Application of Combination Forecasting Model in Aircraft Failure Rate Forecasting

Area of Science:

• Aerospace Engineering
• Data Science
• Reliability Engineering

Background:

• Effective aircraft failure rate prediction is crucial for maintenance planning, reliability, and flight safety.
• Existing models like ARIMA, grey Verhulst, and BP neural networks have limitations in handling complex failure data.
• Combining diverse models can leverage individual strengths for improved predictive performance.

Purpose of the Study:

• To develop and evaluate advanced combined forecasting models for aircraft failure rate prediction.
• To enhance the accuracy and reliability of aircraft equipment failure predictions.
• To provide a robust technical method for aircraft fault prediction and health management.

Main Methods:

• Utilized ARIMA, grey Verhulst, and BP neural network models individually.
• Developed three combined forecasting models using variable weight, sum of squares of errors, Shapley value, and IOWA operator methods.
• Evaluated model performance using metrics such as MAPE, RMSE, MAE, IA, TIC, EC, NSE, and Pearson test.

Main Results:

• Combined forecasting models outperformed single models in prediction precision.
• The IOWA operator-based combination model demonstrated superior evaluation index performance.
• The proposed combined model showed high accuracy, reliability, applicability, and stability compared to other models.

Conclusions:

• The developed combined forecasting models significantly improve aircraft failure rate prediction accuracy.
• The IOWA operator-based approach offers a superior method for weighting model contributions.
• This research provides a practical and valuable technical solution for aircraft fault prediction and safety enhancement.