Analysis Method Of Incomplete Similarity Error Of Steel Structure Semi-rigid Connection And Frame

The scale model is widely used in structural tests due to its economical efficiency,easy operation and short test period.However,structural scale models are generally not completely similar.Take the steel structure beam-column connection node as an example,rolled steel components of full-scale and reduced-scale models are difficult to ensure that they meet completely similar conditions in size,bolted nodes can't be scaled down either.Therefore,almost all steel structure scale models are not completely similar,which need to conduct similarity analysis and design.At the same time,the traditional similarity theory can only solve the design ratio of the main physical quantities in the design of similar models,it cannot solve the similarity error caused by incomplete similarity.Therefore,this article takes the steel structure semi-rigid connection node and frame as the research object,systematically studied the errors caused by the incomplete similarity of the scale model,propose a calculation method that can effectively predict the incomplete similarity error of the scale model.The main work of this article is as follows:(1)Based on the similarity theory and the principle of dimensional harmony,the complete similarity condition of the beam-column connection node of the semi-rigid steel structure was derived,and the semi-rigid node stiffness Ki was introduced as an independent parameter.Clarified the limitations of similar conditions in predicting similar errors and the direction of improvement.Based on Python,the semi-rigid node automatic modeling and post-processing program developed by ABAQUS was compiled to realize the rapid parametric analysis of large batches of non-completely similar nodes.The Spearman nonparametric correlation coefficient was used to screen the main geometric parameters of the typical semi-rigid end plate connection.Through the analysis of the results of 102 noncompletely similar models,4 factors that have the greatest impact on node similarity goals were selected from 28 factors,which provides preconditions for the scientific selection of experimental parameters.(2)Established a refined node finite element model,compared with the full-scale model test results,and analyzed the reliability of the finite element model.The consistency of the stress distribution between the completely similar scaled end plate connection node model and the full-scale model was studied.The influence of end plate thickness,beam section height,column section height,bolt specification and other factors on the stress distribution of beamcolumn end plate connection and T-piece connection node was analyzed.Explain the stress distribution characteristics of the above two non-completely similar semi-rigid connection models.The influence of the degree of plastic development on the stress distribution of the beam-column double web connection node was analyzed.Clarified the effect of each test element on the test results of four semi-rigid connection types: end plate connection,T-piece end plate connection,double web angle steel connection,and cantilever section bolt connection when the joint stiffness is the similar target quantity.The most important factors were obtained.(3)Established a proxy model method for similar error prediction,built a database of node models that are not completely similar.Based on different factors,the similar error prediction problem is divided into three levels,including single-factor error analysis method using least square method,two-factor correlation error estimation method based on response surface method,and multi-factor correlation error analysis method using artificial neural network.Take the connection of the overhanging end plate and the connection of the T-piece as an example,the above method was used to obtain the non-completely similar error prediction formula.(4)Based on the cyclic loading test of 6 semi-rigid connection full-scale models and 6 non-completely similar scale models,considering the influence of non-completely similar on the model cyclic loading test.Compared the full-scale and reduced-scale model test phenomena of four semi-rigid connection types: end plate connection,T-piece end plate connection,double web angle steel connection,and cantilever section bolt connection.Summarized and analyzed the hysteresis curve,skeleton curve,energy consumption capacity,and strain of test points of 12 node model specimens.(5)The complete similarity condition of the semi-rigid steel frame including the joint stiffness characteristics was derived.Applying the method of incomplete similarity error prediction to the error prediction of semi-rigid steel frame.A loop iteration algorithm nested between ANSYS and MATLAB was developed,and the sensitivity distribution of each factor was obtained by calculating 2950 steel frame models.Based on the Monte Carlo method,50,000 sets of non-completely similar steel frame model databases are established,and the similar errors of semi-rigid steel frames are predicted based on the artificial neural network method in the proxy model method.