This paper develops a new on-line parameter estimation strategy, which takes advantage of BFGS formula to update the inverse Hessian matrix and one-dimensional search to accelerate the tracking speed. The dynamic model for fixed-bed readers is lumped to a proper form for identification by double collocation. The gradients of the quadratic objective function with respect to the estimated parameters are obtained in a more accurate way. The method needs low intensive computation, proved to be superior to the traditional method i.e. the recursive least squares algorithm with forgetting factor, in terms of the speed of tracking, robustness in the presence of errors of the model parameters, and the ability to handle nonlinearity of the fixed-bed reactor.