Study On Long-term Performance And Prediction Of Pultruded FRP In Marine Environment

Fiber reinforced polymer(FRP)composites,have relatively higher strength-and stiffness-to-weight ratios and better corrosion resistance.In the harsh construction environments,FRPs have become an ideal alternative for traditional materials.In this regard,the long-term performance of FRPs subjected to environmental effects must be addressed so as to support the life-cycle design and construction of FRP structures.Starting from experiments and data analysis,this work studies the long-term performance behavior of Pultruded FRP Profile in the marine environment,explores the mechanism of long-term performance degradation,and adopted the advanced machine learning algorithm to predict the long-term performance of pultruded FRPs on the basis of large amounts of data.In the end,an online prediction platform was built to support the life-cycle design of FRP structures.The main work and achievements are as follow:(1)A literature review was conducted on the long-term mechanical performance of FRPs under eight types of environmental conditions,and the existing long-term performance prediction models were analyzed and summarized.The existing studies showed a relatively high discreteness in terms of the methodologies and reported results.It is also found that existing theoretical prediction models can only consider a limited number of influential factors,and there is still a lack of universally-applicable prediction model in the field.The degradation of FRPs in actual environment is resulted from the coupling effect of various environmental factors,and thus,the aging mechanism of FRPs is relatively complex,showing a high degree of multi-dimensional nonlinear characteristics.The machine learning algorithms are known to have the advantage of high accuracy in solving the multi-dimensional nonlinear problems,and thus,they are suitable for predicting the long-term performance of FRPs.(2)The natural weathering aging test on a real Island and accelerated aging test in the lab were carried out for pultruded FRP tubes.In particular,the island weathering tests were conducted on the Yongxing island in city of Sansha in the South China sea.The test specimens included two scales,the material(millimeter level)and the component(meter level)scales.Test results showed that the marine environment has little effect on the fiber-dominated properties of FRPs(such as tensile properties),while it has a greater impact on the resin-dominant and fiber-matrix-interface-dominant properties(such as bending properties).The performance degradation of the FRPs being addressed was found to be limited under the high-temperature seawater immersion,and thus,it could be anticipated that the FRPs would show a better corrosion resistance in the practical applications in real environments.(3)Three representative machine learning algorithms,including the artificial neural network algorithm,support vector regression algorithm and XGBoost decision tree algorithm,were analyzed and compared in terms of their capabilities in predicting the long-term tensile performance of pultruded FRPs.The XGBoost algorithm was finally selected to build the prediction model and this model showed excellent prediction relevance and accuracy.Based on the proposed XGBoost prediction model,a long-term performance prediction platform for pultruded FRPs was developed,providing a convenient,fast and reliable method for the field to obtain the long-term tensile properties of pultruded FRPs.The prediction accuracy of this platform has been validated through the test data obtained in this research and those test data in the existing literatures.An excellent accuracy of the platform was observed and thus,this platform can be used as a reference to the life-cycle design of pultruded FRP composites and structures.