Today, machining processes are still not completely understood, due to the complexity of their dynamics and the number of factors involved. One of the major challenges in precision manufacturing is to optimize the surface finishing and the geometric accuracy of machined parts. The optimal settings are usually defined based on "trial and error" and depends on the knowledge level of machine operators in a factory. In order to automate this optimization task, the process model needs to be identified. This paper studies the correlations between the cutting conditions in a hard turning process and the quality (surface finishing and geometric accuracy) of this process. Two techniques will be studied and compared, for the model identification: Response Surface Method (RSM) and an algorithm combining ANN and fuzzy logics. Compared to recent studies on the topic, this paper will consider the validation of the identified models with tool wear, and will also use various measurements, such that physical phenomena of different types can be taken into account. Real data from a hard tuning process with a CNC lathe will be used for model validation.
展开▼
