CN-121997229-A - Self-adaptive shrinkage compensation self-compacting concrete compression resistance prediction method

Abstract

The invention discloses a self-adaptive shrinkage compensation self-compaction concrete compression resistance prediction method, and belongs to the technical field of building materials. The method comprises the steps of firstly collecting environment and raw material data, extracting key parameters through characteristic engineering, then constructing an integrated learning prediction model integrating a support vector machine and a self-adaptive lifting algorithm, and finally dynamically optimizing the mixing amount of intelligent materials such as a water retaining agent, a shrinkage reducing agent and the like based on real-time environment parameters to realize the self-adaptive design of a mixing ratio. The method can accurately predict the compressive strength and working performance of the concrete in different environments, effectively coordinate the mechanical performance and working performance, and remarkably reduce the test period and the void risk.

Inventors

• CHEN YONGGUI
• ZHANG TAO
• LU GUANLI
• DONG XUEHAI
• Zhao Wangcheng
• MENG DESHENG
• LI LINGFENG
• XIE KAIZHONG
• LI YONGLIANG
• ZHANG CHUNHAO
• Shui Jingming
• LI FEI
• ZHANG YUANPENG
• WEI LEI

Assignees

• 中铁二十五局集团第四工程有限公司
• 中铁二十五局集团有限公司
• 广西大学

Dates

Publication Date

20260508

Application Date

20260126

Claims (4)

1. 1. The self-adaptive shrinkage compensation self-compacting concrete compression resistance prediction method is characterized by comprising the following steps of: S1, collecting real-time temperature T and humidity RH of a construction area, and determining the expected change range of T, RH and the upper and lower thresholds of the expected change range by combining meteorological data of the engineering site and forecast of the construction period; S2, deleting samples with 28 d compressive strength attributes missing or repeated in the concrete raw material production data, performing outlier analysis on first class data attributes of the rest samples, deleting the samples if more than 3 attribute values in the same sample are outliers, complementing the missing values in the first class data with 0; S3, performing Z-score standardization on the cleaned data to eliminate dimension, extracting characteristics obviously related to the compressive strength of 28 d by pearson correlation coefficient, combining the usage amount of fine sand, middle sand and coarse sand into the usage amount of fine aggregate, combining the usage amount of small stone and middle stone into the usage amount of coarse aggregate, combining the usage amount of water and reclaimed water into the usage amount of water, keeping the rest characteristics as they are, and finally normalizing all the characteristic data to form a modeling characteristic set; s4, adopting epsilon-SVR to relax variable 、 And regularizing the complexity and training error of the punishment coefficient C balance model, mapping the characteristics to a high-dimensional space through an RBF kernel function, and simultaneously regressing 28 d compressive strength and 60d shrinkage rate in an epsilon-insensitive zone; S5, using the epsilon-SVR agent model obtained in the step S4 as a prediction engine, running in a programmable environment simulation box according to real-time T, RH, using a water retention agent x 1 , a shrinkage reducing agent x 2 and a phase change material microcapsule x 3 as decision variables, constructing a multi-objective optimization function, and adopting a weighted sum form: ; Wherein xs= [ x 1 ,x 2 ,x 3 ] is a decision variable vector, x 1 ,x 2 ,x 3 represents the doping amounts of the water retention agent, the shrinkage reducing agent and the phase change material respectively, T and RH are the temperature and the humidity input in real time, f 6h is a prediction function of the slump expansion degree of 6 hours, f 28d is a prediction function of the compressive strength of 28d, w 1 ,w 2 is a weight coefficient, and w 1 +w 2 =1, wherein w 1 ,w 2 is the weight coefficient; Solving to obtain the optimal xs= [ x 1 ,x 2 ,x 3 ], performing trial mixing and forming according to the proportion, verifying the slump expansion degree of 6h, the compressive strength of 28d and the shrinkage rate of 60d in an environment box, and refluxing the measured data to a sample library to periodically retrain the model to form a closed loop.
2. 2. The method according to claim 1, wherein the normalization in step S3 is performed according to the following formula: ; Wherein, the Is the sample mean value; Is the i-th observation; is the standard deviation of the sample; The pearson correlation coefficient is calculated as: ; wherein r is the sample pearson correlation coefficient, the value is [ -1,1], the closer to 1 the stronger the linear correlation, n is the sample capacity, the absolute value Is the standard deviation of the sample; is the average value of the samples; Is the i-th observation.
3. 3. The method according to claim 1, wherein step S4 is specifically: S41, model structure: by epsilon-SVR, introducing relaxation variables 、 And regularization penalty coefficient C, in epsilon-insensitive zone, simultaneously regressing 28 d compressive strength and 60d shrinkage, and the original characteristics pass through RBF kernel function: ; mapping to a high-dimensional space to complete nonlinear fitting; s42, optimizing targets: the optimization objective function is: ; Wherein, w is a weight vector, C is a regularization penalty coefficient used for balancing model complexity and training errors; And Is a relaxation variable; the constraint conditions are as follows: ; Wherein, the As a weight vector of the weight vector, Is a bias term that is used to determine, And Is a relaxation variable, allowing the sample point to deviate The belt is not sensitive to the motion of the belt, Is a penalty parameter for balancing model complexity and training errors, Is a non-sensitive loss parameter that is not sensitive to, Is a kernel that maps inputs to a high-dimensional feature space, Is the first The true value of the individual samples is calculated, Is the number of training samples; S43.svr prediction function: ; ; s44, quantifying prediction accuracy through regression evaluation indexes: ; ; The calculated model needs to meet R 2 > 0.85 and the RMSE is lower than a preset threshold; S45, super parameter optimizing: performing 5-fold grid cross validation on (C, gamma, epsilon), and obtaining an optimal parameter combination by taking R2 and RMSE as convergence criteria; s46, model output: and obtaining an epsilon-SVR proxy model capable of simultaneously outputting 28d compressive strength and 60d shrinkage, and directly using the epsilon-SVR proxy model for the subsequent multi-objective optimization of the blending ratio.
4. 4. The method of claim 1, wherein the multi-objective optimization function in step S5 satisfies 28d compressive strength ∈design value ∈650mm, initial slump extension ∈650mm, 0.05 +≤x 1 ≤0.3%、1%≤x 2 ≤3.4%、2%≤x 3 +≤8%.

Description

Self-adaptive shrinkage compensation self-compacting concrete compression resistance prediction method Technical Field The invention belongs to the technical field of building materials, and particularly relates to a self-adaptive shrinkage compensation self-compaction concrete compression resistance prediction method. Background The development of the steel pipe concrete arch bridge is very rapid nowadays, and the steel pipe concrete arch bridge has a plurality of advantages of strong universality. The steel tube concrete arch bridge has the advantages of being capable of being considered in a plurality of span ranges, strong in bearing capacity, capable of increasing bearing capacity of the arch rib to meet more engineering demands, strong in adaptability, capable of being considered to be used under the condition of poor soil conditions due to the fact that horizontal thrust received by the arch foot is small due to the arch rib structure, attractive in appearance, various in form and capable of being used as an ornamental bridge in a specific scenic spot. Self-compacting concrete is widely applied to the pouring construction of a steel tube concrete arch bridge due to the excellent fluidity and vibration-free characteristics. The ideal state is that the pipe cavity is filled under the action of dead weight, and the pipe cavity is tightly adhered to the inner wall of the steel pipe to form a cooperative composite structure, and the steel pipe and the concrete can form the effect of 1+1> 2. The self-compacting concrete is special concrete with high fluidity and excellent segregation resistance, can flow and fill each corner in a template only by means of dead weight when pouring, can fully wrap reinforcing steel bars without compacting by external vibration, can ensure uniform and compact structure, avoid the defects possibly caused by the traditional vibration, improve the construction efficiency and reduce noise, and is particularly suitable for engineering structures with complex shapes, such as core tubes of high-rise buildings, prefabricated parts or pouring in arch bridge pipes. The performance of the self-compacting concrete is adjusted mainly by adjusting the water-cement ratio, the sand ratio and the admixture mixing amount, and the measures can effectively adjust the working performance of the concrete. However, the mechanical property and the working property of the self-compacting concrete are inversely proportional to each other, so that the mechanical property and the working property of the self-compacting concrete are required to be coordinated, the mechanical property of the self-compacting concrete can meet the design requirement on the premise of ensuring that the working property of the self-compacting concrete reaches the standard of no blocking, but the workload is relatively large, each test block needs to be maintained for 28 days, the test period is relatively long, and the input time, manpower and material resources are relatively large, so that the requirements of a construction site are difficult to meet. In summary, there is an urgent need to develop a novel self-compacting concrete compression resistance prediction method, which not only can make the designed concrete mix ratio meet the long-distance and high-lift pumping construction requirements, but also can adapt to the external complex environmental conditions, and realize the whole process volume stability from the plastic stage to the hardening stage by intelligently regulating and controlling the working retention, hydration process and shrinkage compensation behaviors of the self-compacting concrete mix ratio, thereby significantly reducing or even eliminating the void diseases. Disclosure of Invention The invention aims to provide a self-adaptive shrinkage compensation self-compacting concrete compression resistance prediction method, which solves the problem of how to realize real-time, dynamic and self-adaptive design of self-compacting concrete mixing proportion, so that the self-compacting concrete compression resistance prediction method can generate optimal mixing proportion before construction based on real-time environment data and raw material states through AI prediction and optimization, thereby ensuring that the concrete performance meets pumping construction requirements under complex and changeable environments and realizing specified strength and volume stability. The compression strength of the shrinkage compensation self-compacting concrete can be accurately and rapidly predicted by extracting the characteristics of the production data related to the concrete raw materials through characteristic engineering and combining two integrated learning methods. The technical scheme of the invention is as follows: a self-adaptive shrinkage compensation self-compacting concrete compression resistance prediction method comprises the following steps: S1, collecting real-time temperature T and humidity RH of a c