Temperature Field Of Mass Concrete Based On BP Neural Network Positive And Negative Analysis Research

As an important structural type of modern bridge construction,the problem of cracking has always been one of the most concerned issues for engineering and technical personnel.In order to guide the construction of large-volume concrete and reduce the occurrence of temperature cracks,it can be simulated and calculated according to the existing temperature field and stress field theory,and specific temperature control measures for each construction stage are formulated.However,in the finite element analysis of the temperature and stress fields of large-volume concrete,the thermal parameters used are mainly obtained through empirical formulas or experiments.Due to the inevitable errors in the empirical formulas,the tests are rarely used because of their high cost and time-consuming.And these parameters are affected by many factors such as meteorological conditions,space-time,loads,construction conditions during the construction period,which often distort the adopted thermal parameters,deviate or even seriously deviate from the actual situation of large-volume temperature control calculations,which is not conducive to engineering The personnel real-time grasped the pouring condition and operation of the large-volume concrete,which misled the engineering personnel to take inaccurate temperature control measures,resulting in more or less cracks in some large-volume concrete structures.In view of the above problems,based on the existing research results,this paper puts forward the idea of "Finite element finite element analysis of temperature field-formulation of temperature control measures-on-site temperature monitoring measurement-temperature field inverse analysis-temperature control measures optimization adjustment" Risk of cracking of large volume concrete in the project.This paper takes the Taihong Yangtze River Bridge scaffold saddle support pier as the engineering background to make related research.The main work and results are as follows:(1)The main characteristics of mass concrete,the causes of temperature cracks,and the research status of temperature control measures are reviewed.Mass concrete cracking is affected by many factors,and each factor is a cause and effect.Due to the difference in the environment of concrete,the requirements for temperature control and crack prevention of concrete are not the same,but the common requirement for various types of large-volume concrete is to control the generation and development of cracks.(2)According to the concrete performance test or empirical formula,the calculation parameters of the temperature field of the bulk concrete are determined;the temperature field positive analysis model is established by using MIDAS/FEA,and the distribution of the temperature field and stress field of each layer of concrete is obtained;the structure The law of temperature change in each layer is roughly the same,which can be summarized as follows:rapid temperature rise and high temperature peaks;large temperature difference between the inside and the surface,and they all go through the process of rising first and then falling;the cooling rate is too fast,and the internal cooling rate is higher than the surface The cooling rate is faster;the stress curve of each feature point is below the allowable stress curve,and the difference is large,that is,the stress value of each feature point is less than the tensile strength of the concrete at the corresponding age,and there is a certain safety factor.(3)Using the strong non-linear mapping ability of BP neural network,the temperature field back analysis model is established,and the neural network toolbox is used to complete the programming of the BP neural network algorithm.In determining the network parameter samples,a uniform design method is introduced to greatly reduce the number of network learning samples,ensure that each level of each factor is done within its value range and only one test is performed,and the network input and output samples are summarized.The unified processing eliminates the difference in the magnitude and dimension of each parameter.(4)Three widely accepted performance indicators:goodness of fit,average absolute percentage error,and root mean square error quantify the training effect of the BP neural network.All three performance indicators are ideal,and the predicted values in the test sample set are concentrated.Good fit with real values indicates that the BP neural network model has higher prediction accuracy and less error for large-volume concrete,and can be used to invert the thermal parameters of large-volume concrete(5)A temperature control and monitoring scheme for the temperature field of the cap of the scaffold saddle pier was developed,and the adiabatic temperature rise,reaction constant,thermal conductivity,and surface heat release coefficient were obtained by inverting the measured temperature at the characteristic points.The calculated temperature based on the modified model agrees well with the measured temperature.The adjustment direction of temperature control measures was clarified,and the temperature control measures were adjusted and verified from the perspective of construction control.