Kanban is an approach to provide an effective relation between different components of a Supply Chain System, such as suppliers, manufacturers, warehouses, distribution centers, retailers and finally customers. Instead of pushing parts, in kanban system each proceeding plant(s) sends a signal to the preceding plant(s) for needed parts to be sent. The plants only produce and deliver desired parts when they receive a signal. Number of kanbans can significantly influence the load balance between processes and the amount of orders suppliers need to obtain from subcontractors. A large kanban size results in large amount of work-in-process(WIP) at each workstation. Although, reducing kanban size serves a decrease in WIP, it leads to transportation increase as well as system throughput rate reduction. In this study a multi-echelon supply chain system is considered. Since the model used in this study is mixed integer non-linear programming and solving by exact algorithms such as branch and bound takes lots of time specially when the scale increases, a heuristic method via Genetic Algorithm(GA) is presented and some problems are solved using the proposed GA to illustrate its performance.