关键词: 数据驱动, 快速预测, 航线运维, 大型客机

Abstract: To address the predictive and condition-based maintenance requirements for the airframe structures of large passenger aircraft, a multi-source data fusion-driven system for line maintenance has been developed. This system integrates engineering data from the type development phase, including flight test-measured loads, full-aircraft finite element internal forces, structural fatigue damage assessment, airworthiness limitation items, and their associated maintenance requirements. By employing methods such as data cleaning, sample reconstruction, data modeling, damage accumulation analysis, and maintenance customization optimization, the system enables comprehensive condition-based maintenance support. Through application on a specific domestic large passenger aircraft model, it has resolved key challenges in data modeling accuracy and prediction efficiency for airframe structure maintenance. The system achieves rapid single-flight prediction of "flight parameters-loads-stress-damage" for critical wing regions. Based on comparison between theoretical damage per flight and actual damage values at 95% confidence level, preliminary estimates indicate that the repeat inspection interval for high-frequency eddy current inspections can be extended as follows: from 5,000 to 8,091 flight cycles for stringer hole edges at the skin-stringer splice details of the lower wing panel, from 5,500 to 8,900 flight cycles for skin hole edges.

Key words: data-driven, rapid prediction, line maintenance, large passenger aircraft