Finally I have found the answer. we should use PSO to Train the neural network and find the weights. by applying this kind of Evolutionary Algorithms the process of training could be short.having said that PSO has strong ability to find global optimist result and Back propagation has ability to find the local.
For doing so reaching the place is not important at all, just we need to check the output and base on it decide which particle is the best.
Hope I explain it properly.