Comparison of Machine Learning Algorithms for Structure State Prediction in Operational Load Monitoring

Area of Science:

• Aerospace Engineering
• Structural Health Monitoring
• Artificial Intelligence

Background:

• Operational load monitoring is crucial for predicting the remaining usability of aerospace structures.
• Maximizing the in-service life of aircraft components requires accurate safety assessments.
• Strain sensors are used to monitor structural integrity, but high sensor counts can be complex due to variable loads.

Purpose of the Study:

• To implement and compare machine learning algorithms for predicting the safety factor of structures.
• To assess the effectiveness of AI in reducing the number of strain sensors required for load monitoring.
• To evaluate the accuracy and computational time of different AI methods for real-time applications.

Main Methods:

• Adaptive Neuro-Fuzzy Inference Systems (ANFIS), Support-Vector Machines (SVM), and Gaussian Processes for Machine Learning (GPML) were trained using simulation data.
• Algorithm effectiveness was validated with experimental data.
• Performance was compared against Artificial Neural Networks (ANN) for sensor reduction in load monitoring.

Main Results:

• The study provides a numerical comparison of accuracy and computational time for ANFIS, SVM, GPML, and ANN.
• AI algorithms demonstrated effectiveness in predicting structural safety factors with reduced sensor data.
• The research quantifies the trade-offs between different AI methods for operational load monitoring.

Conclusions:

• Machine learning and AI offer efficient solutions for operational load monitoring in the aerospace industry.
• These methods can significantly reduce the number of required sensors while maintaining accurate safety predictions.
• The findings support the real-time application of AI for enhancing structural health monitoring and component lifespan.