The research presented in this thesis concern the prediction and improvement of performance of machine tools in parallel structure in order to produce parts that meet the quality required. These works are part of a global optimization problem of the triple play / process / machine. The problem lies in the influence of the structure on the quality of the workpiece. The thesis is illustrated through the analysis of the influence of geometrical and static behavior of parallel structures on the shape of the machined surfaces. This work prppose then advanced along two fundamental axes: Defining a workspace statistically permissible, A proposed method for the identification process. These approaches are proposed in an industrial context, allowing to define the concept of maintenance of the machine geometry and identifying global / local.
