This paper deals with the automations of the machine tools' development in the phase of selecting of the optimal structure's configuration by means of neural networks. For this purpose, an algorithm and a programme were developed. By using of them a set of all possible machine tool structure's configurations on the basis of the relations between the movements of tool, workpiece and machine tool units by the machining is determined. Then, structure's configurations, which meet the needed technical requirements, are extracted from this set. The usage of neural networks, due to their learning ability, makes the performance of these steps not only faster as the manual performance but more flexible as the similar programmes based on Booleans logic.