A mechanism-based approach to experiment design and model building
Tamara Lynn Daugherty · 1992
This research introduces a new approach to the development of models which describe the underlying mechanistic behavior of a system. The approach utilizes an innovative modeling representation that combines digraph modeling and multi-level logic to define the phenomenological causal relationships of a system. Unlike most model representations which have been developed, this representation allows explicit declaration of the mechanisms of a system. The mechanism-based models of all processes which may be of importance in a system can be combined to create a generalized mechanism-based model which describes the behavior of a whole class of systems. Experiment design strategies to determine which mechanisms are of importance in a system have been developed. These methods can be applied to the generalized model of a system to synthesize experiments which isolate individual mechanisms. The experiment design strategies can be used in conjunction with the costs of experiments to determine the optimal experiment for verifying a mechanism. A methodology for estimating experiment costs from the relative accuracies and controllabilities of the variables is introduced. The modeling representation and the experiment design and optimization strategies are implemented in the computer program Edison. The effectiveness of the mechanism-based modeling strategy is illustrated with four example problems: the analysis of packed bed reactor systems, the discovery of electrokinetic lift, the verification of freezing induced motion of rocks below the earth, and the modeling of a glycol cooling system.