| At present,large-span roof structure is widely used in various occasions.This kind of building has the characteristics of high,light and flexible,and the natural frequency is dense,so it is particularly sensitive to wind load,which is the control load affecting the safe use of building structure.In the structural design of long-span building structure,the effect of wind load must be fully considered.In the past,in the wind resistant design of this kind of building structure,the fluctuating wind load is often assumed to be a stationary Gaussian process in Chinese codes.However,recent studies have found that in the negative pressure region of building structure,the higher-order moment of fluctuating wind load is significantly different from the lower order moment.In the part with too much non Gaussian characteristics distribution,if the fluctuating wind is still assumed to be a stationary Gaussian process,It may make the obtained wind-induced response coefficient smaller than the actual value,which may cause the building structure shaking when it is severe,the shedding or transfer of the wind pressure vortex area on the building surface may cause deformation vibration,and in extreme cases may cause the building surface material deformation or structural collapse;When the dynamic displacement exceeds a certain limit,the structural members will be destroyed;Large scale surface vibration will lead to the fatigue damage of the external components of the building structure and its ancillary buildings,which will seriously affect the structural safety of the building for a long time and constitute a potential safety hazard.Based on intelligent algorithm,this paper studies the non Gaussian wind load characteristics of long-span roof structure.At first,the AR autoregressive system and Johnson transform are used to build jt-ar transformation model to simulate the non Gaussian fluctuating wind load,so as to pave the way for the prediction of non Gaussian fluctuating wind load.A hybrid intelligent algorithm,which combines cuckoo search algorithm and particle swarm optimization algorithm,is used to optimize the parameter combination of least squares support vector machine(LSSVM),and to predict and analyze the non Gaussian fluctuating wind load of long-span roof structures with different shapes.The main contents of this paper are as follows(1)Based on the coupling of JT transformation and AR model theory,a jt-ar transformation model is proposed and constructed to simulate the generation of non Gaussian fluctuating wind pressure time history sample data.Compared with the target power spectrum and high-order statistics,the proposed method is used to study the non Gaussian fluctuating wind pressure time history characteristics of long-span structure surface;The reliability and universality of jt-ar transformation model for simulating non Gaussian fluctuating wind are studied and demonstrated by comparing the non Gaussian distribution characteristics of wind tunnel test results on the surface of long-span structures.Before each simulation test,normalization is needed,and the non Gaussian fluctuating wind load sample data is divided into training set and prediction set.(2)Based on the theory of improved cuckoo search algorithm(CS)and particle swarm optimization(PSO),a hybrid intelligent algorithm is proposed to optimize the parameters combination of least squares support vector machine(LSSVM)to form a radial basis function(RBF)kernel function,and a learning machine model based on improved cuckoo search algorithm and particle swarm optimization(CS + PSO-LSSVM)hybrid intelligent algorithm is constructed,The learning and generalization ability of CS + PSO-LSSVM hybrid intelligent algorithm learning machine model and cs-lssvm and PSO-LSSVM intelligent algorithm learning machine model for non Gaussian fluctuating wind pressure time history on the surface of long-span structures are compared,The accuracy and effectiveness of CS + PSO-LSSVM hybrid intelligent algorithm learning machine are proved by calculating the distribution characteristics and higher-order statistics errors between the predicted non Gaussian fluctuating wind pressure and the actual test set wind pressure.(3)Based on the wind tunnel test of long-span cylindrical reticulated shell roof structure and the numerical simulation results of long-span annular roof structure,the superior performance of CS + PSO-LSSVM hybrid intelligent algorithm in the study of non Gaussian fluctuating wind load on the surface of various long-span roof structures is studied,and the characteristics of non Gaussian wind load on the surface of different long-span roof structures are studied,The distribution characteristics of non Gaussian wind load and wind pressure on the surface of long-span roof are analyzed and summarized.The main conclusions are as follows:(1)By analyzing the simulation results of jt-ar conversion model for non Gaussian fluctuating wind pressure time history,the complete non Gaussian fluctuating wind load data can be obtained accurately.It is concluded that jt-ar conversion model can replace wind tunnel test to a certain extent,and this method has universality and reliability.(2)Compared with cs-lssvm and PSO-LSSVM,the advantages of CS + PSO-LSSVM hybrid intelligent algorithm and its learning generalization ability for non Gaussian fluctuating wind are summarized.(3)The advantages and disadvantages of CS + PSO-LSSVM hybrid intelligent algorithm for non Gaussian fluctuating wind load on the surface of many kinds of long-span roof structures are studied and verified.The non Gaussian distribution characteristics of large-span roof surface are summarized,which are related to the building structure form,measuring point location and area.It provides a new solution to simulate the non Gaussian fluctuating wind load on the surface of long-span structure and to train and predict the non Gaussian fluctuating wind load,and provides valuable suggestions and basis for the wind resistant design and safe use of long-span roof structure. |
PDF Full Text Request