Surface roughness modeling and optimization of tungsten–copper alloys in micro-milling processes

• A roughness surface real-time monitoring system for micromilling processes was proposed.
• The z-component of the vibration signal and the feed rate were used as inputs of the model.
• An adaptive neuro-fuzzy inference system (ANFIS) model was implemented.
• A multiobjective genetic algorithm was used and a Pareto front was obtained.

Nowadays, the micrometric and nanometric dimensional precision of industrial components is a common feature of micro-milling manufacturing processes. Hence, great importance is given to such aspects as online metrology and real-time monitoring systems for accurate control of surface roughness and dimensional quality. A real-time monitoring system is proposed here to predict surface roughness with an estimation error of 9.5%, by using the vibration signal that is emitted during the milling process. In the experimental setup, the z-axis component vibration is measured using two different diameters under several cutting conditions. Then, an adaptive neuro-fuzzy inference system model is implemented for modeling surface roughness, yielding a high goodness of fit indices and a good generalization capability. Finally, the optimization process is carried out by considering two contradictory objectives: unit machining time and surface roughness. A multi-objective genetic algorithm is used to solve the optimization problem, obtaining a set of non-dominated solutions. Pareto front representation is a useful decision-making tool for operators and technicians in the micro-milling process. An example of the Pareto front utility-based approach that selects two points close to both extreme ends of the frontier is described in the paper. In the first case (point 1), machine time is of greater importance, and in the second case (point 2), importance is attached to surface roughness. In general terms, users can select different combinations, at all times moving along the Pareto front.