Abstract

Due to the complexity of the high-pressure jet assisted turning, knowledge, and prediction of the cutting forces are essential for the planning of machining operations for maximum productivity and quality. However, it is well known that during processing using this procedure there are difficulties in collecting data. It is required to establish an adequate model that would make it possible to predict the cutting force based on the input parameters. During machining to avoid difficulties in acquisition data, two models have developed based on fuzzy logic that will allow indirect monitoring of the cutting force. This research uses the improved fuzzy logic methods for modeling, whereby it can make predictions of the main cutting force according to the different input parameters. The contribution of this work reflected through the application of two innovative methods based on reducing the number of rules, which leads to better interpretability of models. First is the Mamdani with rule reduction method, and second is the Sugeno sub-clustering method based on the identification of the model structure, it comes down to finding the required number of rules by forming specific clusters. Both approaches differ by reducing the number of rules without affecting the accuracy of the models. The ability to predict the model determined by applying different statistical parameters. It concluded that Mamdani and Sugeno models give an approximate quality of the prediction. The resulting models also have an acceptable error to predict data that did not participate in their creation. Furthermore, obtained models can be used at the generalization stage where the cutting force information is required and where direct measurement is not possible.

中文翻译：

---

预测高压射流辅助切削中主切削力的模糊逻辑和子聚类方法

摘要

由于高压喷射辅助车削的复杂性，切削力的知识和预测对于规划加工操作以实现最大的生产率和质量至关重要。但是，众所周知，在使用此过程进行处理期间，很难收集数据。需要建立一个适当的模型，该模型将使得可以根据输入参数预测切削力。为了避免采集数据时遇到困难，在加工过程中，基于模糊逻辑开发了两个模型，这些模型可以间接监控切削力。本研究使用改进的模糊逻辑方法进行建模，从而可以根据不同的输入参数来预测主切削力。这项工作的贡献是通过在减少规则数量的基础上应用两种创新方法来体现的，从而提高了模型的可解释性。首先是带有规则约简方法的Mamdani，其次是基于模型结构识别的Sugeno子聚类方法，归结为通过形成特定的聚类来找到所需的规则数量。两种方法的不同之处在于减少了规则数量而不影响模型的准确性。通过应用不同的统计参数来预测模型的能力。结论是，Mamdani和Sugeno模型给出了近似的预测质量。结果模型还具有可接受的误差，可以预测未参与其创建的数据。此外，