针对冶炼过程喷溅特征提取及喷溅预测困难的问题，提出基于小波包变换与主成分分析的优化参数模型的支持向量机喷溅预测方法。该方法经小波包变换将冶炼喷溅的噪声和氧枪振动信号分解为不同频带的信号。由于不同频带的信号出现相互干扰和堆叠，因此通过主成分分析将频带能量降维分离成不同频带，进而将这些处理后的信号作为喷溅特征向量。对支持向量机模型参数(C、g)进行遗传算法优化，通过支持向量机对喷溅的分类及预测，验证了该方法的有效性。实验结果表明：经小波包变换和主成分分析获得的特征信号能够准确地反应喷溅特征，提出的支持向量机方法具有较好的分类性能，喷溅预测准确率较高。
Aiming at the problem of splashing feature extraction and splash prediction in refining process,an opti-mal support vector machine method of splash prediction was proposed based on wavelet packet and principal com-ponent analysis.Firstly,wavelet packet transform decomposed the noise and vibration during the process into dif-ferent signal frequency brands.However,the signals with different brands would interfere with each other and stack.Therefore,the different bands energy can be decomposed into different lower dimension frequency bands by principal component analysis,and these processed signals can be regarded as the characteristic vector.Then,ge-netic algorithm is used to optimize the model parameters of support vector machine.The results of prediction and classification in splashing are obtained with support vector machines,and effectiveness of the proposed method is verified.The experimental results show that the characteristic signals obtained by wavelet packet transform and principal component analysis can accurately reflect the character of splash,the proposed method of support vector machine has good classification performance of splash,and the accuracy of splash prediction is high.
参考文献
- 下载量()
- 访问量()
- 您的评分：
-
10%
-
20%
-
30%
-
40%
-
50%