Ill-posed analysis of time-domain deconvolution identification of ice load on the shell structures of ice-going vessels

Load identification algorithms are crucial for the real-time assessment of structural safety, integrity, and durability in ice-going vessel monitoring systems. However, factors such as the ill-conditioning of the response matrix and noise in the measured ice-induced strains can severely degrade the inversion accuracy, thereby compromising the reliable judgment of structural safety. This paper investigates the ill-posed nature of the ice load identification system. A dynamic ice load identification model is first established based on the time-domain deconvolution method which can incorporate the dynamic effects of ice loads, mitigates solution anomalies caused by loading at unexpected locations, and offers broader applicability to various engineering scenarios. Using singular value decomposition (SVD) and Picard theory, a comprehensive analysis is conducted to examine the influencing factors and ill-posed characteristics of the identification system. Subsequently, a low-pass signal filter and a regularization operator are introduced to mitigate the ill-posedness. Finally, three typical ice load identification scenarios are simulated and an experimental validation is executed, and the effectiveness of the proposed mitigation strategies is evaluated. This study may provide supports for the reliability assessment of field measurements.