It is an important method for studying the rheological characteristics of rock to choose an appropriate rheological model and obtain the corresponding parameters based on the triaxial creep experiment of rock in lab. However

there are always a few abnormal values in sample data due to the influence of experiment conditions and random noises; the estimation accuracy may be worse if we use the ordinary least square algorithm to estimate parameters. The robust estimation theory derived from statistics is applied to the parameter identification by using double quadratic iteratively weighted least square algorithm; and it can reduce estimation bias due to the presence of abnormal values. This method is compared with that using ordinary least square estimation; the result shows that the former has high accuracy

uniform convergence and strong robustness

which is better than the ordinary least squares estimation.