Breakage assessment of ship structures based on PCA-BOA-KNN model underwater explosions

Objective To address the issue of assessing structural breach damage in ships under underwater explosion, an breach prediction method based on the PCA-BOA-KNN model is established.

Methods First, finite element models for five-compartment and seven-compartment segments are constructed, and explosion simulation analysis is carried out for 21 sets of underwater explosion conditions. Subsequently, principal component analysis (PCA) is employed to reduce the dimensionality of the peak acceleration, peak velocity, peak displacement, peak stress and peak overpressure values, resulting in two principal features. Finally, the PCA results are integrated into a Bayesian optimization algorithm (BOA) K-Nearest Neighbors (KNN) model. The established breach prediction model is used to predict the breach conditions at different ship cross-sections under a set of conditions.

Results The results show that by using PCA to extract the first two factors, the cumulative contribution rate is 85.165%. Therefore, the first two factors can represent the primary information of the five features. The results obtained using the PCA-BOA-KNN breach prediction model are generally consistent with the simulation results.

Conclusion The proposed prediction model approach is effective for predicting ship structural breaches and has reference value for predicting breachs in ship structures with different principal dimensions.