Chemometric Methods for the Determination of Volatile Organic Compounds with Microsensor Arrays.

Chemometrics;Chemical Sensor Array;Pattern Recognition;Multivariate Calibration;Volatile Organic Compound;Microanalytical System;Chemical Engineering;Chemistry;Mathematics;Public Health;Science (General);Statistics and Numeric Data;Health Sciences;Science;Industrial Health

Jin, ChunguangVincent, James H. ;

University of Michigan

关键词: Chemometrics;    Chemical Sensor Array;    Pattern Recognition;    Multivariate Calibration;    Volatile Organic Compound;    Microanalytical System;    Chemical Engineering;    Chemistry;    Mathematics;    Public Health;    Science (General);    Statistics and Numeric Data;    Health Sciences;    Science;    Industrial Health;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/61627/jincg_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

This research addresses critical chemometric modeling and data analysis functions needed to guide the design and implementation of novel meso-scale and micro-scale instrumentation that incorporates chromatographic separation with microsensor array detection. The first issue addressed relates to the fidelity of the response pattern generated by an array for a chromatographically resolved analyte to its calibrated reference pattern. A statistically rigorous decision rule was developed that accounts for the inherent variability in the array signals and permits assessments of pattern fidelity at a known rate of error.Building on this first study, a more sophisticated and robust approach to peak purity assessment based on fixed-size moving window factor analysis was adapted and evaluated by simulation.With this approach, a minority component with a peak area 0.5% of the primary component could be detected at a primary-peak signal-to-noise ratio as low as 20:1. To address problems involving partial overlap of chromatographic peaks, a self-modeling curve resolution method was applied, which entails an alternating least square algorithm coupled with evolving factor analysis.Response patterns from an array of four sensors were tested with binary co-elutions to evaluate the resolution of mixture components as a function of random noise, chromatographic separation, pattern similarity, and relative composition. In a separate series of studies, the advantages of multi-transducer sensor arrays over single-transducer arrays for vapor recognition were examined. Starting with a database of sensitivities to 11 vapors from 15 microsensors, it was shown by Monte Carlo simulation and principal component regression modeling that optimal MT arrays consistently outperform ST arrays of similar size, and that with judiciously selected 5-sensor MT arrays one-third of all possible ternary vapor mixtures are reliably discriminated from their individual components and binary component mixtures, whereas none are reliably determined with any of the ST arrays.Using the same database, the limits of recognition were determined for various mixtures, revealing that, in general, mixtures cannot be recognized at relative concentration ratios exceeding 20:1 between two components.Collectively, the research reported here has served to help define the limits of performance and interpret the output of microsensor arrays as components of microanalytical systems.