全矢支持向量回归频谱预测方法

The Prediction Method of Frequency Spectrum Based on Full Vector Support Vector Regression

为了对机械设备进行故障类型和故障部位的准确预测,提出了全矢支持向量回归的频谱预测新方法.该方法使用全矢谱信息融合技术对同源双通道信号进行信息融合,采用支持向量回归对全矢谱特征参数进行预测,保证了训练数据信息的全面性以及预测结果的准确性.该方法对振动信号的频谱结构分布情况进行准确预测,从而为对机组的故障类型和故障部位预测奠定技术基础.通过对某1 000MW汽轮机轴振进行频谱预测验证结果表明,该方法在对振动信号频谱结构特征进行预测方面具有较高的预测准确性.

In order to predict the fault type and the fault position of rotating machinery more accurately, a new prediction method of frequency spectrum based on full vector support vector regression is proposed. The new method uses the full vector spectrum technology to merge the homologous double channel signal information, and uses support vector regression to predict the full vector spectrum characteristic parameters, which can en- sure the comprehensiveness of the training data and the accuracy of predict result. This method can forecast the frequency spectrum of the vibration signal accuracy and can then give the technical based for fault type and fault position predict. The experimental results of the frequency spectrum forecast of a 1000 MW steam turbine shaft show that this method can predict fully and accurately the spectrum＇ s structure of vibration signal.