Computational intelligence (CI) has been applied to grinding processes in order to perform more and more efficient operations. Among the techniques that compose the area of CI, it is possible to highlight the artificial neural networks (ANN) and the multiobjective optimization algorithms, such as the NSGA-II. These tools are useful to the computational modeling of several processes. However, until now, the hybrid use of these techniques has not been explored for external cylindrical grinding of superalloys. Filling this gap, this work had the objective of developing a methodology that uses a multilayer perceptron (MLP) ANN associated to the NSGA-II, in the form of an objective function. The method optimizes the grinding process of superalloys with the use of two grinding wheels, one conventional and another cBN superabrasive. The methodology consisted of three phases: the first one was of data collection through experiments for the construction of a dataset, containing input configurations of a grinding process related to output parameters that determine the quality of the machining. These parameters, in addition, are also objectives to be minimized when this problem is mathematically modeled. The second one dealt with the training and validation of an MLP ANN as the simulator of this grinding process, and finally, the third, dealt with the generation of optimized solutions by means of the NSGA-II associated with the MLP ANN already trained. Regarding the results, it could be demonstrated that the simulation does not present evidence of statistical differences when compared to real data. As far as optimization is concerned, the non-dominated solutions presented values, in several scenarios, consistent with the literature of the area. In this sense, this work brings as its main contribution a hybrid, scalable computational method that can be used as a tool for decision making. It is also concluded that it can be useful in the planning of economically efficient operations, since the simulation does not involve logistic costs.

计算智能（CI）已应用于研磨过程，以执行越来越高效的操作。在构成CI领域的技术中，有可能强调人工神经网络（ANN）和多目标优化算法，例如NSGA-II。这些工具对于几个过程的计算建模很有用。然而，直到现在，还没有探索将这些技术混合用于高温合金的外圆磨削。填补这一空白，这项工作的目的是开发一种方法，该方法以目标函数的形式使用与NSGA-II相关的多层感知器（MLP）ANN。该方法使用两个传统的cBN超级磨料砂轮来优化超级合金的磨削工艺。该方法包括三个阶段：第一个阶段是通过实验收集数据以构建数据集，其中包含与确定加工质量的输出参数相关的磨削过程的输入配置。此外，在对问题进行数学建模时，这些参数也是要最小化的目标。第二个涉及作为该磨削过程模拟器的MLP ANN的训练和验证，最后，第三个涉及与已经训练的MLP ANN相关的NSGA-II处理优化解决方案的生成。关于结果，可以证明，与真实数据相比，模拟没有提供统计差异的证据。就优化而言，非主导解决方案提供了价值，在几种情况下，与该地区的文献一致。从这个意义上讲，这项工作的主要贡献是可以用作决策工具的混合，可扩展的计算方法。还得出结论，由于模拟不涉及物流成本，因此它在经济高效的操作计划中可能很有用。