Constructions made of fiber-plastic composites are primarily light and thin-walled. Different semi-finished products (materials, weaving etc.) give the designer a wide range of possibilities. Thus, a component can be dimensioned and manufactured according to the requirements and stresses. But exactly this dimensioning requires a complex determination of the material parameters of the unidirectional single layer. If the strengths and stiffnesses are too low, the structure becomes too heavy, if they are too high, the structure may fail. In a novel approach, the structural properties of the laminate will be predicted by machine learning. By means of an orthotropic damage model a fast nonlinear calculation shall be realized. The goal is to shorten the calculation and development time.