It is assumed that travelers adjust route choices according to updated experience and guidance information from the Ｒeliable Path Searching System in the form of a travel time budget in degradable transport network．With the consideration of bounded rationality and reference dependency，we develop a descriptive dayto-day dynamic model of network flow in the framework of Cumulative Prospect Theory ( CPT) ． This model reveals how travelers learn，update and adjust their travel time budgets as well as route choices from day-to-day．The properties of the day-to-day dynamic model are then discussed，and a solution algorithm is proposed to solve the model． We then conduct numerical examples to illustrate its properties，and it was demonstrated that the day-to-day dynamics can quickly evolve to be convergent when the guidance information is relatively accurate ( with smaller prediction error) ． Furthermore，the convergence state of the day-to-day dynamic model is approximately identical to Wardrop user equilibrium． Such an understanding of complex travel behaviour has important implications on transportation planning and management．