In just-in-time production systems, U-shaped production lines rather than traditional straight lines are often adopted since they have some advantages. The advantages of U-lines over straight lines are that the workstations required can be reduced and the necessary number of workers can be easily adjusted when the demand rates are changed. In this paper, we present a new heuristic based on genetic algorithm to improve the workload smoothness in the U-line. In the proposed algorithm, a new genetic representation is developed which is specific to the problem being solved. To enhance the capability of searching good solutions, genetic operators are designed by using the problem-specific information and heuristics. Extensive experiments are carried out on well-known test-bed problems in the literature to verify the performance of our algorithm. The computational results show that our algorithm is a promising alternative to existing heuristics.