Vehicle routing problem with time window is of profound theoretical research significance and broad practical application value.we propose a membrane algorithm with genetic mechanism to improve the convergence speed or population diversity, because traditional heuristics still have shortcomings in these two problems. In this algorithm, we introduce membrane techniques to increase the diversity of population. We put forward time classifier to further accelerate the evolving speed of each membrane. We propose a new crossover operator in order to further improve the successful probability of crossover operator; Beside,we can also designed an improved roulette mechanism so as to modify the quality of solutions. For membrane algorithm, its most prominent advantage is that the distribution and the parallelism can improve both he ability of local search and global search and the efficiency of this algorithm. The experimental results shows that membrane algorithm with genetic mechanism is competitive with other heuristics.