摘要
及时、准确地预测加工过程产品质量标准差σ,对于及时判断工序状态、调整加工过程因素,进而提高产品过程质量等具有重要意义。文章提出了一种基于加权最小二乘支持向量机(WLS-SVM)的时间序列预测新方法,该方法采用了结构风险最小化原则,较好地避免了人工神经网络等智能方法在小样本学习、预测过程中存在的过学习、泛化能力弱等缺点;并采用"重近轻远"的权重设置原则,提高了预测的精度。实验表明,采用该方法对产品过程质量标准差σ进行预测切实可行,对于产品过程质量控制具有重要意义。
To predict standard deviation a of product quality during process in time and accurately is important for estimating the state of process and adjusting the process factor so as to enhance the product process quality. A new time series prediction method based on the weighted least square support vector machine（WLS- SVM） is put forward. On the one hand, by using the structure risk minimum criterion, it can solve the small- batch learning problem better and avoid the disadvantages of artificial neural networks prediction such as over- training, weak normalization capability, etc. On the other hand, this proposed method is more accurate because it sets larger weight on the nearer sample but smaller weight on the farther. The experimental result shows that the metthod is practical and feasible for predicting standard deviation σ of product process quality and has magnificent significance for quality control of product process.
出处
《合肥工业大学学报（自然科学版）》
CAS
CSCD
北大核心
2013年第2期231-235,共5页
Journal of Hefei University of Technology：Natural Science
基金
国家自然科学基金资助项目(70672096)
合肥工业大学博士学位人员专项基金资助项目(2010HGBZ0302)
合肥工业大学科学研究发展基金资助项目(2010HGXJ0076
2011HGXJ1069)
关键词
标准差σ
加权最小二乘支持向量机
过程质量
预测
standard deviation σ
weighted least squares support vector machine（WLS-SVM）
processquality
prediction
