Computer enhancement of laboratory data collection and analysis
W. A. Schlieper, THOMAS L. ISENHOUR · 1988
Computers are used routinely to collect and analyze experimental data. This dissertation presents various methods of data collection and analysis to aid in the development of analytical chemistry. A prefilter search system based on interferograms is presented. A set of four vectors are generated from four sets of functionally specific data base vectors. These prefilter vectors are used to limit the number of data base entries that need to be searched. The top hit compounds of the prefilter search can be kept identical as a sequential search while only searching a small percentage of the data base. A generalized computer program, ARTS (Analytical Robot Telecommunications Software), has been developed to give the research scientist more flexible control of laboratory robots and instruments. As a stand alone program, ARTS is a complete laboratory control language. ARTS can also be an extension of other software in either a master of slave mode. As master, ARTS can call on other software to perform certain tasks. In a slave mode, ARTS can act as a sensory extension of the calling software. ARTS is a flexible laboratory control language capable of adapting to changing laboratory requirements. The expert system in Chapter three uses ARTS to control laboratory instruments. A microcomputer-based expert system that controls a standard, laboratory robotic system is introduced. The expert system is capable of performing direct complexometric titrations on metal cations in solution. Users of the expert/robotic system can provide instruction by forcing the system to analyze samples under certain conditions. The system can also use heuristic rules, based on conditional stability constants, to make decisions. By storing all titration results, the system is capable of learning from past experience. Building expert systems can be a long and arduous process. The use of the ID3 algorithm as a development tool for building expert systems is described. The ID3 algorithm has been applied to two sets of chemical data resulting in decision trees useful to classify the data. The resulting decision trees were then transformed directly to a set of production rules for an expert system. This procedure allows the attributes of the data set to be represented directly as objects that are descriptive of the actual data and does not require data transformations. (Abstract shortened with permission of author.)