Mix proportion design method of micro-expansion self-compacting high-strength steel tube concrete based on artificial neural network

Abstract

The application discloses a mix proportion design method of micro-expansion self-compacting high-strength steel pipe concrete based on an artificial neural network, which comprises the following steps: step 1, constructing a sample set; step 2, training a high-strength steel pipe concrete volume free shrinkage prediction model; step 3, obtaining a demand value of a high-strength steel pipe concrete shrinkage deformation amount to be controlled and setting influence parameter theoretical data; step 4, obtaining a shrinkage prediction value of the influence parameter theoretical data by using the trained model; step 5, calculating a target value of micro-expansion based on the demand value and the shrinkage prediction value, and obtaining a recommended expansion agent content by using expansion value samples of the high-strength steel pipe concrete under different expansion agent contents; and step 6, performing a volume shrinkage experiment of the high-strength steel pipe concrete, and if the experimental result reaches the shrinkage deformation demand, determining recommended proportions of the self-compacting micro-expansion high-strength steel pipe concrete according to the influence parameter theoretical data and the recommended expansion agent content, otherwise artificially updating the influence parameter theoretical data and returning to step 4.

Description

Technical Field

[0001] This invention belongs to the field of construction engineering, and specifically relates to a mix design method for micro-expansion self-compacting high-strength steel-tube concrete based on artificial neural networks. Background Technology

[0002] In the field of construction engineering, steel-concrete composite materials are composite structural materials in which concrete is poured inside a steel tube to form a joint bearing structure. The "hoop" effect of the steel tube increases its combined load-bearing capacity by 1.7-2.2 times. It uses less concrete, is simple to construct, and is a typical high-strength, low-carbon, and economical building structural material with broad application prospects in the construction of high piers and long-span bridges. Generally, steel-concrete composite materials with a concrete strength grade below C50 are called ordinary steel-concrete composite materials; those with a concrete strength grade above C50 are called high-strength steel-concrete composite materials. Due to the existence of concrete shrinkage, uneven deformation can occur, causing separation between the steel tube and the concrete. This can also lead to stress concentration within the concrete, resulting in cracks, degradation of the concrete structure's stiffness, and a decrease in overall structural integrity. In severe cases, it can cause reinforced concrete members to lose their load-bearing capacity and cease functioning. Therefore, it is necessary to predict and control the volume deformation of steel-concrete composite materials.

[0003] Concrete-in-steel tubes are made by pouring concrete into steel tubes. Therefore, the volume change of the concrete mainly comes from the self-shrinkage of the concrete and the temperature shrinkage caused by temperature difference. At present, the test method for the volume deformation of concrete-in-steel tubes is to first determine the volume change of concrete-in-steel tubes through experiments, and then optimize the mix proportion of concrete-in-steel tubes based on the experimental measurements. This process often takes a lot of time.

[0004] Chinese patent CN105906262B discloses a method for continuous and precise expansion control of steel-tube concrete. This invention effectively addresses the plastic shrinkage, autogenous shrinkage, and internal self-drying shrinkage of steel-tube concrete by combining an expansion agent with a shrinkage-reducing agent and / or an internal humidity compensator. However, this method still has the following shortcomings: 1. Obtaining the concrete mix ratio solely through experimental adjustments, due to the use of various admixtures, leads to high time and experimental costs, a large number of experiments, and increased experimental and time costs; 2. The use of multiple types of admixtures lacks clear guidance in admixture selection.

[0005] Chinese patent CN105906262B discloses an expandable, designable high-strength steel-tube concrete and its preparation method. This invention designs an expandable, designable C50 and C60 high-strength steel-tube concrete and its preparation method, reducing the air content of the concrete and minimizing debonding between the steel tube and concrete caused by gas adhering to the inner wall of the steel tube, thus achieving continuous and stable expansion of the core concrete. However, this method still has the following shortcomings: 1. The dosage of the expanding agent needs to be determined through both experiments and formulas, which is quite cumbersome; 2. This method is only applicable to fixed mix proportions; once the water-cement ratio is changed, this method becomes inapplicable, resulting in a narrow scope of application.

[0006] Currently, engineering methods for predicting the volumetric deformation of high-strength steel-concrete composites primarily rely on existing experimental data and statistical analysis to establish empirical formulas. These methods, which typically use experimental data for calculations or empirical summaries, are subject to numerous limitations, lack versatility, and are unsuitable for rapid and accurate prediction. Neural networks, as powerful pattern recognition and function fitting tools, offer advantages such as high accuracy and wide applicability, and have been widely applied in predictive analysis. However, no methods for optimizing the mix design of steel-concrete composites based on neural networks have been reported, nor have any methods for predicting the volumetric deformation of micro-expansion self-compacting high-strength concrete based on artificial neural networks been reported. Summary of the Invention

[0007] The purpose of this invention is to provide a mix design method for micro-expansion self-compacting high-strength steel-tube concrete based on artificial neural networks, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: A mix design method for micro-expansion self-compacting high-strength steel-tube concrete based on artificial neural networks includes the following steps: Step 1: Collect data on the influencing parameters related to the free volume shrinkage of high-strength steel-concrete composite and the shrinkage values ​​under their influence, and construct a sample set; Step 2: Train a high-strength steel-concrete composite volume free shrinkage prediction model based on an artificial neural network using the sample set; Step 3: Obtain the required value of shrinkage deformation of high-strength steel-concrete composite pipe to be controlled, and set the theoretical data of the influencing parameters; Step 4: Use the trained high-strength steel-concrete composite volume free shrinkage prediction model to obtain the shrinkage prediction values ​​of the theoretical data of the influencing parameters; Step 5: Calculate the target value of micro-expansion based on the demand value and shrinkage prediction value, obtain the expansion value sample of high-strength steel pipe concrete under different expansion agent dosages, and combine the target value of micro-expansion to obtain the recommended expansion agent dosage. Step 6: Conduct a volume shrinkage experiment on high-strength steel-tube concrete using the expansion agent dosage and theoretical data of the influencing parameters, and determine whether the experimental results meet the shrinkage deformation requirements. If so, determine the recommended mix proportion of self-compacting micro-expansion high-strength steel-tube concrete based on the theoretical data of the influencing parameters and the recommended expansion agent dosage; otherwise, manually update the theoretical data of the influencing parameters and return to step 4.

[0009] Preferably, the influencing parameters include water-cement ratio, bone glue ratio, fly ash content, mineral powder content, silica fume content, curing temperature, curing humidity, and curing age.

[0010] Preferably, the shrinkage value is a volume shrinkage value based on a time series, with days as the basic unit.

[0011] Preferably, in the concrete volume shrinkage prediction model, the number of neurons in the input layer is equal to the number of influencing parameters, the number of neurons in the hidden layer is greater than the number of neurons in the input layer, the number of neurons in the output layer is 1, the activation function between the input layer and the hidden layer is the Sigmoid function, and the transfer function between the hidden layer and the output layer is the Purelin function.

[0012] Preferably, step 2 includes: Step 2.1: Divide the sample set into a training set and a test set; Step 2.2: Set training parameters and use the training set to train the high-strength steel-concrete composite volume free shrinkage prediction model in multiple rounds. Step 2.3: Use the test set to test the prediction error of the high-strength steel-concrete composite volume free shrinkage prediction model after training. If it is less than the preset threshold and less than the previous prediction error, then end the process; otherwise, return to step 2.2.

[0013] Preferably, the LM algorithm is used as the neural network model training algorithm during training.

[0014] Preferably, in step 6, the volume shrinkage experiment includes: Based on theoretical data of influencing parameters and suggested dosage of expansive agent, the preparation data of high-strength steel-tube concrete was determined. Based on the preparation data, four high-strength steel-tube concrete specimens were cast in steel molds. Two specimens were selected and connected to the data acquisition system in advance using the lead wires of the non-rebound tester. At this time, the TDS-530 started to read automatically and tested the shrinkage data of 0-1 days. The average shrinkage value of the two specimens at the same time was taken, and the shrinkage curve S1 of 0-1 days was plotted. When the concrete reaches 1 day of age, the steel molds of the other two specimens are removed, and the surfaces of the concrete specimens are immediately sealed with aluminum foil. The non-rebound tester is connected to the data acquisition system with lead wires to test the shrinkage data of the specimens from 1 to 28 days of age. The average shrinkage value of the two specimens at the same time is taken to plot the shrinkage curve S2 from 1 to 28 days. The two curves are combined to obtain the contraction curve for the entire process.

[0015] Preferably, the merging specifically involves: when the cutoff point of the contraction curve S1 and the starting point of the contraction curve S2 are the same, the value is used as the connection point to connect the two contraction curves; when the cutoff point of the contraction curve S1 and the starting point of the contraction curve S2 are different, their average value is used as the connection point, which serves as both the cutoff point of the contraction curve S1 and the starting point of the contraction curve S2, thus connecting the two contraction curves.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: In this invention, a volume free shrinkage prediction model is constructed to predict the shrinkage value corresponding to the set influence parameter value. Experiments are conducted using the influence parameter value and the recommended target admixture. If the experimental results do not meet the shrinkage deformation requirements, the influence parameter value is adjusted. If the requirements are met, the recommended mix proportion is determined. This invention can obtain the corresponding shrinkage value without conducting volume free shrinkage experiments on the influence parameters. It can optimize the concrete mix proportion design more quickly and effectively while significantly reducing the experimental workload and facilitating guidance for engineering practice. Attached Figure Description

[0017] Figure 1 The flowchart is for the invention.

[0018] Figure 2 This is a comparison chart of experimental values ​​and predicted shrinkage values. Detailed Implementation

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0020] Figure 1 This invention illustrates the mix design method for micro-expansion self-compacting high-strength steel-tube concrete based on artificial neural networks, with reference to... Figure 1 As shown, the method includes 6 steps.

[0021] Step 1: Collect data on the influencing parameters related to the free volume shrinkage of high-strength steel-concrete composite, and the shrinkage values ​​under their influence, and construct a sample set.

[0022] In this invention, the free volume shrinkage of high-strength steel-tube concrete is related to factors such as the raw materials used in concrete preparation (e.g., water-cement ratio, bone glue ratio, fly ash content, mineral powder content, silica fume content), curing temperature, and curing age. Changes in any of these factors will cause varying degrees of volume shrinkage in the high-strength steel-tube concrete. Each influencing parameter data point consists of the specific values ​​of multiple influencing parameters, and the shrinkage value is a sequence of volume changes in high-strength steel-tube concrete over a continuous period of 0-28 days. Each influencing parameter data point corresponds to a shrinkage value, and any sample in this sample set consists of the influencing parameter data point and the corresponding shrinkage value.

[0023] In this invention, the influencing parameter data and shrinkage values ​​need to be preprocessed before constructing the sample set to remove abnormal and unreasonable data. In this embodiment, a sample set consisting of 4143 samples is obtained after preprocessing.

[0024] Step 2: Use the sample set to train a prediction model for the free volume shrinkage of high-strength steel-concrete composite pipes based on artificial neural networks.

[0025] In this embodiment, the structure of the artificial neural network-based volume shrinkage prediction model for unconstrained high-strength steel-concrete composite is PSO-ANN8-13-1, where 8 represents the number of input features, which are the parameter values ​​of eight influencing parameters: water-cement ratio, bone glue ratio, fly ash content, mineral powder content, silica fume content, curing temperature, and curing age. 13 represents the number of neurons in the hidden layer, and 1 represents the number of output features, which is the shrinkage value. The activation function between the input layer and the hidden layer is the Sigmoid function, and the transfer function between the hidden layer and the output layer is the Purelin function. The LM algorithm is used as the training algorithm for the neural network model. Specifically, the training process in step 2 is as follows: Step 2.1: Divide the sample set into a training set and a test set; Step 2.2: Set training parameters, including learning rate, maximum number of training iterations per round, and minimum training error. Use the training set to train the high-strength steel-concrete composite volume free shrinkage prediction model for multiple rounds. Here, the learning rate is set to 0.01, the maximum number of training iterations per round is set to 1000, and the minimum training error is set to 0.000001. Step 2.3: Use the test set to test the prediction error of the high-strength steel-concrete tube volume free shrinkage prediction model after training. If it is less than the preset threshold and less than the previous prediction error, then end the test and output the high-strength steel-concrete tube volume free shrinkage prediction model after the current training. Otherwise, return to step 2.2.

[0026] In step 2.2 of this invention, the influence parameter data in the sample is input into the volume free shrinkage prediction model, and the shrinkage prediction value is output. The corresponding shrinkage value in the sample is used as the target output of the prediction model. The target output is compared with the shrinkage prediction value to determine the loss value of the model. The structural parameters of the model are optimized and updated with the goal of minimizing the loss value until the model converges.

[0027] Step 3: Obtain the required value of shrinkage deformation of the high-strength steel-concrete composite to be controlled, and set the theoretical data of the influencing parameters. The required value of the concrete shrinkage deformation is a preset value, which has been given in the specification requirements and actual engineering structural calculations. The theoretical data of the influencing parameters are the initial values ​​of the influencing parameters, that is, the specific values ​​of the eight data: water-cement ratio, bone glue ratio, fly ash content, mineral powder content, silica fume content, curing temperature, curing time, and age. These values ​​are determined manually by referring to the project and combining prior knowledge.

[0028] Step 4: Use the trained high-strength steel-concrete composite volume free shrinkage prediction model to obtain the shrinkage prediction value based on the theoretical data of the influencing parameters. That is, input the theoretical data of the influencing parameters into the volume free shrinkage prediction model and obtain the output value of the model, which is the shrinkage prediction value.

[0029] It is important to note that the required value is the deformation of high-strength steel-concrete composite under the action of the expansion agent, while the shrinkage prediction value obtained by using the initial value of the influencing parameter does not involve the effect of the expansion agent.

[0030] Step 5: Calculate the target value of micro-expansion based on the demand value and the predicted shrinkage value. Obtain samples of expansion values ​​of high-strength steel-concrete composites under different expansion agent dosages. Combine the target value of micro-expansion to obtain the recommended expansion agent dosage. The target value of micro-expansion is the difference between the demand value and the predicted shrinkage value, that is, the amount of shrinkage that the expansion agent needs to compensate for in order to achieve the demand value, based on the volume shrinkage of the theoretical data of the influencing parameters.

[0031] In this embodiment, according to the specifications and the expansion rate of the expanding agent, step 5 of the present invention yields a recommended expansion agent dosage of 48.4 kg / m³. 3 The expansive agent is made from PO42.5 ordinary silicate cement, fly ash, mineral powder, sand, crushed stone, polycarboxylate superplasticizer, and MgO composite expansive agent.

[0032] Step 6: Conduct a volume shrinkage experiment on high-strength steel-tube concrete using the theoretical data of the expansion agent dosage and influencing parameters, and determine whether the experimental results meet the shrinkage deformation requirements. If so, determine the recommended mix proportion of self-compacting micro-expansion high-strength steel-tube concrete based on the theoretical data of the influencing parameters and the recommended expansion agent dosage.

[0033] The volume shrinkage experiment of the high-strength steel-concrete composite tube includes the following steps: a) Based on theoretical data of influencing parameters and suggested dosage of expansive agent, the preparation data of high-strength steel-tube concrete was determined, and four high-strength steel-tube concrete specimens were cast in steel molds using the preparation data. The main process is as follows: The calculation and determination of raw material usage: Ordinary Portland cement is used, with a usage of 290 kg / m³. 3 The fly ash usage is 97 kg / m³. 3 The amount of mineral powder used is 97 kg / m³ 3 Sand usage: 722 kg / m³ 3 The amount of crushed stone used is 1039 kg / m³. 3 Water consumption is 155 kg / m³ 3 Polycarboxylate superplasticizer 1.113 kg / m 3 Expansion agent dosage: 48.4 kg / m³ 3 ; Put ordinary silicate cement, fly ash, mineral powder, sand, crushed stone, and expansion agent into a forced concrete mixer and dry mix for 60 seconds. Pour the polycarboxylate superplasticizer into water and stir evenly, then pour it into a mixer and stir with the raw materials for 120 seconds; The mixed concrete was poured into a 100×100×400mm steel mold in two batches and compacted with a vibrator. Four self-shrinkage specimens were poured, smoothed, and sealed with plastic wrap to prevent moisture evaporation. The poured specimens were then placed in a temperature of 20℃±2℃ and a relative humidity of 60±5% for curing.

[0034] b) Take two specimens (a, b) and connect them to the data acquisition system TDS-530 in advance using the lead wire of the non-rebound tester. At this time, TDS-530 will start to read automatically and test the shrinkage data of 0-1 days. After averaging the shrinkage values ​​of the two specimens at the same time, plot the shrinkage curve S1 of 0-1 days.

[0035] c) When the concrete reaches 1 day of age, remove the steel molds of the other two specimens (c, d), immediately seal all surfaces of the concrete specimens with aluminum foil, and then use a non-rebound tester (GA-5) to connect the lead wire to the TDS-530 to test the shrinkage data of the specimens from 1 to 28 days of age. After averaging the shrinkage values ​​of the two specimens at the same time, plot the shrinkage curve S2 of the specimens from 1 to 28 days.

[0036] d) Data processing: The two curves are merged to obtain the contraction curve for the entire process. Specifically, when the cutoff point of the contraction curve S1 is the same as the starting point of the contraction curve S2, this value is used as the connection point to connect the two contraction curves; when the cutoff point of the contraction curve S1 is different from the starting point of the contraction curve S2, their average value is used as the connection point. This connection point serves as the cutoff point of the contraction curve S1 and the starting point of the contraction curve S2, thus connecting the two contraction curves.

[0037] In step 6 of this invention, it is manually determined whether the shrinkage value obtained from the experiment in step 6 meets the volume shrinkage deformation requirements. If so, a recommended mix proportion of self-compacting micro-expansion high-strength steel pipe concrete that meets the design requirements can be obtained; otherwise, the parameter values ​​are manually adjusted and the process returns to step 4.

[0038] In this embodiment, the recommended raw material and dosage for the final mix proportion is 290 kg / m³ of ordinary Portland cement. 3 The fly ash usage is 97 kg / m³. 3 The amount of mineral powder used is 97 kg / m³ 3 Sand usage: 722 kg / m³ 3 The amount of crushed stone used is 1039 kg / m³. 3 Water consumption is 155 kg / m³ 3 Polycarboxylate superplasticizer 1.113 kg / m 3 Expansion agent dosage: 48.4 kg / m³ 3 Here, to verify the accuracy of the model trained in step 2, a volumetric shrinkage experiment of high-strength steel-concrete composite was conducted. Four high-strength steel-concrete composite specimens were cast in steel molds using the prepared data. The raw materials and quantities in the prepared data, after removing the expansion agent, were consistent with those in step 6(a). After casting, experiments were conducted according to steps 6(b)-6(d) to obtain actual values. These actual values ​​were then compared with the shrinkage prediction values ​​in step 4 to obtain... Figure 2 .Depend on Figure 2 It can be seen that the trained model in this invention has a better prediction effect.

Claims

1. A mix design method for micro-expansion self-compacting high-strength steel-tube concrete based on artificial neural networks, characterized in that, Includes the following steps: Step 1: Collect data on the influencing parameters related to the free volume shrinkage of high-strength steel-concrete composite and the shrinkage values ​​under their influence, and construct a sample set; Step 2: Train a high-strength steel-concrete composite volume free shrinkage prediction model based on an artificial neural network using the sample set; Step 3: Obtain the required value of shrinkage deformation of high-strength steel-concrete composite pipe to be controlled, and set the theoretical data of the influencing parameters; Step 4: Use the trained high-strength steel-concrete composite volume free shrinkage prediction model to obtain the shrinkage prediction values ​​of the theoretical data of the influencing parameters; Step 5: Calculate the target value of micro-expansion based on the demand value and shrinkage prediction value, obtain the expansion value sample of high-strength steel pipe concrete under different expansion agent dosages, and combine the target value of micro-expansion to obtain the recommended expansion agent dosage. Step 6: Conduct a volume shrinkage experiment on high-strength steel-tube concrete using the expansion agent dosage and theoretical data of the influencing parameters, and determine whether the experimental results meet the shrinkage deformation requirements. If so, determine the recommended mix proportion of self-compacting micro-expansion high-strength steel-tube concrete based on the theoretical data of the influencing parameters and the recommended expansion agent dosage; otherwise, manually update the theoretical data of the influencing parameters and return to step 4. The influencing parameters include water-cement ratio, bone glue ratio, fly ash content, mineral powder content, silica fume content, curing temperature, curing humidity, and curing age; The shrinkage value is a volume shrinkage value based on a time series, with days as the basic unit; In the concrete volume shrinkage prediction model, the number of neurons in the input layer is equal to the number of influencing parameters, the number of neurons in the hidden layer is greater than the number of neurons in the input layer, and the number of neurons in the output layer is 1. The activation function between the input layer and the hidden layer is the Sigmoid function, and the transfer function between the hidden layer and the output layer is the Purelin function.

2. The mix design method for micro-expansion self-compacting high-strength steel-tube concrete based on artificial neural networks as described in claim 1, characterized in that, Step 2 includes: Step 2.1: Divide the sample set into a training set and a test set; Step 2.2: Set training parameters and use the training set to train the high-strength steel-concrete composite volume free shrinkage prediction model in multiple rounds. Step 2.3: Use the test set to test the prediction error of the high-strength steel-concrete composite volume free shrinkage prediction model after training. If it is less than the preset threshold and less than the previous prediction error, then end the process; otherwise, return to step 2.

2.

3. The mix design method for micro-expansion self-compacting high-strength steel-tube concrete based on artificial neural networks as described in claim 1, characterized in that, The LM algorithm is used as the training algorithm for the neural network model.

4. The mix design method for micro-expansion self-compacting high-strength steel-tube concrete based on artificial neural networks as described in claim 1, characterized in that, In step 6, the volume shrinkage experiment includes: Based on theoretical data of influencing parameters and suggested dosage of expansive agent, the preparation data of high-strength steel-tube concrete was determined. Based on the preparation data, four high-strength steel-tube concrete specimens were cast in steel molds. Two specimens were selected and connected to the data acquisition system in advance using the lead wires of the non-rebound tester. At this time, the TDS-530 started to read automatically and tested the shrinkage data of 0-1 days. The average shrinkage value of the two specimens at the same time was taken, and the shrinkage curve S1 of 0-1 days was plotted. When the concrete reaches 1 day of age, the steel molds of the other two specimens are removed, and the surfaces of the concrete specimens are immediately sealed with aluminum foil. The non-rebound tester is connected to the data acquisition system with lead wires to test the shrinkage data of the specimens from 1 to 28 days of age. The average shrinkage value of the two specimens at the same time is taken to plot the shrinkage curve S2 from 1 to 28 days. The two curves are combined to obtain the contraction curve for the entire process.

5. The mix design method for micro-expansion self-compacting high-strength steel-tube concrete based on artificial neural networks as described in claim 4, characterized in that, The merging process specifically involves: when the cutoff point of the contraction curve S1 and the starting point of the contraction curve S2 are the same, this value is used as the connection point to connect the two contraction curves; when the cutoff point of the contraction curve S1 and the starting point of the contraction curve S2 are different, their average value is used as the connection point, which serves as both the cutoff point of the contraction curve S1 and the starting point of the contraction curve S2, thus connecting the two contraction curves.

Images