Abstract: Aiming at the characteristic of massive information and integration of qualitative and quantitative knowledge in manufacturing process, the modeling thought by merging knowledge modeling with data mining was proposed. Based on rough set theory, a function relation of knowledge was built, and a new type of rough set model as well as its prediction method was constructed based on indiscernibility-function relations. Thus the machined surface roughness was predicted. The existed knowledge was embeded in data mining model by proposed model, which made the information classification more accurate, the knowledge contained in decision rules more rich, predicted accuracy higher and predicted range wider. Compared with other prediction models, the proposed model only needed the existed data and knowledge, and extra model structure and parameters of model were not designed. Experimental results showed that the proposed model had good exhibition on predicted effectiveness and predicted accuracy.

Key words: roughness prediction, data mining, knowledge modeling, rough sets, indiscernibility—function relation

摘要： 针对当前制造过程海量信息和定性定量知识并存的特性，提出知识建模和数据挖掘技术相融合的建模思想。基于粗糙集模型，首次建立知识的粗糙集函数关系，并构建基于“不可分辨—函数”关系的新型粗糙集模型及预测方法,用以预测加工表面粗糙度。新模型将已有知识嵌入到数据挖掘模型中，其信息划分更精确，获取的决策规则蕴含的知识更丰富，故预测精度更高，预测范围更广。与其他预测模型相比，所建模型仅利用已有知识和信息，不需要建模者额外设计和设定模型的结构形式和参数。实验结果也表明，所建模型在预测有效性和预测精度上均有较好表现。

关键词: 粗糙度预测, 数据挖掘, 知识建模, 粗糙集, 不可分辨—函数关系