The on-board sensor fault detection and isolation (FDI) system is essential to guarantee the reliability and safety of an aero engine. In this paper, a novel online sequential extreme learning machine with memory principle (MOS-ELM) is proposed for detecting, isolating, and reconstructing the fault sensor signal of aero engines. In many practical online applications, the sequentially coming data chunk usually possesses a characteristic of timeliness, and the overdue training data may mislead the subsequent learning process. The proposed MOS-ELM can improve the training process by introducing the concept of memory principle into the online sequential extreme learning machine (OS-ELM) to tackle the timeliness of the data chunk. Simulations on some time series problems and some benchmark databases show that MOS-ELM performs better in generalization performance, stability, and prediction accuracy than OS-ELM. The experiment results of the MOS-ELM-based sensor fault diagnosis system also verify the excellent generalization performance of MOS-ELM and indicate the effectiveness and feasibility of the developed diagnosis system.

The Kalman filter is widely utilized for gas turbine health monitoring due to its simplicity, robustness, and suitability for real-time implementations. The most common Kalman filter for linear systems is linearized Kalman filter, and for nonlinear systems are extended Kalman filter and unscented Kalman filter. These algorithms have proven their capabilities to estimate gas turbine performance variations with a good accuracy, and the studies are done provided that all sensor measurements are available. In this paper, a nonlinear fusion approach with consistent diagnostic mechanism based on unscented Kalman filter is proposed, especially for gas turbine performance monitoring in the case of sensor failure. The architecture of fusion method comprises a set of local unscented Kalman filters and an information mixer. The local unscented Kalman filters are utilized to estimate health parameters of various component combinations, and the results are then transferred to the mixer for the integrated estimation of global health state in fusion structure. The consistent fault diagnosis and isolation logic is designed based on the fusion architecture and combined with the fusing unscented Kalman filter, called an improved fusing unscented Kalman filter. A systematic comparison of the generic linearized Kalman filter, extended Kalman filter, and unscented Kalman filter to their fusion filter kinds is presented for engine health estimation of gradual deterioration and abrupt fault. The studies show that the fusing unscented Kalman filter evidently outperforms the fusing linearized Kalman filter and fusing extended Kalman filter, while the fusing Kalman filters have slightly better estimation accuracy than the basic Kalman filters. In addition, the proposed methodology can reach the reliable performance monitoring with measurement uncertainty while the conventional Kalman filters collapse.

A long-term gas-path fault diagnosis and its rapid prototype system are presented for on-line monitoring of a gas turbine engine. Toward this end, a nonlinear hybrid model-based performance estimation and abnormal detection method are proposed in this paper. An adaptive extended Kalman particle filter (AEKPF) estimator is developed and used to real time estimate engine health parameters, which depict gas turbine performance degradation condition. The health parameter estimators are then pushed into a buffer memory and for periodical renewing baseline model (BM) performance, and the BM is utilized to detect engine anomaly over its life course. The threshold in abnormal detection schemes is adapted to the modeling errors during the engine lifetime. The rapid prototyping system is designed and built up based on the National Instrument (NI) CompactRIO (CRIO) for evaluating gas turbine engine performance estimation and anomaly detection. A number of experiments are carried out to demonstrate the advantages of the proposed abnormal detection scheme and effectiveness of the designed rapid prototype system to the problem of gas turbine life cycle anomaly detection.

Extreme learning machine (ELM) owns the advantages of less computational efforts and simple topology with single-hidden layer structure. However, the performance of plain ELM is sensitive to the input weights, bias, and the number of hidden neurons; and the former two are randomly generated. This paper develops a restricted Boltzmann strategy combined with Moore-Penrose generalized inverse to learn topological parameters in both input and output layers. A novel extreme learning model based on the restricted Boltzmann ELM, constructs a feature mapping and recursively tune the weights between input neurons and hidden neurons. The contribution of this paper is to provide a simple ELM topological network to handle low dimensionality problem with the merit of better accuracy and stability. The proposed methodology is evaluated on University of California Irvine (UCI) benchmark datasets for classification issue, and then extended to gas path fault diagnosis for a turbofan engine. The experimental results confirm the superiority to plain ELM.