Abstract：First of all, this paper uses the particle swarm algorithm, genetic algorithm, and the Monte Carlo method, respectively combined with the dislocation model, to invert three-dimensional sliding rate of the single fault by gravity data simulated by the pure strike-slip fault. By the result, the paper selects the optimal nonlinear algorithm. Then, the three-dimensional sliding rate of Xianshuihe fault is inverted by two phase ground gravity data, recorded from September 2013 to September 2014 and from September 2014 to April 2015. The results show: (1) The particle swarm algorithm, as compared with the genetic algorithm and the Monte Carlo method, has good convergence, strong stability and high precision; but on the whole, they all have certain application value. (2) There are some characteristics of sliding in the Xianshuihe fault zone, which presents a left-lateral strike-slip movement, with thrust fault on the whole and extrusion on the local scale. (3) The strike-slip and dip-slip parameter of Xianshuihe fault inverted by gravity data from September 2014 to April 2015 is greater than the inversion result by gravity data from September 2013 to September 2014.
