辅助动力装置控制系统传感器智能解析余度方法

针对辅助动力装置(APU)控制系统传感器故障,提出了一种基于协方差优化集成极限学习网络(COSELM)的传感器智能解析余度方法。该方法能够根据在线序列预测误差的最小方差来自适应更新单个在线序列极限学习机的权重系数,发挥和权衡各个学习模型的优势,通过提高模型算法的稳定性和泛化性,改善传感器智能解析余度的精度。通过在某辅助动力装置控制系统传感器解析余度的验证表明,提出的COSELM方法可以解决传感器在发生偏置故障时的信号重构问题,重构误差不超过1%,适用于不同辅助动力装置个体,为其提供可靠的解析余度。

Intelligent analytical redundancy method of control system sensors based on APU

For the sensors fault of auxiliary power unit (APU) control system, an intelligent analytical redundancy method of sensors was proposed based on online sequence extreme learning machine by improved covariance optimization algorithm. The covariance online sequence extreme learning machine (COSELM) algorithm method could adaptively update the weight coefficient of a single online sequence extreme learning machine according to the minimum variance of prediction error, exploit and weigh the advantages of each learning model. While increasing the stability and generalization of the model, the COSELM algorithm could improve the intelligent analytical redundancy accuracy of sensors. The simulation experiments using the sensors data of APU were carried out. The results indicated that the proposed COSELM method can solve the reconstruction problem when the sensors encountered the bias fault and the reconstruction error was less than 1%. Consequently, it is suitable for different engines to provide reliable analytical redundancy.