Abstract: Automobile pressing moulds had various types and complex structures, and the cost was hard to estimate, thus the quotation had great difficulty. By using fuzzy mathematics theory, the die cost prediction model was constructed combining fuzzy close-degree analysis with exponential smoothing prediction. To decrease subjective factors of die similarity judgment with fuzzy mathematics, a method to calculate fuzzy nearness combined with rough set was proposed. According to die cost information database, the knowledge was classified by this mothed, and die cost information decision table was created. Objective weight distribution was made for each element of die characteristic set by using rough set attribute importance. Based on subjective weight and objective weight of experts judgment, the comprehensive weights of each element in die characteristic set was determined, and perdiction accuracy was improved. Considering the different influence degree of each die characteristic on price, the empirical formula of adjusting coefficient λ in die cost prediction was improved. The reliability and rationality of proposed method were verified by a group of die samples.

Key words: die cost, fuzzy close-degree, rough set, weight distribution , exponential smoothing prediction

摘要： 汽车冲压模具种类繁多、结构复杂，影响其价格的因素多，难以准确估算成本，导致报价难度大。为此，应用模糊数学理论，将模糊贴近度和指数平滑预测相结合构建模具成本预测模型，为减少应用模糊数学进行模具相似性判断的主观因素，提出结合粗糙集理论计算模糊贴近度的方法。该方法首先根据模具成本信息数据库进行知识分类，构建模具成本信息决策表，应用粗糙集属性重要度对模具特征集合各元素进行客观权重分配，然后根据专家评判的主观权重与客观权重综合确定模具特征集合各元素的综合权重，从而提高预测精度。另外，考虑到模具各特征因素对其价格的影响程度不同，对模具成本预测中的调整系数λ的经验公式进行了改进，使其更加合理。通过一组模具样本实例验证了该方法的可靠性和合理性。

关键词: 模具成本, 模糊贴近度, 粗糙集, 权重分配, 指数平滑预测