Inversion Analysis Model Of Damage Inducement For High-piled Slab&Beam Wharf Based On Support Vector Machine Optimized By Particle Swarm Optimization Algorithm

High-piled slab&beam wharf is a kind of harbor structure mainly built on soft foundation,which has been widely developed and constructed in the coastal areas of China and the lower reaches of the Yangtze River.During the service of the wharf,the wharf structure is easily affected by such adverse factors as over-limit stacking load,nonstandard berthing,and uneven lateral settlement of the structure,which leads to the damage of the wharf structure and poses a serious threat to the overall safety of the wharf structure.The existing health monitoring methods of the wharf structure mainly focus on the identification and location of the damage of the wharf structure.At present,there are few studies on early warning and tracing of the adverse effects of the damage of the highpiled slab&beam wharf.In order to construct inversion model of damage inducement for high-piled slab&beam wharf,the research work and achievements are as follows:(1)Based on the literature research and combined with the characteristics of the high-piled slab&beam wharf,the types,forms,positions and approximate scope of its main damage inducements are determined.A generalized finite element model of highpiled beam-slab wharf is established.Combining the Generalized Structural Stiffness Theory with Abaqus’ life-death element method,this thesis studied the importance of each pile of the wharf under the action of overload,irregular berthing of ships,and lateral settlement of the wharf structure,and determined the monitoring pile foundation of the wharf.According to the general service environment of high-piled slab&beam wharf(high salt fog environment),as well as the current sensor elements and installation technology patents,the monitoring area of the wharf monitoring pile foundation was determined,and the vertical strain of the pile foundation was taken as the monitoring target.In addition,the strain distribution characteristics of the monitoring area ware studied under the specific damage inducing conditions of the monitoring pile foundation.The results show that the strain distribution in this region was obvious and can be used as the information source of the inversion model.(2)Aiming at the technical demands of damage inducement inversion of high-piled slab&beam wharf,an inversion model based on support vector machine was proposed and constructed.For the characteristic parameters of support vector machine: penalty factor C,parameters of RBF kernel function: kernel bandwidth σ,The particle swarm optimization algorithm was used to optimize the two.Research and determine the commonly used model evaluation methods and evaluation indicators for classification and regression problems.(3)To satisfy the requirement of damage inducement inversion model of support vector machine on quantity and quality of inversion sample data set for training,this thesis used Python to carry out secondary development of Abaqus.on the basis of the constructed finite element model of the high-piled plate&beam wharf,realized the submission and calculation of multiple random working conditions,and extracted the strain data of the points in the calculation results,so as to obtained samples of strain data in batches.3000 sets of working conditions of stacking load,9000 sets of working conditions of ship collision and 1000 sets of working conditions of lateral settlement are calculated,and 13000 strain data are extracted for the construction of inversion data set to meet the data requirements of inversion model training.(4)Using the trained damage inducement inversion model,the inversion verification is carried out for the three main damage inducements,the overload,the out-of-control ship collision and the lateral settlement: the identification accuracy of the model for the type of damage inducement was 0.965.The accuracy of the position identification of the overload was 0.94.For the strength identification of the overload,the model index R2 reached 0.998,and the average relative error of the inversion result of the action strength was about 4.34%;For the identification of ship collision action position,the accuracy of the inversion model reached 0.958.For the identification of ship collision action strength,the index R2 of the model was 0.997 in the training set and the test set,and the average relative error of the strength inversion result was 5.64%;For the lateral settlement of the wharf structure,the accuracy of the inversion model for identifying the location of the settlement was 0.925.For the identification of the lateral settlement of the inversion model,the index R2 of the inversion model was 0.999 in the training set and the test set,and the average relative error of the strength inversion result was 1.64%.In conclusion,three main damage inducements can be effectively identified by the damage inducement inversion model of SVM optimized by particle swarm optimization.