CAJ | 학술논문

为了解决柴油机故障预测中大样本、非线性以及高维数据的数据预测问题，避免以向量输入带来的结构信息丢失和数据相关性被破坏等现象，结合支持向量机（SVM ）的学习框架和交替投影的思想，研究基于在线支持张量机（OSTM ）的柴油机故障预测算法和流程，并以测试精度、学习时间和均方根误差作为评价指标，利用远程监测系统采集的数据，分别应用在线支持向量机（OSVM ）和OSTM 进行故障预测和分析。结果表明，与OSVM 方法相比，OSTM方法测试精度较高，学习时间大幅缩短，预测模型的收敛速度较快，能有效在线预测柴油机故障。
위료해결시유궤고장예측중대양본、비선성이급고유수거적수거예측문제，피면이향량수입대래적결구신식주실화수거상관성피파배등현상，결합지지향량궤（SVM ）적학습광가화교체투영적사상，연구기우재선지지장량궤（OSTM ）적시유궤고장예측산법화류정，병이측시정도、학습시간화균방근오차작위평개지표，이용원정감측계통채집적수거，분별응용재선지지향량궤（OSVM ）화OSTM 진행고장예측화분석。결과표명，여OSVM 방법상비，OSTM방법측시정도교고，학습시간대폭축단，예측모형적수렴속도교쾌，능유효재선예측시유궤고장。
To resolve the prediction problems with giant sample size ,nonlinear and high dimensional data for diesel engines ,this paper ,aided by the framework of support vector machine (SVM ) and the alternating projection method ,studied the algorithm and process of diesel engine failure prediction on the basis of online support tensor machine (OSTM ) to avoid loss of structural information and damage to data correlation resulting from vector input .Prediction accuracy ,learning time and mean square er‐ror (MSE) were employed as evaluation indicators ,and data collected by distant monitoring system were used in diesel engine failure prediction on the basis of OSVM and OSTM ,respectively .The re‐sults show that compared with OSVM ,OSTM is more accurate in failure prediction ,boasting less learning time ,higher convergence speed and greater efficiency .