Enhancing Aircraft Safety through Advanced Engine Health Monitoring with Long Short-Term Memory.

Area of Science:

• Artificial Intelligence
• Aerospace Engineering
• Mechanical Engineering

Background:

• Predictive maintenance is essential for managing costs and safety in industries like aviation.
• Engine sensor data is key to assessing wear and tear for proactive maintenance.
• Accurate prediction of engine lifespan optimizes operational efficiency and safety.

Purpose of the Study:

• To forecast the remaining operational lifespan of aircraft engines.
• To evaluate the effectiveness of a Long Short-Term Memory (LSTM) architecture for this task.
• To benchmark LSTM performance against other predictive maintenance methodologies.

Main Methods:

• Utilized a Long Short-Term Memory (LSTM) neural network architecture.
• Employed the NASA Turbofan Engine Corruption Simulation dataset for model training and evaluation.
• Compared LSTM performance with alternative predictive maintenance techniques.

Main Results:

• The LSTM model achieved a classification accuracy of 98.916%.
• The model demonstrated a low mean average absolute error of 1.284%.
• LSTM outperformed alternative methods in predicting engine remaining useful life.

Conclusions:

• LSTM is a highly effective deep learning model for aircraft engine predictive maintenance.
• The developed model offers significant improvements in accuracy and error reduction.
• This research supports enhanced safety and operational efficiency in aviation through advanced prognostics.