After steel industry and oil industry casting industry is one of the most important industries. Casting parts trading is a large market and it has large amount in international trade. A suitableproduction planning is an important problem in the foundries. Production planning helps to good operations like, appropriate response to the needs of customers, reduce the production costs in the foundries. In this dissertation, simultaneous lot-sizing and scheduling in small foundries is studied. The properties of small foundries include, only one furnace is usually operating at any point in time, the preparation of sand molds is a manual process, slight use of automation and the existence of finishing step for correction of casting parts. In recent decade, one of the significant visions is major increase in level of customer service in Just-In-Time (JIT) in make to order systems. According to this vision, reducing the cost of delay orders and holding products as one of the goals of this dissertation is considered. In this study, according to the existing constraints in the production system of small foundries, a mathematical model for the simultaneous lot-sizing and scheduling in small foundries, on the basis of GLSP is developed. The objective function of this model is minimizing the total costs of preparing the furnace, work-in-process and final product holding and delay orders.Features of this model include decision-making about the entire process of production simultaneously, considering constraints on production system and coordinating between molding and pouring steps. So far in the literature, production planning in foundries with all of mentioned features has not yet been considered. Since this is a NP-Hard problem, so the mathematical model cannot solve large size problems. Two heuristic algorithms based on fixing and optimizing and a heuristic algorithm based on rolling horizon are developed to solve the large scale problems. These algorithms are evaluated by solving 120 instances. The average time to solve these problems in the algorithm based on rolling horizon is 40.22 seconds and in the first and second algorithms based on fixing and optimizing is 62.13 and 48.18, respectively. Computational results show while these algorithms reduce solution time significantly, the solutions quality is suitable. In terms of the solution quality, the average distance the solutions obtained by the algorithm based on rolling horizon and the first and second algorithms based on fixing and optimizing from solution obtained by implementation of the proposed mathematical model in the duration 7200 seconds, is 4.47, 2.52 and 2.73 in percentage, respectively. So in terms of solution time, the algorithm based on rolling horizon and in terms of the solution quality, the first algorithm based on fixing and optimizing have a better performance.