基于PCA SVM的多品种小批量产品质量预测方法研究

Research on qualitative forecasting for diversified small-quantity production based on PCA-SVM

摘要:

对于少样本的多品种小批量产品复杂生产系统,生产过程中影响质量的因素有很多。以往研究忽略这些影响因素对质量结果的直接影响导致预测精度受限。本文将主成分分析与支持向量机结合应用到多品种小批量产品的质量预测。首先确定多品种小批量产品生产过程中的定量影响因素,并将其作为初始影响因素集；然后利用主成分分析方法降低因素集的维度,同时提取关键主成分；最后将关键主成分作为影响因素集并建立针对于多品种小批量生产的支持向量机质量等级预测模型。算例分析表明,与传统的支持向量机分类模型相比,主成分分析与支持向量机结合的模型预测准确率及稳定性均有显著提高,说明模型具有更好的预测性能。

Abstract:

As a complicated production system with small sample, diversified small-quantity production have many influencing factors in the process of manufacture. Ignoring the direct effect on quality results of this influencing factors cause a limitation of forecast accuracy. A method combined the principal component analysis (PCA) with the support vector machine (SVM) is applied to the qualitative forecasting for diversified small-quantity production in this paper. Firstly, quantifiable influencing factors in the process of manufacture were selected as the initial influencing factor set；Secondly, the dimension of the initial influencing factor set was reduced by the method of PCA to simplify operations；Lastly, a SVM regression model for diversified small-quantity production was constructed by using the key principal components as the influencing factor set. Case study results illustrate that, compared with the SVM model, both the accuracy rate and the stability were improved. It indicates that the PCA-SVM model has better predictive performance.