Abstract: Stochastic back analysis of thermal parameters, including coefficient of thermal conduction, coefficient of heat conduction as well as coefficient of superficial radiation of transient temperature field of mass concrete, is presented using the Bayesian parameter estimation theory. An error function is established, by which, the cumulative time effect can be taken into account. DFP optimization method is implemented in the stochastic back analysis. Corresponding formulas as well as the computational methods and steps are also presented. The thermal problem of a typical mass concrete block is taken as an example to demonstrate the suggested method for the solution of random thermal parameter identification.