans and resolution of two sensors were defined as 64 and 4 cm-1,respectively.Before the sample introduction,a blank control was scanned in order to remove any air interference.

    For each sample,approximately 0.5 g powder was weighed using an electronic balance(Sartorius,Germany)and put into a sample cell of FT-NIR and FT-MIR instruments.For FT-MIR,an additional attenuated total reflection accessory was connected to enable sample powder to be directly detected without complicated preparation.Each spectrum was scanned in triplicate,and the average spectrum was used for final analysis.

    3.Geographical traceability strategy

    3.1.Spectral pretreatments

    The spectral quality is susceptible to environmental factors.Many interference factors,including baseline drift and light scattering,decrease the analytic accuracy.Several pretreatments were conducted to optimize spectral data.Baseline correction was applied to produce a stable spectral baseline,and a smoothing algorithm(15 points)was used to remove the tinysignals which were useless for the next analysis.Multiplicative scatter correction eliminated the effect of light scattering caused by the particle size of powder[22].

    Fig.4.Comparison of twelve feature selection models based on the KAP coefficient displaying the different efficiency of various feature selection models.

    Fig.5.The first 100 important variables of FT-NIR and FT-MIR spectra,respectively,based on different models.

    3.2.Outlier diagnosis

    Anomaly samples can negatively impact model accuracy of geographical traceability.Therefore,two outlier detection tools were jointly used to ensure that the analyzed samples were free from abnormal points.

    The first such tool can be regarded as a conventional clustering method based on Hotelling's T2distribution[23].Based on the principal component analysis,Hotelling's T2displays a confidence ellipse 95% confidence limit.Samples outside of this ellipse were generally regarded as outliers in our study.

    The second method used to accomplish this goal was also an unsupervised algorithm called isolation forests(iForest),which is a state-of-the-art technique for handling high-dimensional data[24].It is an ensemble method utilized to combine many isolation trees.In brief,this method randomly selectedψpoints as sub-sampling size.For each tree,these partition points were recursively partitioned by randomly chosen attributes.The process was complete when all the samples were divided into single isolated subspaces.The average path length over selected trees was then recorded as iForest score for each sample.A sample with a low iForest score was classified as an outlier.In this study,two parameters of subsampling sizeψand tree number were set as 256 and 100,respectively.

    3.3.Feature selection

    Different types of feature selection techniques always perform various levels of efficiency for simplifying spectral data of natural medicines.Three feature selection theories(filter,wrapper,and embedded models)were utilized in our study to pick out the informative spectral variables according to their importance comparatively.

    Filter models evaluate each variable according to their criteria instead of a specific machine learning classifier.Two unsupervised feature selection techniques of Laplacian Score(LS)[25]and Unsupervised Multi-Cluster Feature Selection(U-MCFS)[26]were first applied.The other two were supervised feature selection techniques which were called Supervised Multi-Cluster Feature Selection(S-MCFS)[26]and Infinite Latent Feature Selection(ILFS)[27].

    Embedded models combine the superiorities of filter and wrapper models.This type of model always performs a high ef ficiency for feature selection.Least Absolute Shrinkage and Selection Operator(LASSO)[32]and Variable Importance in Projection(VIP)[33]were used as two linear embedded models because they were embedded into linear classifies.Additionally,Permutation importance(PIMP)[34]and Gini coefficient(Gini)[35]based on decision tree theory were also used to propose the best one.

    3.4.Evaluation of feature selection model

    The evaluation of feature selection models was an essential step in selecting the most useful spectral information to reflect the regional variation of emblic medicines.A repeated 10 fold crossvalidation procedure[36](three times)was used to evaluate the performance of each feature selection model.Because superabundant variables always enlarge the size of search space and lead toan over fitting model,we only used the first 400 variables according to their score ranking.These variables were circularly evaluated with an interval of 10.The best feature selection models were confirmed according to the accuracy of cross-validation regarding FT-NIR and FT-MIR datasets,respectively.

    Fig.6.mtryoptimization process for an RF model according to the lowest OOB estimate.

    Table 2 The performance of the geographical traceability model of emblic fruits.

    3.5.Data fusion and RF model

    Data fusion was conducted on the feature level,because the feature selection models have been performed to select the informative variables from two spectral datasets,respectively.Based on data fusion theory,a combined data matrix related to regional variation of emblic medicines was generated.

    RF algorithm is an ensemble learning algorithm combined by a certain number of tree classifiers(ntree),which are mutually independent of each other.It also has excellent performance against over fitting and noise resistance because the training process is random.First,the bootstrap sampling method is used to select a random number of samples for each tree classifier.Besides,a random subspace of variables(mtry)of each sample is applied for each tree classifier.The results of all tree classifiers are exported,and a majority vote is performed for a final decision.ntreeand mtryare determined according to the out-of-bag(OOB)estimate ntreein advance[37].

    Four parameters of kappa(KAP),accuracy(ACC),sensitivity(SEN)and specificity(SPE)were together used for a balanced evaluation of our geographical traceability model.Overall,the model has several primary advantages over previous studies:(1)interferential and redundant signals were removed as much as possible;(2)multi-source descriptive data were well-utilized;(3)our proposed strategy is complete and can be effectively generalized.A simple data flow diagram for feature selection,data fusion,and the RF model is shown in Fig.1.

    4.Results and discussion

    4.1.Quality variation of emblic materials

    Primary and secondary metabolites are the basis of natural medicines exerting their healthcare and medicinal functions.Some phenolics in emblic fruits are mainly responsible for their antioxidant activities and natural vitamin C source displays an important function on the prevention of cancers[38,39].

    Nowadays,chromatographic analysis is the most fundamental technique for the quality assessment of medicinal plants because it can quantifymultiplebioactivecomponentssimultaneously[40,41].We first determined the six main metabolites(gallic acid,corilagin,chebulagic acid,ellagic acid,quercetin,and vitamin C)to investigate the quality variation of emblic materials originating from seventeen geographical origins.Chromatographic plots are exhibited in Fig.S2.Each calibration curve was established by plotting its peak area against the standard concentration(Table S2).Methodological examination, including precision, stability,repeatability and recovery,was conducted(Table S3).These results demonstrated that the HPLC-UV method could be applied to determine the quality variation of emblic medicines.

    The levels of determined active compositions are shown in Table 1.The concentrations of gallic acid,corilagin,chebulagic acid,ellagic acid,quercetin and vitamin C of emblic materials originated from different geographicalorigins are 4.48—61.00 mg/g,0.77—9.82 mg/g, 4.82—32.31 mg/g, 0.90—13.00 mg/g,0.34—3.60 mg/g and 0.47—14.56 mg/g,respectively.Gallic acid in this product from CX origin is almost 14 times that from ST origin,and the vitamin C concentration from MY origin is 30 times more than that from HZ origin.This result shows an obvious quality variation of these fruits from different growing environments.

    從Langmuir模型擬合參數(shù)中可得,花生殼的理論最大吸附量為 136.064 mg/g,碳纖維的最大吸附量為 487.363 mg/g。在Freundlich模型擬合參數(shù)中,花生殼試驗值的擬合k值為 14.030,n值在 0.1 與 0.5 之間,說明花生殼對Cd2+的吸附能力強(qiáng)、易于發(fā)生吸附。

    A PLS-DA model was developed to visualize their quality variation.The regional variation of these products is apparent because they are divided into several groups obviously(Fig.2).According to the loading plot,these six compounds play an essential role in this classification model(Fig.S3).The conclusion can be determined that geographical origins have a significant influence on the quality of emblic products.Because these species are extensively distributed in the world,an effective geographical traceability strategy is very essential for their quality assessment.Chromatographic techniques are always time-consuming,pollution producing and inaccurate to deal with this problem.Hence,two spectroscopic techniques were applied for a better solution for a geographical traceability model of emblic medicines.

    4.2.Spectral pretreatment and outlier diagnosis

    The raw FT-NIR and FT-MIR spectra of emblic products are visualized in Fig.S4.These original spectral signals are sensitive to the operating environment.The optimized spectra based on baseline correction,smoothing and multiplicative scatter correction are displayed in Fig.3.Comparatively,these approaches are effective in improving the spectral quality not only for the visualization of metabolic characterization but also for the subsequent data analysis.Many typical absorption peaks were raised,indicating that their metabolic characterizations are similar.Hence,the metabolic variation of emblic medicines from different geographical origins mainly reflects on the level of metabolic products,which can be partly explained by chromatographic results.

    Two methods were together used for the outlier diagnostic.The result of Hotelling's T2distribution is shown in Fig.S5.Six observations of FT-NIR spectra and three observations of FT-MIR spectra are out of the 95% confidence limit.The result of iForest indicates that the scores of six FT-NIR observations are lower than 2.86 and three FT-MIR observations are smaller than 2.65,respectively(Table S4).Using 2.86 and 2.65 as the threshold scores regarding FTNIR and FT-MIR spectra,four outliers were additionally detected by this algorithm.Summary,a total of ten samples were identified as abnormal individuals and thus they were not used for the subsequent analysis.

    4.3.The results of feature selection

    After spectral pretreatment and outlier diagnosis,two preliminary data matrixes concerning FT-NIR(245×1556)and FT-MIR(245×1789)have been produced.They were too sizable to analyze directly.Twelve feature selection models( filter,wrapper and embedded)were comparatively used to simplify these data structures.

    Fig.S6 shows the performance of four filter models.U-MCFS model performs the highest accuracy for FT-NIR with the ACC and KAP of 92.68% and 92.19% ,respectively,using the first 200 features.Regarding FT-MIR spectra,95.03% and 94.68% of ACC and KAP are calculated using the S-MCFS model with the first 100 features.

    For wrapper models(Fig.S7),the first 120 FT-NIR features perform the best accuracy using the Boruta model,with ACC and KAP of 96.10% and 95.82% ,respectively.For FT-MIR spectra,the intelligence optimization algorithm of GA shows the best accuracy based on the first 40 features.The ACC and KAP are 94.82% and 94.46% ,respectively.

    The results of embedded feature selections are presented in Fig.S8.Compared with two linear models of LASSO and VIP,nonlinear models of PIMP and Gini perform better results.The former performs 92.86% ACC and 92.36% KAP for FT-MIR spectra using the first 120 features.The latter technique performs 95.84% ACC and 95.55% KAP regarding FT-NIR spectra using the first 150 features.

    4.4.The comparison of feature selection

    We applied twelve different feature models,including filter,wrapper,and embedded models.A 3 times cross-validation procedure was performed to propose the best one for the optimization of sizable spectral datasets.They were well evaluated because a total of 30 random samplings were performed.The comparison of their KAP accuracy is displayed in Fig.4.

    Feature selection models exhibit different performance for different datasets.Filter models have a weaker performance than other methods with a significant variation(P<0.05)for the FT-NIR dataset.The variation between the wrapper and embedded models was not significant.Considering feature number,validation accuracy and cost time together,we chose Boruta as the best method to simplify the FT-NIR dataset.

    For the FT-MIR dataset,the variation among different types of feature selections is not significant.LS is the worst algorithmwith a significant variation to others(P<0.05).Comparatively,the S-MCFS model was selected as the optimized method to simplify this dataset.

    The first 100 important variables of FT-NIR and FT-MIR spectra are visualized,respectively,in order to further compare the performance of feather selection models.As seen in Fig.5,LS and ILFS mostly focus on the local region of spectral data.This can explain why they performed a bad result.A local search strategy may be powerless regarding sizable spectral data.Conversely,SA and GA are the randomly global optimization algorithms.Their performances were acceptable,but too much time is needed to achieve these algorithms.For several excellent feature selections such as Boruta and S-MCFS,the features selected by them are mainly distributed on the informative spectral region of 7000-4000 cm-1for the FT-NIR dataset and 2000-500 cm-1for FT-MIR dataset.

    Different feature selection models had different efficiencies when they were used to simplify spectral data.Multiple models need to be together applied to propose the best one for the spectral data optimization of natural medicines.Eventually,120 FT-NIR features and 100 FT-MIR features were selected using Boruta and S-MCFS models,respectively.This is the first time that filter,wrapper,and embedded feature selections were together used for spectral datasets of natural medicines.

    4.5.Development of the geographical traceability model

    An optimized data matrix was successfully generated via spectral pretreatment,outlier diagnosis,feature selection,and data fusion in turn.It contained 245 rows and 220 columns,which was simple,representative and informative.Such a data matrix could contribute to constructing an accurate and robust geographical traceability model of emblic medicines.

    The OOB estimate is based on a bootstrap sampling procedure,which is an unbiased measurement.This parameter was closely related to a model fitting degree,and it can effectively enhance the generalization ability of the model.So it was used to adjust the parameters of an RF model.94 treeshad the best performance,with the lowest averagederrorof 0.029(Fig.S9).Then,a roughset from1 to 100 was designed to select the best mtry.As seen in Fig.6,mtry=65 has the best performance with an error of 0.023.Via the parameter adjustment process,the calibrated geographical traceability model was successfully developed with the OOB error reduced from 0.036 to 0.023.

    A well-chosen external validation set based on Kennard-Stone sampling was imported into the calibrated model to evaluate its generalization performance[42].The confusion matrixes were produced in Table S5 and Table S6.Four samples from QXN,DL and CX groups are misclassified in the calibrated model.The SEN,SPE,and ACC are 97.65% ,99.85% ,and 97.63% ,respectively.All samples in the external validation set are correctlyclassified,with 100% of SEN,SPE,and ACC,respectively(Table 2).

    5.Conclusion

    Natural products are always the complex mixtures that consist of diversified chemical constitutes.Their metabolize characterizations are difficult to illustrate completely.Spectroscopic techniques have many advantages because they can contribute to a rapid and green quality detection for natural medicines.The time for collecting the FT-NIR and FT-MIR spectra of an emblic sample is less than 1 min without sample loss.

    However,spectral data need to be carefully optimized before their application.When the feature subset is well prepared,spectroscopic techniques show a huge potential for the quality assessment of natural medicines on both qualitative and quantitative levels.These techniques should play a more important role in the field of quality assessment for Chinese medicine.

    In this study,we presented an integrated analysis process of two spectral datasets to develop an effective geographical traceability model for emblic medicines.This model performed a 100.00% predicted accuracy for these medicines originating from seventeen geographical origins.These optimization steps included spectral pretreatment,outlier diagnosis,feature selection,data fusion,and machine learning algorithm.This analysis strategy also can be used in quantitative respect and is worth to generalize for the quality assessment of other natural multi-source medicines.

    Conflicts of interest

    The authors declare that there are no conflicts of interest.

    Acknowledgments

    This work is financially supported by the National Wild Plant Germplasm Resources Infrastructure which is the follow-up work of a project called Standardization and Community for the Collection and Preservation of Important Wild Plant Germplasm Resources(2005DKA21006).

    Appendix A.Supplementary data

    Supplementary data to this article can be found online at https://doi.org/10.1016/j.jpha.2019.12.004.

    猜你喜歡
    花生殼碳纖維理論
    花生殼磁性生物炭對水體中Cr(Ⅵ)的吸附研究
    堅持理論創(chuàng)新
    神秘的混沌理論
    理論創(chuàng)新 引領(lǐng)百年
    相關(guān)于撓理論的Baer模
    一種碳纖維加固用浸漬膠的研究
    上海建材(2019年4期)2019-05-21 03:13:02
    花生殼有妙用
    科教新報(2018年48期)2018-06-11 07:19:21
    HP-RTM碳纖維復(fù)合材料中通道加強(qiáng)板研究
    中間相瀝青基碳纖維及其在飛機(jī)上的應(yīng)用
    碳纖維增強(qiáng)PBT/ABS—g—MAH復(fù)合材料的力學(xué)性能和流變行為
    中國塑料(2016年6期)2016-06-27 06:34:16
    亚洲国产毛片av蜜桃av| 久久精品熟女亚洲av麻豆精品| 亚洲精品国产av成人精品| 国国产精品蜜臀av免费| 91精品伊人久久大香线蕉| 水蜜桃什么品种好| 80岁老熟妇乱子伦牲交| 欧美精品人与动牲交sv欧美| freevideosex欧美| 国产在线免费精品| 日韩在线高清观看一区二区三区| 一个人免费看片子| 久久久久久人妻| 人妻系列 视频| 热99国产精品久久久久久7| 亚洲精品中文字幕在线视频 | 乱系列少妇在线播放| 最近2019中文字幕mv第一页| 欧美精品亚洲一区二区| 天美传媒精品一区二区| 中文精品一卡2卡3卡4更新| 一区在线观看完整版| 精品人妻视频免费看| 亚洲av中文av极速乱| 亚洲电影在线观看av| kizo精华| 欧美激情国产日韩精品一区| 国产又色又爽无遮挡免| 男人添女人高潮全过程视频| 亚洲性久久影院| 亚洲电影在线观看av| 日韩人妻高清精品专区| 自拍偷自拍亚洲精品老妇| 欧美日韩在线观看h| 国产成人91sexporn| 丰满人妻一区二区三区视频av| 国产真实伦视频高清在线观看| 人人妻人人澡人人爽人人夜夜| 黑丝袜美女国产一区| 国产在线男女| .国产精品久久| av天堂中文字幕网| 久久国产精品大桥未久av | av国产免费在线观看| 看非洲黑人一级黄片| 欧美精品人与动牲交sv欧美| 国产精品一区二区在线观看99| 全区人妻精品视频| .国产精品久久| 亚洲欧美精品自产自拍| 你懂的网址亚洲精品在线观看| 天天躁夜夜躁狠狠久久av| 99热国产这里只有精品6| 中文精品一卡2卡3卡4更新| 人人妻人人添人人爽欧美一区卜 | 亚洲精品第二区| 99久久综合免费| 久久精品久久精品一区二区三区| 激情 狠狠 欧美| 国产一区亚洲一区在线观看| av免费在线看不卡| 欧美另类一区| 亚洲欧美一区二区三区国产| 午夜视频国产福利| av在线蜜桃| 日本与韩国留学比较| 亚洲av中文av极速乱| 身体一侧抽搐| 一个人看的www免费观看视频| 亚洲国产日韩一区二区| kizo精华| 少妇的逼水好多| 国产av精品麻豆| 日产精品乱码卡一卡2卡三| 日韩不卡一区二区三区视频在线| 国产精品欧美亚洲77777| av卡一久久| 乱系列少妇在线播放| 国产高清有码在线观看视频| 18禁在线无遮挡免费观看视频| 97热精品久久久久久| 国产精品嫩草影院av在线观看| 99热全是精品| 国产 精品1| 久久久久国产精品人妻一区二区| 51国产日韩欧美| 人妻系列 视频| 亚洲av不卡在线观看| 晚上一个人看的免费电影| 最近2019中文字幕mv第一页| 51国产日韩欧美| 亚洲精品456在线播放app| 精品视频人人做人人爽| 午夜福利高清视频| 男女国产视频网站| 人妻少妇偷人精品九色| 在线亚洲精品国产二区图片欧美 | 国产精品99久久久久久久久| 免费看光身美女| 一级毛片aaaaaa免费看小| 久久精品熟女亚洲av麻豆精品| 蜜桃在线观看..| 男女免费视频国产| 夫妻午夜视频| a 毛片基地| 久久久久久久大尺度免费视频| 亚洲精品第二区| 男人爽女人下面视频在线观看| 熟女人妻精品中文字幕| 亚洲av综合色区一区| 直男gayav资源| av黄色大香蕉| 97超碰精品成人国产| 自拍偷自拍亚洲精品老妇| 精品少妇久久久久久888优播| 国产v大片淫在线免费观看| 亚洲成色77777| 国产精品国产三级国产av玫瑰| 久久97久久精品| 狂野欧美白嫩少妇大欣赏| 黄色视频在线播放观看不卡| 国产久久久一区二区三区| 亚洲无线观看免费| 永久网站在线| 日韩不卡一区二区三区视频在线| 少妇裸体淫交视频免费看高清| 亚洲国产欧美在线一区| 偷拍熟女少妇极品色| 王馨瑶露胸无遮挡在线观看| 韩国高清视频一区二区三区| 成人综合一区亚洲| .国产精品久久| 九九久久精品国产亚洲av麻豆| 美女xxoo啪啪120秒动态图| 日本一二三区视频观看| 在线亚洲精品国产二区图片欧美 | 欧美日韩综合久久久久久| 中文天堂在线官网| 亚洲av国产av综合av卡| 国产午夜精品久久久久久一区二区三区| 交换朋友夫妻互换小说| 亚洲精品,欧美精品| 久久久久久久久大av| 最近最新中文字幕免费大全7| 18禁在线播放成人免费| 亚洲国产色片| 超碰97精品在线观看| 国产毛片在线视频| 日产精品乱码卡一卡2卡三| 精品一区在线观看国产| 国产av国产精品国产| av国产久精品久网站免费入址| 国产69精品久久久久777片| 日日啪夜夜爽| 国产成人午夜福利电影在线观看| 人妻一区二区av| 寂寞人妻少妇视频99o| 日韩av在线免费看完整版不卡| 精品国产三级普通话版| 777米奇影视久久| 亚洲aⅴ乱码一区二区在线播放| a级毛色黄片| 在线观看人妻少妇| 国产久久久一区二区三区| 成年av动漫网址| 亚洲丝袜综合中文字幕| 麻豆成人av视频| 美女脱内裤让男人舔精品视频| 自拍欧美九色日韩亚洲蝌蚪91 | 男女免费视频国产| 一本色道久久久久久精品综合| 日日摸夜夜添夜夜爱| 久久人人爽人人片av| 欧美精品亚洲一区二区| 在线观看一区二区三区激情| 亚洲经典国产精华液单| 国产 一区精品| 爱豆传媒免费全集在线观看| 超碰av人人做人人爽久久| 成人无遮挡网站| 蜜桃亚洲精品一区二区三区| 精品亚洲乱码少妇综合久久| 一级毛片aaaaaa免费看小| av专区在线播放| av福利片在线观看| 日本猛色少妇xxxxx猛交久久| 国产精品久久久久久精品古装| 黄色日韩在线| 夫妻午夜视频| 狠狠精品人妻久久久久久综合| 天堂中文最新版在线下载| 国产精品爽爽va在线观看网站| 国产精品国产三级国产专区5o| 国产av一区二区精品久久 | 永久免费av网站大全| 中文字幕av成人在线电影| 精品午夜福利在线看| 久久人人爽av亚洲精品天堂 | 久久久久国产网址| 欧美日韩一区二区视频在线观看视频在线| 久久久久久伊人网av| 最近的中文字幕免费完整| 看十八女毛片水多多多| 777米奇影视久久| 老司机影院成人| 在线观看一区二区三区激情| 国产在线男女| 成人一区二区视频在线观看| 中国美白少妇内射xxxbb| 久久国产精品大桥未久av | 少妇的逼水好多| 国产午夜精品久久久久久一区二区三区| 大片免费播放器 马上看| 高清在线视频一区二区三区| 亚洲国产日韩一区二区| 亚洲成色77777| 精品熟女少妇av免费看| 久久久午夜欧美精品| 精品国产乱码久久久久久小说| 成人特级av手机在线观看| av国产久精品久网站免费入址| 国产深夜福利视频在线观看| 国产免费一级a男人的天堂| 国产精品一及| 日本爱情动作片www.在线观看| 高清欧美精品videossex| 久久久久性生活片| 国产av一区二区精品久久 | 又粗又硬又长又爽又黄的视频| 亚洲精品色激情综合| 一级毛片我不卡| 在线播放无遮挡| 成人国产麻豆网| 少妇猛男粗大的猛烈进出视频| 干丝袜人妻中文字幕| 国产精品蜜桃在线观看| 久久久欧美国产精品| 永久免费av网站大全| 欧美日韩视频精品一区| 大片电影免费在线观看免费| 在线观看国产h片| 久久久久精品性色| 国产色爽女视频免费观看| 久久久久久久久久久免费av| 国产亚洲欧美精品永久| 成年av动漫网址| 精品人妻偷拍中文字幕| 国产精品精品国产色婷婷| 欧美3d第一页| 美女中出高潮动态图| 国产在线免费精品| 国内少妇人妻偷人精品xxx网站| 最新中文字幕久久久久| 97热精品久久久久久| 香蕉精品网在线| 久久久精品免费免费高清| 欧美区成人在线视频| 777米奇影视久久| 三级国产精品片| 国产毛片在线视频| 午夜福利在线在线| 九九久久精品国产亚洲av麻豆| 一级二级三级毛片免费看| 毛片女人毛片| 精品一品国产午夜福利视频| 欧美日韩一区二区视频在线观看视频在线| 日本wwww免费看| 亚洲成人手机| 成年免费大片在线观看| 1000部很黄的大片| 在线观看免费高清a一片| 亚洲精品一区蜜桃| 美女脱内裤让男人舔精品视频| 婷婷色综合大香蕉| 欧美日韩视频高清一区二区三区二| 国产深夜福利视频在线观看| 边亲边吃奶的免费视频| 成人影院久久| 99热这里只有是精品在线观看| 日韩,欧美,国产一区二区三区| 日本wwww免费看| 18禁在线无遮挡免费观看视频| 新久久久久国产一级毛片| 制服丝袜香蕉在线| 国产精品精品国产色婷婷| 麻豆精品久久久久久蜜桃| av专区在线播放| 日韩av不卡免费在线播放| 国产精品女同一区二区软件| 黄色视频在线播放观看不卡| 免费人成在线观看视频色| 身体一侧抽搐| 国产精品人妻久久久影院| 妹子高潮喷水视频| 人妻制服诱惑在线中文字幕| 精品久久久久久久久亚洲| 色哟哟·www| av国产久精品久网站免费入址| 香蕉精品网在线| 欧美日韩综合久久久久久| 国产亚洲91精品色在线| 亚洲四区av| 午夜视频国产福利| 午夜福利网站1000一区二区三区| 日韩制服骚丝袜av| 春色校园在线视频观看| 日韩欧美一区视频在线观看 | 免费观看av网站的网址| 国产精品无大码| 在线观看美女被高潮喷水网站| 国产精品99久久99久久久不卡 | 在现免费观看毛片| 亚洲国产毛片av蜜桃av| 国产精品国产三级专区第一集| 国产国拍精品亚洲av在线观看| 97精品久久久久久久久久精品| 亚洲国产成人一精品久久久| 十八禁网站网址无遮挡 | 国产精品一区二区在线观看99| 一区二区三区精品91| 一边亲一边摸免费视频| 中文乱码字字幕精品一区二区三区| 人人妻人人看人人澡| 久久精品久久久久久噜噜老黄| 国产av国产精品国产| 亚洲精品456在线播放app| 久久这里有精品视频免费| 天堂中文最新版在线下载| 欧美少妇被猛烈插入视频| 成年人午夜在线观看视频| 联通29元200g的流量卡| 天天躁夜夜躁狠狠久久av| 美女国产视频在线观看| 高清欧美精品videossex| 午夜免费观看性视频| 国产精品无大码| 男女边吃奶边做爰视频| 亚洲va在线va天堂va国产| 精品亚洲成a人片在线观看 | 日韩制服骚丝袜av| 毛片一级片免费看久久久久| 国产亚洲av片在线观看秒播厂| 人体艺术视频欧美日本| 国产精品三级大全| 国产精品久久久久久久久免| av福利片在线观看| 亚洲美女视频黄频| 亚洲色图综合在线观看| 国产91av在线免费观看| 性高湖久久久久久久久免费观看| 久久99精品国语久久久| 狂野欧美激情性xxxx在线观看| 日韩 亚洲 欧美在线| 欧美一级a爱片免费观看看| 亚洲精品亚洲一区二区| 日韩免费高清中文字幕av| 久久久久久久久久久丰满| 女人久久www免费人成看片| 国产 一区精品| 啦啦啦视频在线资源免费观看| 99久久人妻综合| 亚洲国产高清在线一区二区三| 丰满乱子伦码专区| 在线亚洲精品国产二区图片欧美 | 美女内射精品一级片tv| 国产大屁股一区二区在线视频| 熟女电影av网| 国产精品欧美亚洲77777| 人妻系列 视频| 欧美日韩综合久久久久久| 一级二级三级毛片免费看| 涩涩av久久男人的天堂| 久久久亚洲精品成人影院| 久久 成人 亚洲| 秋霞伦理黄片| 亚洲av.av天堂| 国产精品伦人一区二区| 欧美三级亚洲精品| 亚洲av电影在线观看一区二区三区| av黄色大香蕉| 欧美xxⅹ黑人| 99九九线精品视频在线观看视频| 欧美3d第一页| 亚洲av欧美aⅴ国产| 午夜福利视频精品| 中文字幕久久专区| av播播在线观看一区| 亚洲高清免费不卡视频| 欧美97在线视频| 黄片无遮挡物在线观看| 女人十人毛片免费观看3o分钟| 成年av动漫网址| 国产精品国产三级专区第一集| a级毛片免费高清观看在线播放| 51国产日韩欧美| 国产成人一区二区在线| 亚洲精品乱久久久久久| 建设人人有责人人尽责人人享有的 | 麻豆乱淫一区二区| 欧美激情极品国产一区二区三区 | 乱码一卡2卡4卡精品| 寂寞人妻少妇视频99o| 最近最新中文字幕免费大全7| 97超视频在线观看视频| 亚洲精品成人av观看孕妇| 日韩国内少妇激情av| 久久久久视频综合| 国产69精品久久久久777片| 黄色日韩在线| 男女国产视频网站| 国产 一区 欧美 日韩| 久久久亚洲精品成人影院| 99九九线精品视频在线观看视频| 又爽又黄a免费视频| 午夜福利高清视频| 王馨瑶露胸无遮挡在线观看| 91精品一卡2卡3卡4卡| 一级片'在线观看视频| 久久久久久久国产电影| 亚洲精品aⅴ在线观看| 伦理电影免费视频| 蜜桃久久精品国产亚洲av| 男女免费视频国产| 成人18禁高潮啪啪吃奶动态图 | 国产av一区二区精品久久 | 高清欧美精品videossex| 欧美区成人在线视频| 国产免费视频播放在线视频| 赤兔流量卡办理| 国内揄拍国产精品人妻在线| 能在线免费看毛片的网站| 日韩视频在线欧美| 免费观看a级毛片全部| 国产精品久久久久久精品电影小说 | 亚洲中文av在线| 亚洲av电影在线观看一区二区三区| 观看免费一级毛片| 精品人妻视频免费看| 毛片女人毛片| 伦理电影大哥的女人| 丰满人妻一区二区三区视频av| 国产片特级美女逼逼视频| 欧美激情极品国产一区二区三区 | 99热国产这里只有精品6| 国产成人aa在线观看| 成人亚洲欧美一区二区av| 麻豆国产97在线/欧美| 高清在线视频一区二区三区| 日本与韩国留学比较| 91久久精品国产一区二区三区| 极品少妇高潮喷水抽搐| 美女主播在线视频| 欧美丝袜亚洲另类| 欧美三级亚洲精品| 亚洲精品久久午夜乱码| 国产成人精品一,二区| 我的女老师完整版在线观看| 天天躁日日操中文字幕| 亚洲综合精品二区| 内射极品少妇av片p| 日韩av免费高清视频| 亚洲av在线观看美女高潮| 中文欧美无线码| 精品久久久久久久末码| 日韩在线高清观看一区二区三区| 国产成人免费观看mmmm| 蜜桃久久精品国产亚洲av| 亚洲精品乱码久久久久久按摩| 精品久久久精品久久久| 网址你懂的国产日韩在线| 精品少妇久久久久久888优播| 免费人成在线观看视频色| 亚洲美女视频黄频| 麻豆精品久久久久久蜜桃| 99精国产麻豆久久婷婷| 高清不卡的av网站| 99热这里只有是精品50| 中文欧美无线码| 国产免费一区二区三区四区乱码| 99国产精品免费福利视频| 蜜桃久久精品国产亚洲av| 亚洲av国产av综合av卡| 免费观看的影片在线观看| 欧美激情极品国产一区二区三区 | 在线观看免费日韩欧美大片 | 亚洲经典国产精华液单| 日韩强制内射视频| 亚洲人成网站高清观看| 免费观看a级毛片全部| 亚洲精品成人av观看孕妇| 你懂的网址亚洲精品在线观看| 在线观看免费高清a一片| 不卡视频在线观看欧美| 九色成人免费人妻av| 国产 一区 欧美 日韩| 亚洲av国产av综合av卡| 欧美精品亚洲一区二区| 国产成人精品久久久久久| 美女福利国产在线 | 国产精品久久久久久久电影| 国产免费一区二区三区四区乱码| 亚洲欧美日韩东京热| 成人影院久久| 精品久久久噜噜| 蜜臀久久99精品久久宅男| 乱系列少妇在线播放| 成人亚洲精品一区在线观看 | 下体分泌物呈黄色| 在线天堂最新版资源| av在线播放精品| 日韩成人伦理影院| 我的女老师完整版在线观看| videos熟女内射| 日韩精品有码人妻一区| 下体分泌物呈黄色| 亚洲成人av在线免费| 久久青草综合色| 伦理电影大哥的女人| 各种免费的搞黄视频| 亚洲av在线观看美女高潮| 亚洲av成人精品一二三区| 国产v大片淫在线免费观看| 亚洲美女黄色视频免费看| 少妇熟女欧美另类| 男的添女的下面高潮视频| 欧美日韩国产mv在线观看视频 | 男女国产视频网站| 亚洲丝袜综合中文字幕| 最新中文字幕久久久久| 欧美极品一区二区三区四区| 免费人妻精品一区二区三区视频| 少妇人妻精品综合一区二区| 国产av码专区亚洲av| 国模一区二区三区四区视频| 久久亚洲国产成人精品v| 九色成人免费人妻av| 男人添女人高潮全过程视频| 美女国产视频在线观看| 国产精品久久久久久久久免| 国产亚洲5aaaaa淫片| 熟女人妻精品中文字幕| 国产免费福利视频在线观看| 国产在视频线精品| 午夜福利在线观看免费完整高清在| 亚洲av免费高清在线观看| 高清午夜精品一区二区三区| 欧美成人a在线观看| 国产一区二区三区综合在线观看 | 国精品久久久久久国模美| 亚洲精华国产精华液的使用体验| 国产国拍精品亚洲av在线观看| 免费黄频网站在线观看国产| 哪个播放器可以免费观看大片| 久久精品久久久久久久性| 久久久久久九九精品二区国产| 精品久久久久久久末码| .国产精品久久| 91久久精品国产一区二区成人| 中文资源天堂在线| 欧美少妇被猛烈插入视频| 午夜视频国产福利| 亚洲精品国产成人久久av| 国产又色又爽无遮挡免| 久久久成人免费电影| videos熟女内射| 大码成人一级视频| 国产伦理片在线播放av一区| 最近手机中文字幕大全| 少妇人妻一区二区三区视频| 少妇精品久久久久久久| 五月玫瑰六月丁香| 国产午夜精品一二区理论片| 狂野欧美激情性xxxx在线观看| 国产色爽女视频免费观看| 国产精品一及| 国产乱人偷精品视频| 国产精品av视频在线免费观看| 超碰97精品在线观看| 免费播放大片免费观看视频在线观看| 亚洲国产成人一精品久久久| 3wmmmm亚洲av在线观看| 亚洲av不卡在线观看| 日本wwww免费看| 在线观看人妻少妇| 国产亚洲91精品色在线| 亚洲国产日韩一区二区| 51国产日韩欧美| 一区二区av电影网| 欧美一级a爱片免费观看看| 多毛熟女@视频| 亚洲性久久影院| 久久久久人妻精品一区果冻| 国产精品人妻久久久久久| 噜噜噜噜噜久久久久久91| 欧美日本视频| 亚洲精品久久午夜乱码| 免费大片黄手机在线观看| 久久 成人 亚洲| 大陆偷拍与自拍| 毛片女人毛片| 国产爱豆传媒在线观看| 久久人人爽人人片av| 亚洲欧美日韩无卡精品| 欧美高清性xxxxhd video| 各种免费的搞黄视频| 国产精品国产av在线观看| 免费观看性生交大片5| 久久国产精品男人的天堂亚洲 | 高清不卡的av网站| 老司机影院毛片| 五月天丁香电影| 亚洲精品aⅴ在线观看| 在线 av 中文字幕| freevideosex欧美| 纯流量卡能插随身wifi吗|