This doctoral thesis aims to present real case studies developed to address industrial problems. Taking into consideration that in real factory environments the production process may suffer variations from several sources of uncertainty, the focus of the study is to build models considering such variations. To address the problems efficiently, we first propose the use of a new technique: Chance-constraint with Risk Allocation(CCRA). In addition to the CCRA technique, we address the stochastic two-stage technique, widely used in the literature to solve production planning problems with stochasticity. Both techniques are compared and their advantages are reported. The first case study represents the reality of an Industry of Production of Pots and Ampoules (IPPA) located in Minas Gerais. The main decision to be made consists in dimensioning the quantity of ampoules and pots that must be produced over a time horizon, in order to meet a certain demand. To solve the problem it was applied the exact method Branch-and-Cut (B&C), a simple Genetic Algorithm (GA) and a Multi-population Genetic Algorithm (MPGA). The results showed that the metaheuristics are able to find the optimal solutions obtained by the exact B&C method. In groups of more complex instances, the metaheuristics outperform the B&C method. In this context, the Production of Pots and Ampoules tool was also developed, to assist in the development of optimized production schedules. In the second case study we are dealing with a production process in an oil extraction industry, where it is necessary to determine the optimal settings in order to obtain the highest efficiency in soybean oil extraction. The entire extraction process is complex, so we will deal with only one main piece of equipment, the rolling mill, whose function is to transform the broken soybeans into small flakes. This equipment is manually controlled by applying hydraulic pressure to the rolls. With the objective of improving production efficiency, aiming to obtain pressure adjustments and flake measurements in an intelligent and automatic way, this paper proposes a stochastic mathematical model that aims to obtain optimal pressure setpoints for a rolling mill, in order to maintain the flake thicknesses in the ideal operation pattern. The results obtained have proven to be superior to the measures that are currently adopted in the factory, increasing annual profits.