基于蜻蜓算法和最小二乘向量机的小批量生产质量预测

Quality Prediction of Small Batch Products Based on LSSVM with Dragonfly Algorithm

摘要:

本文采用蜻蜓算法（DA）和最小二乘支持向量机（LSSVM）的方法，解决生产过程中小批量产品在质量预测方面的问题。首先以汽车变速箱轴承内圈孔直径的尺寸作为预测数据，连续观测12个单位时间，并记录每个单位时间轴承内圈孔直径的尺寸数据，进行归一化处理；其次采用LSSVM对变速箱轴承内圈孔直径加工过程变化进行量化分析，并采用蜻蜓算法优化LSSVM参数；最后将DA-LSSVM综合方法与多种预测模型进行对比分析。结果表明，DA-LSSVM方法可以提高预测模型的训练预测精度，缩短训练时间。

Abstract:

In this paper, the Dragonfly Algorithm (DA) and the least Squares support vector machine (LSSVM) are used to solve the problem of quality prediction of small batch products. Firstly, the size of inner ring hole diameter of automobile gearbox bearing is used as the forecast data, and continuous observation of 12 unit time and recording of dimensional data of the inner ring hole diameter of each unit time bearing for normalization treatment. Secondly, LSSVM is used to quantitatively analyze the change of the diameter of the inner ring hole of gearbox bearing, and the LSSVM parameter is optimized by Dragonfly algorithm. Finally, the DA-LSSVM synthesis method is compared with many kinds of forecasting models. The results show that the DA-LSSVM method can improve the precision of prediction model and shorten the training time.