Delamination Detection In Fiber Reinforced Polymer Beams Based On Frequency And Fiber Bragg Grating

Fiber Reinforced Polymer(FRP)composite materials which have advantages such as high specific strength,specific modulus,designability,excellent fatigue resistance,corrosion resistance,vibration resistance,etc.,have been widely used in aerospace,energy and transportation,marine engineering,sports equipment,construction and medical engineering.The widely used FRP composite materials are mainly used in the form of laminated structure,and the bonding force between laminate layers is determined by the resin matrix of which the mechanical properties are weak.Thus,the delamination damage of FRP materials occurs frequently in the case of medium and low speed impact during the processing,storage,transportation,service and maintenance.As the delamination propagates,the carrying capacity of FRP laminated structures may decrease rapidly,and even cause the structures collapse and lead to severe loss of life and property.Therefore,it is of great significance to have an early diagnosis and identification of delamination damage in FRP laminated structures.In this thesis,the FRP composite laminated beams are the object of current study.The project mainly focuses on delamination detection for the FRP laminated beams through changes in vibration parameters such as modal frequencies as well as the wavelength of fiber bragg gratings.Several machine learning algorithms,i.e.,artificial neural network,support vector machine,extreme learning machine and genetic algorithm were built for detecting the delamination parameters in FRP laminated beams,i.e.,damage location,size and interface.Moreover,numerical and experimental verification of machine learning algorithms were carried out,respectively.(1)In the numerical validation,theoretical model of FRP beam,finite element modelling of FRP composites and two surrogate models based on artificial neural network and support vector machine were constructed to simulate the vibration of the beams.These numerical models were used to generate ‘delamination damage parameters-frequency shifts' database and the numerical test cases to validate and compare the machine learning algorithms in terms of the identification accuracy and prediction time.(2)In the experimental validation,undamaged and delaminated FRP composite beam specimens were manufactured,and then modal testing and the optic wavelengh test with fibre bragg gratings were conducted under cantilever boundary condition.The changes in measured vibrational frequency and optic wavelength were used as the target values to be achieved through the machine learning algorithms.In addition,the adjustable parameters of each algorithm during the damage identification through frequency changes are discussed,including database size,hidden layer node number,normalization,input and output forms and network parameters.The results of numerical verification based on frequency show that the prediction effect of BP neural network is the best,with interface prediction accuracy as high as 98.15%,while the prediction error of damage location and size being only 0.14% and 0.17%,respectively.The experimental results show that the prediction accuracy is the highest for support vector machine,which is followed by the extreme learning machine,genetic algorithm based on the surrogate model and BP neural network.To sum up,the absolute predicition error of four kinds of machine learning algorithms all less than 3.3%.However,for the interface of delamination,the prediction accuracy of support vector machine has achieved 100%.The experimental results based on optic wavelength show that both BP neural network and support vector machine can predict the relative position between loading point and delamination location.In conclusions,the four kinds of machine learning algorithms based on the changes of frequency and wavelength can successfully identify the FRP beam delamination damage with good accuracy,among them support vector machine has the best prediction accuracy.In this thesis,the high-efficiency prediction of frequency were combined with the fiber bragg gratings to predict the delamination,which can save the work to distribute the fiber bragg gratings in a very fine grid,and also to further verifiy and confirmthe delamination parameters predicted through the frequency shift.