Applications of pattern recognition and artificial intelligence to some problems in analytical chemistry

Chapter I. A quantitative measure of library search reliability was developed. Applications of the Quantitative Reliability Metric (QRM) to measuring the reliability of library searches for unknown target spectra and using this measure to detect the failure of a library search caused by noise, contaminant peaks and missing library spectra are discussed. The effect of noise and mixture infrared composite spectra on the QRM is examined for test sets of 561 infrared spectra. The QRM is also used to evaluate the search performance of an infrared library compressed by eigenvector projection.;Chapter II. Closure is caused by normalizing a data set and affects any multivariate analytical method applied to that data set. Two common methods of normalizing infrared spectra (IR), to unit maximum absorbance and to unit vector length, are evaluated by measuring library search performance. Search performance is evaluated, by using the Quantitative Reliability Metric (QRM), as a function of noise frequency and noise level.;Chapter III. A brief history of infrared spectral abbreviation methods is presented. Different methods of data preprocessing were evaluated for the compression of infrared spectral libraries by eigenvector projection. The effect of noise on compressed library searches was examined. A compressed infrared library achieved an 81% reduction in size without any loss in search performance.;Chapter IV. An algorithm was devised to calculate robust eigenvectors for the specific purpose of compressing spectral libraries. Infrared libraries compressed with robust eigenvectors were compared to libraries compressed with conventional eigenvectors and a non-compressed library. A method for locating poor quality infrared spectra in large databases is also discussed.;Chapter V. The Temporal Optimizer of Robotic Task Sequences (TORTS) expert system was devised as a programming aid and a precursor to the merging of laboratory robotics and artificial intelligence. This program predicts the feasibility and run times of various robotic program configurations. The TORTS system can explore different configurations of robotic programs to minimize the procedure completion time and efficiently allocate resources.