微型电脑应用

传统单一模型对复杂生产过程的质量预测存在精度低、泛化能力弱的问题。提出一种联合主成分分析(PCA)、卡尔曼滤波(KF)和支撑向量机(SVM)的组合模型实现对复杂生产过程的高精度预测。利用PCA对质量特性数据进行滤波分析实现噪声干扰抑制，将其作为SVM和KF模型的输入，分别建立SVM和KF预测模型实现质量预测，采用Stacking集成学习算法实现对2种预测结果的综合，从而获得最终预测结果。以铸铁生产过程中的化学成分预测为例进行试验验证，与传统单一模型的预测性能对比，验证了所提出的方法的有效性。

The traditional single models have the problems of low precision and weak generalization ability for the quality prediction of complex production process. A combined model combining principal component analysis(PCA), Kalman filter(KF) and support vector machine(SVM) is proposed to achieve high-precision prediction of complex production process. PCA is used to filter and analyze the quality characteristic data to achieve noise interference suppression. It is used as input to the SVM and KF models, and SVM and KF prediction models are established to achieve quality prediction.The Stacking ensemble learning algorithm is used to synthesize the two prediction results to obtain the final prediction result. Taking the prediction of chemical composition in the production process of cast iron as an example, the effectiveness of the proposed method is verified by comparing the prediction performance with the traditional single model.