Continuous integrated solutions from CAD down to the preparation of NC programs were developed in the recent years. However, if tolerances should be considered, the interaction of human experts is still necessary. A way to fill this gap in the production process is shown in this thesis. The study builds a relationship between the given design tolerances and including these tolerances in machining by generating respective G and M codes. The study focuses on physical phenomena and their inter-relationship while manufacturing. For example how the speed of machining, torque, power, depth of cut, etc. influences machining under specified tolerances. Artificial neural networks (ANN) have been used to generate required outputs because of their capability to learn from a given set of data points. Four different kinds of neural networks, as a module, have been used. with different kinds of learning rules (algorithms) depending on the type of inputs and outputs. The whole model incorporates retrieval of tolerances from a CAD software and running the algorithms for (i) Dimensional tolerance analysis, (ii) Control of feed rate, spindle speed, depth of cut and cutting forces, (iii) Propagation of errors in multistage machining, and (iv) Vectorization of geometrical tolerances. Machining processes would include (i) Milling, (ii) Turning, and (iii) Drilling. Then the corresponding outputs are interpreted and analyzed to generate G and M codes. This study has shown how ANN can revolutionize NC machine manufacturing. A case study illustrates the effectiveness of the proposed method.