A boom damage prediction framework of wheeled cranes combining hybrid features of acceleration and Gaussian process regression

Abstract

In a real-life environment, construction machinery and other large equipment are subjected to dynamic loads that cause structural fatigue or failure. Therefore, it is necessary to assess the fatigue damage accurately or predict its structural damage in real-time. However, testing and structural damage monitoring for large boom structures are inconvenient and difficult. In this study, we solve this problem by proposing a boom damage prediction framework for wheeled cranes. The framework design highlights three aspects: (1) A limited number of sensors usage, (2) Extraction of hybrid features, and (3) Model transferable. Time/frequency domain features and fatigue features of finite accelerometers are captured, and more efficient feature sets are obtained by feature mining to prevent model overfitting. The Gaussian process regression (GPR) model is applied to obtain more satisfactory point and interval prediction results, thus improving the accuracy of the model. In addition, the potential of the model transferability is also verified by actual boom damage data, which demonstrates how it can serve real engineering applications.