The invention relates to a variable-mass broken-rock permeability parameter computing method. According to the variable-mass broken-rock permeability parameter computing method, a genetic algorithm is designed, and the power exponent function relationship among the permeability, the non-Darcy flow beta factor and the acceleration coefficient satisfies the equations (please see the equations in the specification); the reference value k [r] of the permeability, the reference value beta [r] of the non-Darcy flow beta factor, the reference value c [ar] of the acceleration coefficient, the power exponent n [beta] and the power exponent n [c] are optimized, so that the difficulty that a tester lacks experience is avoided, and a reasonable decision variable can be obtained without referring to previous test results; the bit string, the k [r], the beta [r], the c [ar], the n [beta], the n [c], the seepage velocity time curve, the optimal individual permeability, the non-Darcy flow beta, the acceleration coefficient and the fitness of an initial population and all new populations of each new generation can be output; according to a pressure gradient time sequence and a seepage velocity time sequence, the permeability, the non-Darcy flow beta factor and the acceleration coefficient at each sampling moment are worked out, and therefore conditions are provided for analysis of the rule of variation, caused by mass loss, of the broken-rock permeability parameter.