Analytical redundancy modeling of onboard sensors for engine parameter estimation using flight data

In the field of fault diagnosis and engine functional health management, onboard sensor analytical redundancy modeling serves as a critical technology. However, the modeling of analytical redundancy models for engine parameters is challenged by hardware constraints and the demand for high precision. This paper proposes a modeling method for Onboard Sensor Analytical Redundancy Model (OSARM) based on flight data, which combines the advantages of high precision, fast computation, and small storage footprint. The steady-state portion is established using the Kullback-Leibler (KL) Divergence Minimization K-Means Algorithm (KLDK-Means), with the baseline model of the engine obtained through a Physics-Based Model (PBM). Considering the impact of inlet operating conditions on the distribution of steady-state data, the accuracy of the steady-state baseline model is further improved by constructing a Modified Baseline Model (MBM). The dynamic modeling portion innovatively introduces the concept of nonlinear levels and employs the Nonlinear Levels Seeker Optimization Algorithm (NL-SOA) to construct a dynamic model, which improves upon traditional dynamic error optimization methods. The proposed method is validated using real flight data. The results show that compared to the Hammerstein model, OSARM achieved a reduction in maximum error of 32.4%. The modeling approach presented in this paper meets the requirements for precision modeling while satisfying the need for fast computation and small storage footprint.