Engine life prediction method based on multi-head attention

By combining multi-head attention mechanisms and long short-term memory networks, the problem of sensor contribution differences and uncertainties in the prediction of the remaining service life of aero-engines is solved, achieving accurate multi-dimensional feature capture and uncertainty quantification, thereby improving the credibility and accuracy of the prediction results.

CN122113574APending Publication Date: 2026-05-29HEBEI UNIV OF TECH
0 Cites 0 Cited by

Patent Information

Application Number
CN202610041436.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-01-13
Publication Date
2026-05-29

AI Technical Summary

Technical Problem

Existing methods for predicting the remaining service life of aero-engines suffer from several issues in data-driven approaches, including differences in sensor contributions, neglect of multidimensional correlations, and insufficient quantification of data noise and uncertainty. These issues result in unreliable prediction results and make it difficult to provide a reliable reference for maintenance decisions.

Method used

A multi-head attention mechanism is used to weight the spatiotemporal dimensions. Combined with long short-term memory networks and Monte Carlo regularization methods, the uncertainty of prediction is quantified. Redundant features are removed through data visualization analysis, and the kernel density estimation method is used to quantify the uncertainty of prediction.

Benefits of technology

It accurately captures multi-dimensional key degradation features, reduces random errors, quantifies uncertainty, provides probability density distribution and confidence interval, improves the credibility and accuracy of prediction results, adapts to single-condition multi-failure modes, and takes into account engineering practicality.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to the technical field of engine life prediction, in particular to an engine life prediction method based on multi-head attention, and the method comprises the steps: carrying out the normalization processing of the original data of an aero-engine sensor, and removing redundant features; space-time two-dimensional self-adaptive weighting is achieved through a multi-head attention mechanism, and key degradation information is enhanced; inputting the weighted data into a long short-term memory network to learn a long-term dependency relationship, and completing the preliminary prediction of the remaining service life; and finally, combining Monte Carlo regularization and kernel density estimation to generate a plurality of prediction values and quantify uncertainty, and taking a mean value as a final prediction result. Experimental verification shows that the method effectively highlights key degradation characteristics, reduces the uncertainty influence, remarkably improves the prediction performance in a single-working-condition multi-fault mode, and provides reliable data support for preventive maintenance of an aero-engine.
Need to check novelty before this filing date? Find Prior Art