This thesis deals with numerical simulations of gravity sand casting processes for the production of large steel parts. The entire manufacturing process is numerically modeled and evaluated, taking into consideration mould filling, solidification, solid state cooling and the subsequent stress build up. The thermal analysis is then combined with evolutionary multi-objective optimization techniques in a search for the optimal thermal aspects and conditions for producing sound and competitive castings. The goals of the optimization procedure are related to the casting and rigging design and to defects occurrence. In other words, it is desired to eliminate all of the potential casting defects and at the same time to maximize the casting yield. The numerical optimization algorithm then takes these objectives and searches for a set of the investigated process, design or material parameters e.g. chill design, riser design, gating system design, etc., which would satisfy these objectives the most. The first step in the numerical casting process simulation is to analyze mould filling where the emphasis is put on the gating system design. There are still a lot of foundry specialists who ignore the importance of a good gating system design. Hence, it is common to see, especially in gravity sand casting, “traditional gating systems” which are known for a straight tapered down runner a well base and 90º bends in the runner system. There are theories supported by experimental results claiming that flow patterns induced by non-optimal gating systems can cause a variety of defects which are generally not considered to be filling related, such as hot tears and channel segregates. By improving the gating technology in traditional gating systems it is possible to achieve much higher casting integrity with less defects and also to reduce the amount of metal to be re-melted, hence reducing the energy consumption for melting in foundries. Guidelines on how to approach gating system design are given together with examples on how the separate elements constituting gating systems which affect filling patterns and subsequent defects occurrence. Investigation and optimization of thermal behavior of steel castings is the main focus area of this thesis. The intention is to show and discuss a relation among three well-known casting defects, namely centerline porosity, hot tears and macrosegregation. It occurs that all of these defects depend on thermal gradients, cooling rate, pressure drop over the mushy zone and solidification pattern. It is not a standard procedure in daily foundry practice to run convection, macrosegregation and stressstrain calculations on all projects to identify macrosegregation and hot tearing, because of insufficient computational power in many foundries. As a result, many castings are being produced without any knowledge of these two defects. The consequences are known to everybody. The methodology or approach, adopted during the study, lies in utilizing the prediction of centerline porosity for the subsequent assessment of hot tears and macrosegregation. The Niyama criterion is used for this purpose. If there are any narrow areas or channels with high Niyama values, they indicate a presence of high thermal gradients over a narrow region and hence high thermal straining which may lead to hot tearing. If Niyama values are very low, flat thermal gradients are present in this area, which means that there is not a directional and progressive solidification pattern and there will be issues with centerline porosity. Moreover, flat thermal gradients imply large extent of the mushy zone which may promote macrosegregation and especially channel segregates. One can now see that indeed macrosegregation and hot tearing can be addressed by standard thermal calculations, just by using the Niyama criterion. This should not be understood as stress-strain analysis or convection calculations can be entirely omitted, not at all. But they can be reasonably substituted in the initial (informative) calculation or when there is not time to run proper stress-strain or convection calculations. The methodology applied in the thermal analysis is then exploited in numerical optimization. It is shown and verified that it is possible to eliminate hot tears and macrosegregation issues by minimizing centerline porosity and by establishing the directional and progressive solidification pattern towards the heaviest area of the casting. Multi-Objective Genetic Algorithms are applied to handle this task. Three industrial case studies are presented in which minimization of riser volumes and minimization of the three aforementioned defects are pursued by modifying the riser and chill designs and their placement.