Frequency-based Delamination Detection In Stiffened Fiber Reinforced Polymer Plates

FRP composite materials which have advantages such as high strength/stiffness ratio,ease of shaping,excellent fatigue resistance,good corrosion resistance,etc.,have been widely used in aerospace,civil,energy,transportation,and marine engineering.One of the main application forms of FRP is stiffened plate or shell structures,in which the stiffeners are added to the main structures(laminated plates or shells)through certain way to greatly improve the bearing capacity of the FRP structure.Stiffened FRP plates have become a basic as well as common type of FRP structures applied on modern aircraft and spacecraft.Delaminations or debonding may occur between the stiffeners and the base plates,or internally embedded within the base plates in such stiffened FRP laminates either during manufacturing or in service caused by the impact from sandstone,hail,bullets or maintenance tools.Such delamination damage will greatly weaken the carrying capacity of the stiffened FRP plates,or even lead to structural failure causing big loss of life and properties.Therefore,it is significant to diagnose and identify the delaminations in the widely applied FRP stiffened plate structures.Present work focus on assess delamination damage in stiffened FRP plates through the changes in frequencies after damage occurs.Two inverse algorithms,namely artificial neural network(ANN)and genetic algorithm,were developed to predict the location and size of delaminations in the stiffened FRP plates using a series of frequency shifts.The efficiency and accuracy of the frequency-based detection algorithms are validated both numerically and experimentally.Firstly,the stiffened plates were modelled using finite element software and the two types of delminations were considered(delamination between stiffeners and base plate and internal delamination in base plate).The numerical frequency shifts from FEM were then used to verify the two reverse algorithm.As for the experiments,undamaged and delaminated stiffened plate specimens were manufactured,and modal testing was conducted to extract the frequencies and mode shapes.The measured frequencies of the stiffened plate specimens were input into the inverse algorithms for validation purpose.To study the robustness of the inverse algorithms,different levels and different sample sizes of noise were added to the numerical frequencies to simulate the experimental errors.The results of numerical verification show that the two kinds of reverse detection algorithms can successfully identify the two types of delaminations with good accuracy in predicting location and size.Compared to use the genetic algorithm directly,surrogate-assited optimization can greatly enhance the prediction efficiency and maintain good accuracy.The experimental results show that the optimization algorithm with the surrogate model can predict delaminations to have overlap with the actual ones indicating satisfactory accuracy,in particular,the sizes can be predicted with better accuracy than the locations.The artificial neural network was found to be very sensitive to experimental errors,and could not give meaningful delamination information when inputting the measured frequency shifts.A further sensitivity analysis show that the surrogate-assisted optimization algorithm was less sensitive to the artificial noise and could still predict delaminations with satisfactory accuracy with noised frequency shifts,while ANN was vulnerable to the noise in the inputting frequency shifts and performed much worse in delamination prediction.In conclusion,frequency-based method was validated numerically and experimentally to be able to predict delaminations successfully and the surrogate-assisted genetic algorithm was recommended to be applied since it is more robust.