Date of Award

Spring 2008

Document Type


Degree Name

Doctor of Philosophy (PhD)


Mechanical Engineering

Committee Director

Jen-Kuang Huang

Committee Member

Chuh Mei

Committee Member

Abdelmageed Elmustafa

Committee Member

Julie Zhili Hao


The pharmacodynamics and pharmacokinetics of caffeine have been well characterized. In this study, a caffeine dynamic model is developed to describe its pharmacodynamic effects on vigilance performance. Validated biomathematical models developed to address both individual and group fatigue and alertness in a non-laboratory setting represent a tremendous commercial opportunity. First, a test data set with caffeine effects isolated from circadian and homeostatic effects is created. Then a modeling approach for input and output effects is developed and different model structures for the caffeine effects are considered. Observer/Kalman filter Identification (OKID) algorithm is proposed and developed to identify the caffeine model from the created input/output data. The identified caffeine model is then tested to fit for the test data. In this caffeine model, five system parameters [a 1,a2,c1, c2,d0] can be identified by using the proposed OKID algorithm. Identification of the individualized caffeine model shows that the first two coefficients [a1, a2] have small variations for users of both low and high amounts of caffeine among all doses. The 100 mg model has a statistically higher caffeine response as compared to the response of the 200 mg or 300 mg models based on the individualized caffeine models identified from test data. The result also shows that users of both low and high amounts of caffeine users have comparable responses based on the 100 mg model. However, the responses of the 200 mg or 300 mg models show that users of high amounts of caffeine have a statistically higher response to caffeine. In conclusion, the results suggest that the caffeine dosage and habitual usage do not have much impact on the individualized caffeine model dynamics, and the proposed individualized caffeine model can be modified by adding a dose factor to the input of the model to improve the prediction of the performance of other caffeine doses.