Abstract: Abstract：Based on the PSO algorithm to design the inversion program of prediction parameters of proba-bility integral method. By designing experiments, these four experiments ,including the study of reliability of the PSO algorithm, resisting the effect of gross errors, the losses of observation points and the reliance on initial value are researched.The results show that the use of PSO algorithm can accurately inverse the prediction parameters of probability integral method ,which has a strong anti-interface ablility to the influences of random errors in observed values and the losses of observation points. When the given initial values’ error is no more than 10%, the first inversion can get satisfactory results; When the initial error is no more than 50%,the inversion result of subsidence coefficient and the main affect of tangent is not affected by initial value. Further analysis show that putting inversion parameters into the second inverse can get a more accurate results. Meanwhile, moderately increasing the scales of the initial population can improve the reliability and accuracy of parameter inversion.

Key words: psoalgorithm, probability integral method, parameter inversion, design of experiment