Machine learning-based method for analyzing stress distribution in a ship.

Area of Science:

• Naval Architecture and Marine Engineering
• Computational Science
• Data Science

Background:

• Ship structural health monitoring (SHM) is crucial for operational safety.
• Analyzing stress distribution via pressure sensors enhances SHM performance and prevents failures.
• Understanding stress data is key to improving ship integrity.

Purpose of the Study:

• To develop a machine learning computational method for analyzing crane ship stress distribution.
• To infer stress values using data from multiple pressure sensors.
• To identify critical sensors for effective stress monitoring.

Main Methods:

• Utilized extreme gradient boosting (XGBoost) to assess relationships between pressure sensor data.
• Employed multilayer perceptron (MLP) to build regression models for stress prediction.
• Developed optimal MLP models for individual sensors to infer stress from related sensor data.

Main Results:

• MLP regression models showed strong fitting performance on training and test datasets.
• The method successfully identified key pressure sensors crucial for stress recovery.
• Ablation tests confirmed the robustness and effectiveness of the proposed approach.

Conclusions:

• The machine learning method provides effective stress distribution analysis for ship SHM.
• Identifying key sensors enhances the interpretability and efficiency of monitoring systems.
• The approach offers valuable insights into SHM mechanisms and improves structural safety.