Comparison results (MSE) of different models.

To make the puzzle of aero-engines complete, understanding the law of the compressor geometric variable system is a vital part. Modeling all aspects of aero-engines quickly has been a continuous area of research since the advent of artificial intelligence (AI). However, diagnosing or predicting faults is an ancient adage, and it is vital to explore key system forecast research, particularly since traditional forecasting techniques do not account for future information of non-target parameters. In this article, based on the feasibility of forecasting the compressor geometric variable system, an enhanced ConvNeXt model utilizing the Sliding Window Algorithm mechanism is proposed. And this method takes into account the future information of non-target parameters. With the novel strategy, the issue of the forecast’s error increasing with forecast length is alleviated. As a result, in a particular condition, the error we obtained only accounts for 20.07% of that of the standard forecast approach. Additionally, it is verified that this method can be applied to various aero-engines. Finally, experiments on several aero-engine states involving the transition state and the steady state are conducted to strengthen the plausibility and credibility of our theories. It should be noted that the foundation of each experiment is data from actual flights.若要全面掌握航空发动机的整体运行特性，厘清压气机几何可变系统（Compressor Geometric Variable System）的运行规律是至关重要的一环。自人工智能（Artificial Intelligence, AI）问世以来，快速构建航空发动机全维度模型始终是研究领域的持续热点。尽管故障诊断与预测技术已有较长的发展历史，但开展核心系统预测研究仍具备重要价值——尤其是传统预测方法无法兼顾非目标参数的未来信息。本文基于压气机几何可变系统预测的可行性，提出了一种融合滑动窗口算法（Sliding Window Algorithm）机制的改进型ConvNeXt模型，该方法可充分考虑非目标参数的未来信息。借助该创新策略，预测误差随预测长度增加而增大的问题得到有效缓解。实验结果显示，在特定工况下，本文方法的预测误差仅为标准预测方法的20.07%。此外，验证表明该方法可适配多款不同型号的航空发动机。最后，本文针对包括过渡工况与稳态工况在内的多种航空发动机运行状态开展实验，以验证理论的合理性与可靠性；需特别说明，所有实验数据均来源于实际飞行采集的真实数据。