模式识别用于压电晶体传感器阵列识别可燃物质

用７个压电晶体组成传感器阵列,每个晶体上分别涂有不同种类的气相色谱固定液,通过测定各种可燃物质燃烧时放出的混合气体来识别所燃物质,在识别中分别应用了人工神经网络法（ＡＮＮ）和逐步判别分析法（ＳＤＡ）。讨论了解决神经网络开始训练时不收敛或产生麻痹现象的方法,提出了训练数据选取的新方法－训练集逐步扩展法,实验证明：人工神经网络对被测物质的识别准确率达１００％,高于逐步判别分析法（８３％）。     

压电晶体传感器阵列用于混合有机溶剂蒸气中官能团的识别

为模拟生物化学传感体系, 提出了可用于识别有机官能团的传感器阵列, 用作人工气味识别系统。该阵列由八个压电晶体传感器组成, 每个传感器涂以具有广谱响应性能的不同吸附活性材料, 阵列对常见小分子有机溶剂混合蒸气的响应频移数据采用逐步判别分析(SDA)处理, 选出五个供信能力最佳的判别变量, 以此构成的阵列用于小分子有机溶剂混合蒸气中醇羟基、羰基与其它官能团的识别, 并采用主成分分析(PCA)法降维投影, 在二维空间含相同官能团的物质聚为一类; 阵列可用于酒类、软饮料的识别。     

压电晶体传感器阵列用于含乙醇的蒸气检测

将多个表面涂有不同选择性吸附材料的压电晶体传感器组成阵列，用于混合有机溶剂蒸气中乙醇的检测。对阵列输出的数据，采用逐步判别法进行二类判别，可以检测混合蒸气中少量的乙醇，采用主成分回归法，可由阵列输出数据预测未知样品中乙醇的浓度。     

环境致癌物的计算机预测研究

环境致癌物可诱发人类或哺乳动物体内的肿瘤,建立环境致癌物的计算机预测模型对环境风险评价和生态安全具有重要的意义．通过构建了3780个化合物的数据集,随机选取其中3024个作为训练集,其余756个作为外部验证集;基于定量构-效关系（QSAR）方法,采用逐步判别分析和主成分分析建立数学模型．结果表明训练集非致癌物预测正确率为86．0％,可能致癌物的预测正确率为88．O％,而采用主成分建模时,非致癌物和可能致癌物的预测正确率分别为74．2％和73．1％．说明逐步判别分析法的结果优于主成分判别分析．同时确定了可能致癌物和非致癌物的分子结构参数,阐明了两者结构差异．以上结果为预测和评估环境致癌物提供参考依据．     

烃类混合气体的神经网络模型检测

八十年代末科学家模仿生物鼻研制一种传感器阵列与计算机模式识别的气体检测系统．传感器阵列相当于生物鼻的嗅觉细胞，计算机模式识别系统相当于嗅泡和大脑「‘］．传感器阵列对气体的响应是一个多维空间的响应模式，这种响应模式经过一定的数学处理后可以实现气体的种类识别或浓度检测［’－‘j．传感器的响应和混合气体浓度之间呈非线性关系，这一特性给定量检测多组分气体混合物造成很大的限制．应用人工神经元网络技术（ANN）可以克服这一缺陷，并使检测气体的选择性大大提高．本工作运用ANN中的反向传播（BP）算法识别由16个不同…     

陶工尘肺辅助性诊断的方法研究——测定发中微量元素 建立判别方程

测定了陶工Ｉ期尘肺患者的头发中Ｚｎ、Ｃｕ、Ｓｅ、Ｎｉ、Ｆｅ、Ｍｎ、Ｍｇ７种金属元素，用陶工可疑患者作对照，建立Ｆｉｓｈｅｒ判别方程，对陶工尘肺进行判别。自身回代正确率９４．４４％，前瞻性回代达８３．３３％，各其流行病学指标计算均属良好。     

改进的人工神经网络在传感器阵列信号处理中的应用

本文提出传感器阵列信号处理的人工神经网络模型，以Cu^ 2/Ca^ 2系统为研究对象．尝试了神经网络方法的效果。其最大相对误差不超过5．％，最大相对预测误差不超过2．4％，结果表明，该方法性能良好，在各种传感器阵列的信号处理方面有广泛的应用前景。     

基于随机森林与Chemistry Development Kit描述符的P-gp底物识别

应用随机森林方法、开放源代码软件-CDK(Chemistry Development Kit)描述符与170个化合物的训练数据集[其中96个为磷糖蛋白(P-gp)底物], 建立了P-gp底物的识别模型. 研究了CDK描述符与P-gp底物识别的关系, 结果表明, 原子极化性和电荷偏面积等分子属性对P-gp底物识别起到重要作用. 该模型对训练集的预测正确率为99.42%; 对外部测试集(42个化合物, 其中24个为P-gp底物)的预测结果为P-gp底物、非底物及总测试集的识别正确率分别为87.50%, 83.33%和85.71%. 212个化合物数据集上的Leave-One-Out交叉验证识别正确率为77.4%.     

相转移催化下双媒质体系对醇类的选择性间接电氧化

在相转移催化剂（Ｂｕ４ＮＨＳＯ４）作用下，用Ｃｒ（Ⅵ）／Ｃｒ（Ⅲ）及Ａｇ（Ⅱ）／Ａｇ（Ⅰ）双媒质体系对醇类进行间接电氧化，产率为７５．４～９７．５％，电流效率达６０．２～７５．８％，双媒质体系可重复使用。     

冠心病微量元素谱的计算机模式识别

用计算机多元分析研究冠心病微量元素谱，识别冠心病患者与健康者；非线性映射法判别率男性８６．６％，女性９６．２％；ＳＩＭＣＡ差别法正确回判率男性８５．０％，女性８８．３％。     

Boronic acid based peptidic receptors for pattern-based saccharide sensing in neutral aqueous media, an application in real-life samples

The advent of the alternative sweeteners market has signaled a demand for chemosensors which target multiple saccharides and saccharide derivatives, in aqueous media at physiological pH. This demand has largely been unmet as existing molecular receptors for saccharides have generally not shown sufficient degrees of affinity and selectivity in aqueous media. A chemosensor array for saccharides and saccharide derivatives, fully operational in aqueous media at physiological pH, has been developed and is reported herein. Boronic acid based peptidic receptors, derived from a combinatorial library, served as the cross-reactive sensor elements in this array. The binding of saccharides to these receptors was assessed colorimetrically using an indicator uptake protocol in the taste-chip platform. The differential indicator uptake rates of these receptors in the presence of saccharides were exploited in order to identify patterns within the data set using linear discriminant analysis. This chemosensor array is capable of classifying disaccharides and monosaccharides as well as discriminating compounds within each saccharide group. Disaccharides have also been distinguished from closely related reduced-calorie counterparts. This linear discriminant analysis set was then employed as a training set for identifying a specific saccharide in a real-world beverage sample. The methodology developed here augurs well for use in other real-world samples involving saccharides as well as for sensing other desired analytes.     

Recognition of benzene, toluene and xylene using TGS array integrated with linear and non-linear classifier

Three volatile organic compounds (VOCs): benzene, toluene and xylene were measured with an array of six Taguchi gas sensors in the air with variable humidity content. The recognition of single compounds was performed, based on measurement results. The principal component analysis (PCA) pointed at humidity as the main classification factor in the measurement data set. The linear discriminant analysis (LDA) was applied to overcome this drawback and enforce classification with respect to benzene, toluene or xylene. It was shown that discriminant function analysis (DFA), which is an LDA method allowed for 100% success rate in test samples recognition of benzene. It did not allow for accurate recognition of test samples of toluene or xylene. Following, the non-linear classifier, radial basis function neural network (RBFNN) was applied. A specific configuration of input ‘s was found, which provided for successful recognition of each single compound: benzene, toluene or xylene in air with variable humidity content.     

Recalibration of a gas-sensor array system related to sensor replacement

Univariate multiplicative drift correction and multivariate component correction were applied for recalibration of long-term measurement data acquired with a solid-state gas-sensor array system. The efficiency of the methods was evaluated by classifying recalibrated measurement data using k-nearest neighbor classification and partial least-squares discriminant analysis. For the measurement data in this experiment both multiplicative drift correction and component correction appeared to be useful for recalibration of measurement data from the new gas-sensor array with regard to measurement data acquired with the old replaced gas-sensor array.     

Investigation of crimson-dyed fibres for a new approach on the characterization of cochineal and kermes dyes in historical textiles

The colorant behaviour of cochineal and kermes insect dyes in 141 experimentally-dyed and 28 artificially-aged samples of silk and wool was investigated using ultra-high performance liquid chromatography coupled to photodiode array detector (UHPLC-PDA), liquid chromatography electrospray ionisation mass spectrometry (LC-ESI-MS) and image scanning electron microscopy – energy dispersive X-ray spectroscopy (SEM-EDX). Partial-least squares discriminant analysis (PLS-DA) was then used to model the acquired UHPLC-PDA data and assess the possibility of discriminating cochineal insect species, as well as their correspondent dyed and aged reference fibres. The resulting models helped to characterize a set of 117 red samples from 95 historical textiles, in which UHPLC-PDA analyses have reported the presence of cochineal and kermes insect dyes.     

Development and validation of a high performance liquid chromatography-diode array detection method for the determination of aging markers in tequila

A solid-phase extraction procedure followed by analysis by high performance liquid chromatography (HPLC) with UV-vis photodiode array detection (DAD) is proposed to simultaneously determine 11 aging markers in tequila. The method showed good intraday (n=5) and interday (n=3) precision, RSD<1.6% in both cases, for each of the identified compounds. The calibration curves were linear at the tested ranges (R(2)>0.999). Good recoveries (84.2-108.5%) were obtained for 10 of the 11 compounds studied; and the LOD and LOQ ranged from 0.62 to 4.09 microg/mL and 1.9-12.4 microg/mL, respectively. The proposed methodology was applied to a set of 15 authentic tequila samples grouped by aging state (blanco, reposado and a?ejo). An ANOVA analysis combined with discriminant analysis with stepwise backward variable selection was used to differentiate between the various aging groups based on their oak related compounds content.     

Chemotaxonomic study of Chrysobalanus icaco Linnaeus (Chrysobalanaceae) using ultra‐high performance liquid chromatography coupled with diode array detection fingerprint in combination with multivariate analysis

We investigated a strategy for the chemotaxonomy study of Chrysobalanus icaco Linnaeus (Chrysobalanaceae) based on ultra‐high performance liquid chromatography coupled with diode array detection fingerprint in combination with multivariate analysis. Two models using principal component analysis and partial least squares discriminant analysis were developed, and the samples could be successfully classified into two classes: Class 1 (red morphotype) and Class 2 (white and black morphotypes). Furthermore, ultra‐high performance liquid chromatography coupled with diode array and electrospray ionization tandem mass spectrometry was used to identify the main compounds responsible for class separation. The partial least squares discriminant analysis model accurately classified the C. icaco samples using an external validation subset with prediction ability of 100% and revealed the existence of two chemotypes. The most important finding obtained in this study is that the three morphotypes distinguished by the mature fruit color (white, red, and black) are not all phytoequivalent to each other.     

Differentiation of Mint (Mentha haplocalyx Briq.) from different regions in China using gas and liquid chromatography

In this study, complex substances such as Mint (Mentha haplocalyx Briq.) samples from different growing regions in China were analyzed for phenolic compounds by high‐performance liquid chromatography with diode array detection and for the volatile aroma compounds by gas chromatography with mass spectrometry. Chemometrics methods, e.g. principal component analysis, back‐propagation artificial neural networks, and partial least squares discriminant analysis, were applied to resolve complex chromatographic profiles of Mint samples. A total of 49 aroma components and 23 phenolic compounds were identified in 79 Mint samples. Principal component analysis score plots from gas chromatography with mass spectrometry and high‐performance liquid chromatography with diode array detection data sets showed a clear distinction among Mint from three different regions in China. Classification results showed that satisfactory performance of prediction ability for back‐propagation artificial neural networks and partial least squares discriminant analysis. The major compounds that contributed to the discrimination were chlorogenic acid, unknown 3, kaempherol 7‐O‐rutinoside, salvianolic acid L, hesperidin, diosmetin, unknown 6 and pebrellin in Mint according to regression coefficients of the partial least squares discriminant analysis model. This study indicated that the proposed strategy could provide a simple and rapid technique to distinguish clearly complex profiles from samples such as Mint.     

Identification of combustible material with piezoelectric crystal sensor array using pattern-recognition techniques

A promising way of increasing the selectivity and sensitivity of gas sensors is to treat the signals from a number of different gas sensors with pattern recognition (PR) method. A gas sensor array with seven piezoelectric crystals each coated with a different partially selective coating material was constructed to identify four kinds of combustible materials which generate smoke containing different components. The signals from the sensors were analyzed with both conventional multivariate analysis, stepwise discriminant analysis (SDA), and artificial neural networks (ANN) models. The results show that the predictions were even better with ANN models. In our experiment, we have reported a new method for training data selection, 'training set stepwise expending method' to solve the problem that the network can not converge at the beginning of the training. We also discussed how the parameters of neural networks, learning rate eta, momentum term alpha and few bad training data affect the performance of neural networks.