Fragility curves represent the conditional probability that a structure’s response may exceed the performance limit for a given ground motion intensity. One of the most important problems in the development of fragility curves is the high number of analyzes used to evaluate random parameters. The Monte Carlo technique usually requires a relatively large number of simulations in order to obtain a sufficiently reliable estimate of the fragilities making it computationally expensive and time consuming. Hence, methods such as the HDMR and Cornell’s method are proposed to estimate the fragility curves with much less time history analysis. They are used to replace the algorithmic performance-function with an explicit functional relationship, fitting a functional approximation, thereby reducing the number of expensive numerical analyses. After the functional approximation has been made, Monte Carlo simulation is used to obtain the fragility curve of the system. The idea of ??these methods is very suitable for reducing the number of analyzes, but due to the selection of inputs of uncertainties without initial analysis, their compatibility with the Monte Carlo method is not high at all performance levels. On the other hand, because the fragility curves are developed at different levels, it is necessary that input points be selected compatible with each performance level in order to increase the accuracy of the generated curves at all performance levels. In this study a new method has proposed to increase accuracy and decrease the numbers of non-linear time history analysis. The proposed method is based on the selection of suitable points in uncertaites and also the more accurate calculation of the dispersion in the fragility function. In this study the defects of previous methods have been eliminated so that for selection of inputs the scientific reasons with analytical support have been proposed. The responses are approximated with more relationships and consequently with greater accuracy, so that their adaptation to the Monte Carlo method is very high. Also for each performance level, the input points have been selected compatible with that level to the accuracy of the fragility curve at all performance levels. To control the validity of this method, a special study was carried out on a SDOF system and a reinforced concrete frame under various ground motions. Then the fragility curves are developed and compared at different performance levels using different methods. The results show that compatibility of the proposed method with Monte Carlo simulation is more than other methods. Keywords Fragility Curve, Monte Carlo method, Compliant response functions, Cornell’s method, HDMR method, Performance levels.