A deep learning platform for evaluating energy loss parameter in engineering structures

Abstract

This paper proposes a new parameter to evaluate and identify changes in the condition of a structure. The energy consumption coefficient (C0) proposed in this paper can be used to evaluate changes in various types of structures. This study improves the model of Hooke’s traditional linear mechanical change using a nonlinear model to truly express the nature of structures in reality and adds to the current framework regarding changes in the mechanical properties of a material, according to the Kelvin-Voigt model (C0). The sensitivity of C0 is much higher than parameters proposed in similar researches and is suitable for many different types of bridge spans, such as simple bridge spans, concrete bridge spans, cable-stayed bridge spans and overpass spans. Based on a deep learning platform, C0 can identify mechanical changes over a structure’s operation time, while most of the previous researches followed Hooke’s law. In addition, to make the new parameter work effectively and accurately during evaluation, training is conducted on a deep learning platform by using the amplitude acceleration data of the structure. This manuscript was based on a large number of signals measuring the acceleration of vibration of some large bridges in Vietnam many years ago. The results show that the training process helps eliminate interference values and errors. The results demonstrate that C0 is highly sensitive for most types of structures in reality, while previous parameters were only significantly sensitive for specific types of structures. In the future, the coefficient C0 could be applied to many different types of structures. It will become a new and reliable parameter for the structural health monitoring of the complex working conditions of structures.