Scholarlink

In order to digitally assist aero-engine ground test so as to reduce the number and cost of the tests, this paper suggests a hybrid model connecting and combining Convolutional Neural Network (CNN), Bidirectional Long Short-Term Memory network (BiLSTM), and attention mechanisms for aero-engine thrust estimation during ground test, which achieved feature extraction from sensor data, as well as time series modeling and an accurate prediction whose mean absolute percentage error is 2.58%. A comparison of the CNN-BiLSTM-Attention hybrid model to the single deep learning models, such as Long Short-Term Memory network and Temporal Convolutional Network, reveals a significant reduction in root mean square error, with a 23.81% and 39.62% decrease, respectively. This model is of great value in engineering applications.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Study on Aero-Engine Thrust Prediction During Ground Test Using CNN-BiLSTM-Attention

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Yuze Jiang,
  • Yang Wang,
  • Su Zhao,
  • Chen Wang

摘要

In order to digitally assist aero-engine ground test so as to reduce the number and cost of the tests, this paper suggests a hybrid model connecting and combining Convolutional Neural Network (CNN), Bidirectional Long Short-Term Memory network (BiLSTM), and attention mechanisms for aero-engine thrust estimation during ground test, which achieved feature extraction from sensor data, as well as time series modeling and an accurate prediction whose mean absolute percentage error is 2.58%. A comparison of the CNN-BiLSTM-Attention hybrid model to the single deep learning models, such as Long Short-Term Memory network and Temporal Convolutional Network, reveals a significant reduction in root mean square error, with a 23.81% and 39.62% decrease, respectively. This model is of great value in engineering applications.