An Intelligent Prediction Method of Aero-Engine Gas Path Performance Parameters

Liu, Lei; Jiang, Jipeng; Lü, Feng

doi:10.1007/s40032-021-00662-2

Cited by 5 publications

(1 citation statement)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are many methods that can be used to predict the health states of an aero-engine gas path system, such as the Kalman filter method [5,6], grey theory [7][8][9], neural network method [10,11], support vector machine [12,13], analytic hierarchy process [14,15], hidden Markov model [16,17], and expert knowledge [18][19][20]. A two-way kernel extreme learning machine was proposed to predict the health states of aero-engine gas path system by one parameter [21]. The literature [22] can denoise the collected health state parameters and predict the health states of a gas path system by using the denoising health state parameters.…”

Section: Introductionmentioning

confidence: 99%

Health State Prediction of Aero-Engine Gas Path System Considering Multiple Working Conditions Based on Time Domain Analysis and Belief Rule Base

et al. 2021

View full text Add to dashboard Cite

The gas path system is an important part of an aero-engine, whose health states can affect the security of the airplane. During the process of aircraft operation, the gas path system will have different working conditions over time, owing to the change of control parameters. However, the different working conditions which change the symmetry of the system will affect parameters of the health state prediction model for the gas path system. The symmetry of the system will also change. Therefore, it is important to consider the influence of variable working conditions when predicting the health states of gas path system. The accuracy of the health state prediction results of the gas path system will be low if the same evaluation standard is used for different working conditions. In addition, the monitoring data of the gas path system’s health state feature quantity is huge while the fault data which can reflect the health states of the gas path system are poor. Thus, it is difficult to establish a health state prediction model only by using the monitoring data of the gas path system. In order to avoid problems, this paper proposes a health state prediction model considering multiple working conditions based on time domain analysis and a belief rule base. First, working condition is divided by using time domain characteristics. Then, a belief rule base (BRB) theory-based health state prediction model is built, which can fuse expert knowledge and fault monitoring data to improve modeling accuracy. The reference value of the feature is given by the fuzzy C-means algorithm in a model. To decrease the uncertainty of expert knowledge, the covariance matrix adaptive evolution strategy (CMA-ES) is used as the optimization algorithm. Finally, a NASA public dataset without labels is used to verify the proposed health state model. The results show that the proposed health prediction model of a gas path system can accurately realize health state prediction under multiple working conditions.

show abstract