This paper presents an exponential-based spiral dynamic algorithm (SDA) as an improved version of the original SDA. A simple structure with a unique feature of a spiral model in SDA leads the algorithm achieving fast convergence speed and short computation time. However, adopting a constant radius and a constant angle into the spiral model of SDA causes the algorithm to trap into local optima solution. To overcome the problem, both radius and angle of the spiral model is adaptively varied throughout the search process. It can be done by incorporating an exponential-based radius and an exponential-based angle into the spiral model of SDA. The algorithm is applied to solve dynamic modeling problem of a single-link flexible manipulator system. Results of the dynamic model show the proposed algorithm and the original SDA estimate adequate model and capture good dynamic behaviour of the system. However, the proposed algorithm acquires better model for the system compared to its predecessor algorithm.