发动机传感器数据偏差检测子空间方法研究

Study on Subspace Model-based Approach to Engine Sensor Data Error Detection

检测了应用子空间方法检测具有线性关联的多通道信号偏差。方法原理是利用正常工况数据的主元分析建立主元-残差子空间模型,计算新数据在模型子空间的统计量,以是否超越阈值判别数据中是否包含传感器偏差。导出了数据偏差幅值与主元-残差子空间结构参数影响偏差检测性的量化关系,并用一种实际涡轮试验数据进行验证,结果表明发展的子空间方法可有效检测线性关联多通道信号中的加性传感器偏差。

This paper presents a subspace model based approach to sensor data error detection for gas turbine test.The normal subspace model is generated by principal component analysis(PCA)of multi-sensor data with linear correlation under normal regimes,and statistics thresholds in principal component subspace and residual subspace are quantitatively connected with error magnitude and subspace characteristics.Sensor data errors can be detected by comparing statistics thresholds with calculated values of new data.The effectiveness is demonstrated for practical gas turbine test data validation.