Application of Adaptive Network based Fuzzy Inference System on Predicting Grinding Surface Quality

[[abstract]]磨削，是切削加工中，常用的精加工方法。而在磨削加工的過程中，砂輪轉速、進給速度、砂輪切深、砂輪磨粒大小等參數往往是影響磨削後工件品質的主要變因。在傳統的磨削加工中，常常參考技術手冊及機台操作者本身的經驗，利用嘗試錯誤法來選擇這些參數。也因為如此，每一批次磨後工件的表面品質，往往因為操作者的不同，而變的不易掌控。因此，再要求提升品質的前提下，建立一個準確有效的加工模型，以便在得到加工量及磨後品質的同時，即可提供一組良好的加工參數，為控制工件品質與加工效率的一個好方法。 本研究目的為建立加工參數與表面品質間的準確加工模型。利用適應性類神經模糊系統(ANFIS)，建立KG砂輪磨削SKD11模具鋼表面品質的預測模型，用磨削加工參數做為輸入參數，磨後工件的表面品質當作輸出目標值，並依據實際的加工實驗的量測數據來建購加工預測模型。經由網路的訓練，預測模型對表面粗糙度的誤差值小於1.2%，而對表面微硬度的誤差值小於0.31%。[[abstract]]Grinding is commonly used precision work method in the machining. But in grinding process, grinding parameter, such as the grinding wheel rotational speed, feed rate, depth of cut, grinding wheel grit size often is the influence of work piece quality. In the tradition grinding processing, frequently refers to the technical manual and the operator itself experience, is choosing these parameters by try-and-error method. As this reason, each period of work piece surface quality, often changes because of operator''s difference, and don’t easily control. Therefore, in order to improve quality, it is a good method to establish an accurate effective processing model, provides group of good processing parameters. The model help us to controls the work piece quality and processing efficiency. To establish model with processing parameter input and surface quality output is the goal of this research. Using the Adaptive Network based Fuzzy Inference System (ANFIS), establishes the KG grinding wheel to grind the SKD11 molding steel model to forecast surface quality. The processing forecast model which with grinding parameter as the input parameter, work piece surface quality as the output target value, constructs by measure the data based on the actual processing experiment. By way of the network training, forecast the model is smaller than 1.2% to the surface roughness erroneous value, but is smaller than 0.3% to the surface micro-hardness erroneous value