ABSTRACT
Ice accretion on cables of bridges poses serious safety risks to the structure as well as to vehicular traffic on the bridge when the ice falls onto the road. This is a common concern for cable-stayed bridges located in Northern region and no effective ice build-up prevention tool has yet been designed. This work proposes to use a structural health monitoring approach to develop an ice accretion detection algorithm based on time series modeling of cable acceleration. The objective is to develop a model that is sensitive to small mass changes and robust to noise.
When ice accumulates on the cable, it significantly changes its linear mass, thus modifying the vibration signature. We propose to use an autoregressive (AR) model to capture the variations in frequency. The three first AR coefficients are extracted as damage sensitive features (DSF). The correlation between the AR coefficients and the amount and/or the rate of ice accumulation is studied to formulate a predictive model for ice accretion.
The analysis is based on the assumption that the stiffness of the cable does not change significantly during ice formation, thus changes in the AR coefficients are attributed to changes primarily in the mass of the cable. An analytical formulation was first developed showing the relationship between changes in AR coefficients and changes in cable mass. It was determined that fitting a power law function leads to minimum error.
Then the algorithm was tested on numerically generated signals. The structure was modeled as a linear dynamic single degree of freedom system to generate response accelerations using computer simulation. After processing the data and applying the AR model, we used regression analysis methods to observe the behavior of the features extracted. A strong correlation is observed between the increase in mass and the first AR coefficient, which is consistent with the analytical result derived.
Finally, a wind tunnel experiment on a cable with gradually increasing ice formation was conducted at the Denmark Technical University/Force Climatic Wind Tunnel facility in Lyngby, Denmark and the algorithm was applied to the collected signals. Results from numerical simulations and experiments are in excellent agreement. The DSF extracted from the models are significantly correlated with the mass variation of the system, regardless of the level of noise induced by the experiment conditions. These initial results are very encouraging and require further investigation and experiments, especially for larger quantity of ice accretion. Ultimately, field experimentation will be necessary to affirm the validity of the algorithm.
