Multiple surface crack distributed randomly along a weld toe influences strongly on the fatigue crack propagation life of welded joint. It is investigated by using statistical approaches based on series of systematic experiments. From the statistical results, initial crack numbers and its locations follow the normal distribution, and the probability of initial crack depths and lengths can be described well by tile Weibull distribution. These characteristics are used to calculate the fatigue crack propagation life, in which the mechanisms of mutual interaction and coalescence of the multiple cracks are considered as well as the Mk-factors obtained from a parametric study on the crack depths and lengths. The automatic calculation is achieved by the NESUSS, where the parameters such as the number, location and size of the cracks are all treated as random variables. The random variables are dealt through the Monte-Carlo simulation with sampling random numbers of 2,000. The simulation results show that the multiple cracks lead to much shorter crack propagation life compared with those in single crack situation. The sum of the simulation and tile fatigue crack initiation life derived by the notch strain approach agrees well with the experiments.