Abstract: Based on Helmert's variance components estimation formula,the rationality of target function of joint inversion with weight ratio is demonstrated and a sequential algorithm of joint inversion is put forward in this paper.This algorithm,inherited the statistical unbiased property of Helmert's formula,is able to avoid negative variance caused by Helmert's simplified formula sometimes.It does not need to retain extra process data inversed from each kind of observations.This overcomes the shortcomings,such as huge memory occupied and calculation cumbersome,of original Helmert's formula.Otherwise,linear formula conducted in this paper,is convenient for program calculating.A simulation example,underground fault's joint inversion by displacement and gravity observations data,is given in the end.It shows that,combined with other nonlinear mean,this algorithm can be applied equally to non-linear joint inversion problems.