An inverse transient method by genetic algorithm (GA) was applied to leak detection in pipe line. Transient flow caused by valve operation was calculated using the characteristics method. Measured pressure waves at one or two points and calculated pressure waves at those points are evaluated to minimize the difference between measured and calculated values. The location and discharge of leak were selected as calibration parameters of GA. The unsteady pipe friction was approximated to be quasi-steady for the calculation of unsteady pressure. Moreover, the effect of the errors in the pipe friction factors on leak detection was discussed. As the result of the present study, leak detection was found to be possible at the node as well as any point in the pipeline. Furthermore, the pressure data at multi-points for GA method was found to be effective to improve the leak detection accuracy.