A quantitative prediction method for ultrasonic wave velocity in concrete inside arch bridge pipes based on a parallel multi-chain DRAM model
By optimizing the iterative process using a parallel multi-chain DRAM model and a multi-process algorithm, the problems of computational complexity and low accuracy of the DRAM model in predicting the compactness of concrete inside the arch bridge pipe were solved, achieving more efficient and accurate monitoring of concrete compactness.
Patent Information
- Application Number
- CN202311225146.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-09-21
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2043-09-21
AI Technical Summary
Existing stochastic uncertainty models, such as the DRAM model, have shortcomings in computational complexity and efficiency, making it difficult to accurately predict the concrete density in steel-concrete composite arch bridges. They are also prone to getting trapped in local optima, resulting in low accuracy of the calculation results.
A parallel multi-chain DRAM model is adopted, and a parallel multi-process algorithm is used to iteratively optimize the DRAM model. Multi-core computing and adaptive iterative update algorithm are used to optimize the iterative process of the DRAM model through multi-process algorithm to find the global optimal solution.
This method improves the computational efficiency and accuracy of predicting ultrasonic wave velocity in concrete inside arch bridge pipes, overcomes the computational complexity and slow convergence of DRAM models, and achieves more efficient and accurate monitoring of concrete compactness.
Smart Images

Figure CN117250261B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of concrete-filled steel tube structure technology, and in particular to a quantitative prediction method for ultrasonic wave velocity in concrete inside arch bridge tubes based on a parallel multi-chain DRAM model. Background Technology
[0002] The synergistic performance of steel-concrete composite structures has become a key factor restricting the development of steel-concrete composite arch bridges towards larger spans. However, due to factors such as concrete material quality, pouring process, and temperature changes, the core concrete is prone to varying degrees of debonding and voiding. Furthermore, the heterogeneous nature of concrete materials and the influence of construction deviations lead to significant objective uncertainties in these influencing factors, making it impossible for existing models to accurately establish the relationship between ultrasonic wave velocity and the compactness of the concrete inside the tube.
[0003] Therefore, in order to accurately and comprehensively analyze the compactness of concrete-filled steel tubes, the Metropolis-Hastings algorithm and the adaptive Metropolis-Hastings algorithm, which consider stochastic uncertainty models, have been applied in the field of civil engineering, verifying that stochastic uncertainty models may be a suitable tool for solving problems in the field of concrete materials.
[0004] Existing stochastic uncertainty models are limited by the inherent defects of the prediction models themselves. For example, the lack of an adaptive iteration interval can cause the results to fall into local optima. The second-order acceptance probability discrimination increases the rejection rate of model iteration, which greatly reduces computational efficiency. Single-process iterative analysis cannot guarantee that the iteration process will reach the global optimum, which may lead to extremely low accuracy of the results.
[0005] Currently, some more advanced stochastic uncertainty models, such as the DRAM model, have gradually achieved successful applications in various industries due to their advantages such as fast processing speed, good adaptive iteration effect, and good control of overfitting.
[0006] However, the DRAM model suffers from computational complexity and low efficiency, and manual parameter tuning in a single process often fails to find the globally optimal solution. Therefore, how to fully utilize the number of computer processor cores to improve computational efficiency and accuracy has become a pressing issue that needs to be addressed to realize the full potential of the DRAM model. Summary of the Invention
[0007] The purpose of this invention is to provide a quantitative prediction method for ultrasonic wave velocity in concrete inside arch bridge pipes based on a parallel multi-chain DRAM model, addressing the problems existing in the prior art.
[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides a method for quantitative prediction of ultrasonic wave velocity in concrete inside an arch bridge pipe based on a parallel multi-chain DRAM model, comprising the following steps: Step 1: Obtain the engineering parameters of the concrete inside the arch bridge pipe, including the water-cement ratio of the concrete inside the pipe. Aggregate particle size Age ; Step 2: Obtain measured ultrasonic test data of the matched cubic test block in the construction site laboratory. The test data is the ultrasonic wave velocity. ; Step 3: Input the engineering parameters and measured ultrasonic wave velocity into the pre-built parallel multi-chain DRAM model; the pre-built parallel multi-chain DRAM model is a model obtained by iteratively optimizing the DRAM model using a parallel multi-process algorithm; Step 4: The parallel multi-chain DRAM model outputs a prediction result of dense, relatively dense, or not dense based on the detection data; The specific steps for building the parallel multi-chain DRAM model are as follows: S1. Obtain the engineering parameters of the concrete inside the arch bridge pipe, including the water-cement ratio of the concrete inside the pipe. Aggregate particle size Age ; S2. Obtain the measured ultrasonic test data of the matched cubic test block in the construction site laboratory. The test data is the ultrasonic wave velocity. ; S3. Input the engineering parameters as deterministic parameters into the deterministic model to obtain the prior model parameter values, and determine the following model:
[0009] In the formula, This is the predicted value of ultrasonic wave velocity. The water-cement ratio of the concrete inside the pipe. t For age period, For aggregate particle size, These are the prior model parameter values; S4. The measured ultrasonic test data of the matched cubic test block in the construction site laboratory is used as prior information and input into the parallel multi-chain DRAM model. The DRAM model is then iteratively trained using a parallel multi-process algorithm to determine the optimal parameter combination of the DRAM model, thus obtaining the DRAM model. The DRAM model includes:
[0010] In the formula, This refers to the parameter vector of the uncertainty model that is updated based on the measured ultrasonic test data of matched cubic test blocks in the construction laboratory. The objective uncertainties are caused by geometric, material, and environmental parameters, etc. To account for subjective uncertainty caused by incomplete factors, and To conform to a standard normal random distribution, and These are systematic errors caused by objective factors and subjective factors, respectively.
[0011] As a preferred technical solution of the present invention, based on the parallel multi-chain DRAM model and combined with confidence levels... The compaction performance of the concrete inside the pipe was evaluated using values of 0.50 and 0.95, respectively, with the specific settings as follows: When the test data is within the 50% confidence interval, the compactness is considered very good. When the test data is within the 95% confidence interval, the density is considered good. When the test data is outside the 95% confidence interval, the density is considered poor.
[0012] As a preferred technical solution of the present invention, step S4 performs model iterative analysis on the DRAM model using a parallel multi-process algorithm, specifically including the following steps: S41. Given the iteration parameters of the DRAM model, set the range of values for the iteration parameters; S42. Given the initial parameters of the parallel multi-process algorithm This place makes It follows a standard random normal distribution; S43. Given prior values, here we take the engineering parameters and the measured ultrasonic test data of the matched cubic test block in the construction site laboratory as the prior values for model calculation. S44. Follow the steps below until a Markov chain with a low rejection rate and a tendency to stabilize is generated. S441. Let the number of parallel chains be the number in the first iteration. If there are 8 groups, then according to step S42, the initial parameters with 8 different values are calculated; S442. During the iteration process, substitute the 8 sets of initial parameters from step S441 into the receiving probability respectively. In the process, the difference between the receiving probability and the standard normal random number is calculated and denoted as the Nth order receiving probability difference. If the Nth order receiving probability difference is greater than 0, the initial parameter is replaced with the initial parameter for the new round of iteration; if the Nth order receiving probability difference is less than 0, the receiving probability is increased to the N+1th order, and step S442 is continued, and so on, until the discrimination requirement of step S442 is met. The proposal function is a function that follows a log-normal distribution. S443. Proceed to the second iteration analysis. Based on the 8 sets of Markov chains calculated in step S442, select the Markov chain with a low rejection rate and a good steady-state distribution. Then proceed to step S442 to finally obtain a steady-state Markov chain that is infinitely close to ergodicity, stationarity, and aperiodicity.
[0013] As a further preferred technical solution of the present invention, during the parallel multi-process algorithm iterative parameter process, the Markov chain that first reaches a steady state can be used as the initial value of the remaining 7 chains for recalculation.
[0014] As a further preferred technical solution of the present invention, the iteration parameters in step S41 include: Adaptive iteration step size The value range is 10 to 100, and the increment is set to 10. Combustion period The value range is 1000~10000, and the specified increment is 1000; Acceptance probability order The value range is 4 to 8, and the specified increment is 2. initial error The value ranges from 0.01 to 0.10; Number of iterations The range is 50,000 to 100,000.
[0015] As a further preferred technical solution of the present invention, the rejection rate in step S44 is 0.4 for the first iteration analysis and 0.2 for the second iteration analysis.
[0016] As a further preferred technical solution of the present invention, the steady state described in step S443 is the Markov chain iteratively tending to a certain stable value; The prior values include Markov chains with low rejection rates and tending towards stability, obtained from on-site laboratory matching cubic specimens.
[0017] In a second aspect, the present invention also provides an electronic device, including at least one processor and a memory, and a number of processor cores matching parallel multiprocessing, and not less than eight, to realize the quantitative prediction method of ultrasonic wave velocity in concrete inside arch bridge pipes based on the parallel multi-chain DRAM model as described in any of the above.
[0018] Thirdly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the quantitative prediction method for ultrasonic wave velocity in concrete inside an arch bridge pipe based on a parallel multi-chain DRAM model as described in any of the above claims.
[0019] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. The present invention discloses a quantitative prediction method for ultrasonic wave velocity in concrete inside arch bridge pipes based on a parallel multi-chain DRAM model. This method utilizes a parallel multi-process algorithm to automatically optimize the Markov chain iteration of the DRAM model. By employing an adaptive delay rejection algorithm (DRAM model), prior information is used to update the model parameters. The parallel multi-process algorithm, with its multi-core synchronous computation and strong global search capabilities, optimizes the DRAM algorithm's Markov chain iteration optimization method. This overcomes the shortcomings of the DRAM model, such as numerous parameters, slow convergence, and the fact that manual parameter tuning is not always globally optimal. This improves the efficiency and accuracy of the DRAM model in the field of monitoring the compactness of concrete inside arch bridge pipes. 2. The present invention provides a quantitative prediction method for ultrasonic wave velocity in concrete inside arch bridge pipes based on a parallel multi-chain DRAM model. Existing methods for predicting and calculating ultrasonic wave velocity in concrete inside arch bridge pipes involve various complex parameters. This invention, through theoretical analysis and parameter sensitivity analysis, selects the water-cement ratio of the concrete inside the pipe. Aggregate particle size Age As an index system for ultrasonic wave velocity in concrete inside arch bridge pipes, all indicators are relatively easy to obtain, making it highly usable and applicable in actual engineering prediction. 3. The present invention describes a quantitative prediction method for ultrasonic wave velocity in concrete inside an arch bridge pipe based on a parallel multi-chain DRAM model. The parallel multi-chain DRAM model and the DRAM model are compared under two parameter tuning methods: default iteration (DVI) and random iteration (RI). Under DVI tuning, the DRAM model exhibits poorer iteration rejection rate, Markov chain steady-state performance, computational efficiency, and accuracy, and its calculation results are prone to getting trapped in local optima. Under RI tuning, the DRAM model gains some randomness and adaptability, improving the overall iterative analysis performance. However, due to the limited sample size and relatively small amount of iterative analysis, the computational stability of the DRAM model is significantly reduced. When dealing with large amounts of sample data and iterative analysis, the DRAM model can obtain a better Markov chain through massive computation, but the running time will increase exponentially. In contrast, the parallel process algorithm used in this paper performs well in all aspects of the model's performance indicators. It has good iteration rejection rate, Markov chain steady-state performance, computational efficiency and accuracy under both DVI and RI parameter tuning conditions. The parallel multi-chain DRAM model not only makes full use of the multi-core and multi-process computing capabilities of the processor, greatly maximizing the computer's runtime space, but also brings out the randomness of the initial iteration value and the iteration process, making the random uncertainty optimization advantage of the DRAM model more closely related to the global optimum. 4. The quantitative prediction method for ultrasonic wave velocity in concrete inside arch bridge pipes based on the parallel multi-chain DRAM model described in this invention, compared with the prediction accuracy of other stochastic uncertainty models used in the field of predicting the compactness performance of concrete inside arch bridge pipes, except for the parallel multi-chain DRAM model and DRAM model research of this invention, the others are all non-adaptive models, that is, iterative processes. The parameter tuning indicators such as covariance matrix and reception probability are not adjusted with the stability of Markov chain iteration. The MH model (Metropolis-Hastings algorithm) performs iterative analysis by using deterministic fitting parameters as the initial values of iteration. The Markov chain has high stability. However, the disadvantage of this model is that the Markov chain often exhibits a "false steady state", that is, it falls into the misconception of local optimal solutions. As a result, the calculation results are only applicable to predicting the ultrasonic wave velocity values of a small number of measuring points that are less affected by subjective and objective uncertainties. The AM model (Adaptive Metropolis-Hastings algorithm) introduces an adaptive covariance matrix for iterative update based on the MH model, which introduces the concept of randomness to a certain extent. However, compared with the MH model, the overall performance is not significantly different. The parallel multi-chain DRAM model uses a parallel multi-process algorithm to generalize and tune the initial values and iterative process of the DRAM model. It also uses an automatic optimization function to find the Markov chain that first reaches a relatively "steady state" and interacts and merges with the other chains. Compared with existing models, the computational accuracy and efficiency are greatly improved. It can achieve more accurate predictions in the shortest time. It is more advanced in terms of both the complexity of the model and its performance. 5. The present invention describes a method for quantitative prediction of ultrasonic wave velocity in concrete inside an arch bridge pipe based on a parallel multi-chain DRAM model. This method uses a parallel multi-process algorithm for iterative analysis, establishes multiple initial values for iteration based on the principle of randomness, and replaces the model parameter values of the remaining parallel chains with the model parameter values of the Markov chain that first reaches a good "steady state" during the iteration process of each parallel chain. This further reduces the waste of computational resources caused by errors in the direction of parallel chain iterations. Furthermore, by setting a threshold condition for terminating the iteration, the iterative analysis is terminated after the iterative steady state reaches the threshold, further improving computational efficiency while maintaining prediction accuracy. Attached Figure Description
[0020] Figure 1 This is a schematic diagram illustrating the fusion logic of the iterative process of a parallel multi-process algorithm. Figure 2 A schematic diagram of the optimal Markov chain iteration path for a parallel multi-chain DRAM model; Figure 3 This is a schematic diagram of the second iteration analysis of the parallel multi-chain DRAM model; Figure 4 A graph showing the parameter information of the Markov chain model in the second iteration; Figure 5 A comparison chart of performance metrics for different stochastic uncertainty models; Figure 6 A comparison chart of model-calculated values and field test values; Figure 7 This is a schematic diagram of in-pipe concrete defect identification based on 95% and 50% confidence intervals. Detailed Implementation
[0021] The present invention will now be described in detail with reference to the accompanying drawings.
[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0023] Example 1 like Figures 1 to 7 As shown, the present invention provides a method for quantitative prediction of ultrasonic wave velocity in concrete inside an arch bridge pipe based on a parallel multi-chain DRAM model, comprising the following steps: Step 1: The identification of concrete density inside the arch bridge pipe is closely related to ultrasonic acoustic parameters. In order to ensure the accurate prediction of concrete density performance inside the pipe, a scientific and reasonable index system is established with ultrasonic wave velocity as the starting point to ensure the accurate evaluation of concrete inside the arch bridge pipe.
[0024] The main prediction indices used in the existing prediction model for the compaction performance of concrete inside the arch bridge tube are shown in Table 1.
[0025] Table 1 Commonly used formulas and indices for predicting the compactness of concrete inside pipes
[0026] Table 1 f c V represents the compressive strength of concrete (MPa) and V represents the ultrasonic wave velocity (m / s). t The age (d) is the water-cement ratio (W / B). d The aggregate particle size is (mm).
[0027] Table 1 shows that the main index systems involved in predicting the ultrasonic wave velocity of concrete compaction performance inside arch bridge pipes are as follows: f c V represents the compressive strength of concrete (MPa) and V represents the ultrasonic wave velocity (m / s). t The age (d) is the water-cement ratio (W / B). d Let V be the aggregate particle size (mm), where the water-cement ratio is a microscopic manifestation of the compressive strength of concrete. Therefore, V can be used as the ultrasonic wave velocity (m / s).t The age (d) is the water-cement ratio (W / B). d A deterministic prediction model for ultrasonic wave velocity in concrete inside the arch bridge pipe is established based on aggregate particle size (mm). The deterministic model includes:
[0028] In the formula, The ultrasonic wave velocity value is obtained by detecting a non-metallic ultrasonic detector. The water-cement ratio of the concrete is taken as the 28-day compressive strength value of a cube made with the same mix proportions and materials, provided by the site laboratory; t The ages were determined by testing at 7, 14, and 28 days of curing time for the cubic test blocks. Aggregate particle size, provided by the site laboratory; These are the prior model parameter values.
[0029] Step 2: Establish a stochastic uncertainty prediction model for the ultrasonic wave velocity inside the concrete of the arch bridge pipe. The stochastic uncertainty prediction model includes:
[0030] In the formula, The parameter vector for updating the uncertainty model based on measured ultrasonic test data of matched cubic test blocks in the construction laboratory; These are objective uncertainties caused by geometric, material, and environmental parameters; To account for subjective uncertainty caused by incomplete factors; and The value follows a standard normal random distribution; and These are systematic errors caused by objective factors and subjective factors, respectively.
[0031] Step 3: Obtain the engineering parameters of the concrete inside the arch bridge pipe, including the water-cement ratio of the concrete inside the pipe. Aggregate particle size Age ; Obtain measured ultrasonic test data of a matched cubic test block in the construction site laboratory, wherein the test data is the ultrasonic wave velocity. .
[0032] The engineering parameters and measured ultrasonic wave velocity are input into a pre-built parallel multi-chain DRAM model; the pre-built parallel multi-chain DRAM model is a model obtained by iteratively optimizing the DRAM model using a parallel multi-process algorithm.
[0033] The iterative analysis of the DRAM model using a parallel multi-process algorithm specifically includes the following steps: A. Given the iteration parameters of the DRAM model (adaptive step size) Combustion period Order of reception probability Initial error Number of iterations ), set the range of values for the iteration parameters.
[0034] The iteration parameters include: A1. Adaptive Iteration Step Size The value is 100; A2. Combustion Period The value is 1000; A3. Order of acceptance probability The value is 8; A4. Initial Error The value is 0.01; A5. Number of iterations The value is 100000.
[0035] The pseudocode is as follows: Step 1. Adaptive Step Size options.adaptint=100; Step 2. Combustion Period options.burnintime = 1000; Step 3. Order of reception probability options.ntry=8; Step 4. Initial Error model.sigma2=0.01; Step 5. Number of iterations : nsimu=100000.
[0036] B. Given the initial parameters of the parallel multi-process algorithm This place makes They follow a standard random normal distribution. Taking 8 parallel chains as an example, the pseudocode is as follows: 8groups.theta0=[abs(randn()) params{1}{1,2}; abs(randn()) params{2}{1,2}; abs(randn()) params{3}{1,2}].
[0037] C. Given the initial covariance matrix Assuming that the parameters of the probability model are independent of each other, and according to An initial covariance matrix is constructed using a pre-defined initial standard deviation, and an adaptive iteration step is set. The covariance matrix is updated using the following formula:
[0038] In the formula, For small variables, s d This is the scaling factor. I d It is a d-dimensional identity matrix; Let be the covariance matrix of the historical samples. The pseudocode is shown below: Step 1. if isempty(qcov) qcov = thetasig.^2; ii = isinf(qcov)|isnan(qcov); qcov(ii) = [abs(value(ii)) 0.05].^2; qcov(qcov==0) = 1; qcov = diag(qcov); end Step2.if isempty(adascale)||adascale<=0 qcov_scale = 2.4 / sqrt(npar); else qcov_scale = adascale; end [cm,cn]=size(qcov); Step 3. if min([cm cn]) == 1 s = sqrt(qcov(parind)); R = diag(s); qcovorig = diag(qcov); qcov = diag(qcov(parind)); else qcovorig = qcov; qcov = qcov(parind,parind); R = chol(qcov); end Step 4. global invR global A_count A_count = 0; Step 5. if dodram RDR = getpar(options,'RDR',[]); Step 6. if ~isempty(RDR) for i=1:Ntry invR{i} = RDR{i}\eye(npar); end R = RDR{1}; else RDR{1} = R; invR{1} = R\eye(npar).
[0039] D. Given prior values, here we use the engineering parameters and the measured ultrasonic test data of the matched cubic test block in the on-site laboratory as the prior values for model calculation, as shown in Table 2 below: Table 2 Sample Data
[0040] E. Follow the steps below until a Markov chain with a low rejection rate and a tendency to stabilize is generated.
[0041] E1, Let the number of parallel chains be in the first iteration. If there are 8 groups, then according to step B, 8 groups of initial parameters with different values are calculated, and then from the first-order proposal distribution... Construct suggested values According to Find Where Z is a random number that follows a standard normal distribution. First-order proposed distribution standard deviation Represents the first-order proposal distribution Variables.
[0042] E2. Calculate the first-order acceptance probability. :
[0043] In the formula, The proposal function is to follow a log-normal distribution. Let be the prior distribution function.
[0044] E3. Randomly select a number from 0 to 1. If satisfied If so, then accept the suggested values for the model parameters. , that is to say If yes, proceed to step E6; otherwise, proceed to step E4.
[0045] E4. Proceed to the second-order acceptance probability discrimination and adjust the covariance matrix. From the second-order proposed distribution Construct new suggested values According to Find ,in, For the second-order proposed distribution The standard deviation, where · represents the second-order proposed distribution. Variables.
[0046]
[0047] E5. Similarly, randomly select a number from 0 to 1. If satisfied If so, then accept the suggested values for the model parameters. , that is to say If yes, proceed to step E7; otherwise, proceed to step E6.
[0048] E6, and so on, proceeding to the next... The order of acceptance probability discrimination, and the corresponding adjusted covariance matrix From the first Order of suggestion distribution Constructing new candidate values for model parameters According to Find ,in, For the first Order of suggestion distribution Standard deviation;
[0049] Similarly, a number is randomly selected from 0 to 1. If satisfied If so, then accept the suggested values for the model parameters. , that is to say Then proceed to step E6, otherwise let Repeat step E6.
[0050] E7, if If the iteration stops, then continue; otherwise, increase the iteration count and return to step E1. This is the current variable.
[0051] F. During the parallel multi-process algorithm iteration process, the optimal Markov chain is automatically achieved according to the following steps.
[0052] F1. The Markov chain that first reaches a steady state can be used as the initial value for the remaining 7 chains to be recalculated, and the fusion logic is as follows: Figure 1 As shown.
[0053] F2. After the first round of iterative analysis terminates, the optimal Markov chain path is selected based on the rejection rate and computational efficiency, such as... Figure 2 As shown.
[0054] F3. Using the best Markov chain selected in the first iteration as prior information for the second iteration analysis, a Markov chain that is infinitely close to the global optimum can be obtained, such as... Figure 3 and Figure 4 As shown.
[0055] In this embodiment, the termination condition for the iterative algorithm is set to the Markov chain reaching a preset number of iterations, 100,000. After reaching 100,000 iterations, the optimal posterior parameter combination of the parallel multi-chain DRAM model is output, such as... Figure 3 and Figure 4 As shown, after 20,000 to 40,000 first iterations, the fluctuation range, rejection rate, and mean of the model parameters all reached a stable state. The optimal search path was the fourth Markov parallel chain, with a rejection rate of 0.3056 and a computational efficiency of 75 seconds. In the second iteration, the Markov chain showed no significant change, indicating that the selected fourth Markov chain was the globally optimal iteration path. The parameter combination information of this Markov chain model is shown in Table 3.
[0056] Table 3 Optimal Model Parameter Combinations for Parallel Multichain DRAM Model
[0057] This example uses a parallel multi-chain DRAM model, combined with several other common model optimization methods, namely... The parameter tuning results of the MH model (Metropolis-Hastings algorithm) and the AM model (Adaptive Metropolis-Hastings algorithm) under stochastic iteration (RI) are compared. The initial values of the iteration, the iteration mean, and the standard deviation are (1,1,1,1), (1,1,1,1), and (1000,1000,1000,1000) respectively.
[0058] During randomized iteration (RI), the initial value, mean, and standard deviation of the iteration are all randomly selected from a standard normal distribution. The selected model evaluation metrics are rejection rate, accuracy, computation time, and Markov chain stability.
[0059] The comparison results are as follows Figure 5 As shown, by Figure 5 It can be seen that in DVI mode, the MH model has a high rejection rate and computation time, reaching 0.99 and 91 seconds respectively, and the model's computational accuracy is also poor. The AM model improves the various indicators of the MH model to some extent, but the overall performance difference is not significant, indicating that the optimization ability of the MH and AM models is poor in DVI mode. In addition, the parallel multi-chain DRAM model shows excellent performance in all aspects after the second iteration of optimization, with a low rejection rate, high accuracy, and short computation time, indicating that the parallel multi-chain DRAM model is suitable for predicting the ultrasonic wave velocity of concrete inside the arch bridge pipe.
[0060] Step 3: Use the ZBL-U5200 non-metallic ultrasonic testing instrument to test the steel pipe concrete structure. The test data includes sound time, wave velocity, wave amplitude, and frequency.
[0061] Step 4: Input the detected ultrasonic wave velocity into a stochastic uncertainty model for quantitative evaluation.
[0062] The predictive accuracy of this model was verified by using test specimens prepared on-site with the same raw materials, mix proportions, and construction grouting processes. Figure 6 As shown, the calculated values from the probability model are in good agreement with the values from the on-site test, indicating that the above-mentioned random uncertainty model has high accuracy in identifying defects in the concrete inside the CFST arch bridge.
[0063] The predictive accuracy of this model was verified by using test specimens prepared on-site with the same raw materials, mix proportions, and construction grouting processes. Figure 7 As shown, the calculated values of the stochastic uncertainty model are in good agreement with the on-site test values, indicating that the stochastic uncertainty model has high accuracy in identifying defects in the concrete inside the CFST arch bridge.
[0064] like Figure 7As shown: When the detection data is within the 50% confidence interval of the stochastic uncertainty model, the cross-sectional detection data has good density. When the test data is within the 95% confidence interval of the random uncertainty model, the concrete in this section has slight voids, but it is acceptable, and the test data of this section has good compactness. When the detection data is outside the 95% confidence interval of the stochastic uncertainty model, the concrete in that section is severely voided, the detection data of that section has poor density, and an early warning is issued.
[0065] Step 5: Treat the voided areas of the entire bridge that were previously exposed to the warning, which can be reinforced by drilling and grouting.
[0066] In summary, this invention provides a quantitative prediction method for ultrasonic wave velocity in concrete inside arch bridge pipes based on a parallel multi-chain DRAM model. This method proposes a parallel multi-process algorithm to automatically optimize the Markov chain of the DRAM model. Through an adaptive iterative update algorithm (DRAM model), the model parameters are updated and optimized one by one based on prior information. Then, the excellent global search capability and flexibility of the multi-process algorithm are utilized for automatic Markov chain optimization. This overcomes the shortcomings of the DRAM model, such as computational complexity, slow convergence, and poor stability, thus improving the practicality of the DRAM model in the field of monitoring the compactness of concrete inside arch bridge pipes.
[0067] Example 2 An electronic device comprising at least one processor and a memory, and a number of processor cores matching parallel multiprocessing, and not less than eight, to implement the quantitative prediction method for ultrasonic wave velocity in concrete inside arch bridge pipes based on a parallel multi-chain DRAM model as described in any of the above.
[0068] As a preferred embodiment of this invention, the electronic device may include a processor, a memory, and may also include one or more of a multimedia component, an input / output (I / O) interface, and a communication component.
[0069] The processor controls the overall operation of the electronic device to complete all or part of the steps in the above-mentioned quantitative prediction method of ultrasonic wave velocity in concrete inside arch bridge pipes based on the parallel multi-chain DRAM model.
[0070] Memory is used to store various types of data to support the operation of the electronic device. This data may include, for example, instructions for any application or method used to operate on the electronic device, as well as application-related data. Memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.
[0071] The multimedia component may include a screen and an audio component, wherein the screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals; for example, the audio component may include a microphone for receiving external audio signals, the received audio signals may be further stored in memory or transmitted via a communication component; the audio component may also include at least one speaker for outputting audio signals.
[0072] I / O interfaces provide interfaces between the processor and other interface modules, such as keyboards, mice, buttons, etc.; these buttons can be virtual buttons or physical buttons.
[0073] The communication component is used for wired or wireless communication between the electronic device and other devices; wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G or 5G, or one or more combinations thereof, and the corresponding communication component may include: Wi-Fi module, Bluetooth module, NFC module, mobile communication module.
[0074] As a preferred embodiment, the electronic device may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the above-described method for quantitative prediction of ultrasonic wave velocity in concrete inside arch bridge pipes based on a parallel multi-chain DRAM model.
[0075] In addition, the computer-readable storage medium provided in this embodiment can be the memory including program instructions, which can be executed by the processor of an electronic device to complete the above-mentioned method for quantitative prediction of ultrasonic wave velocity in concrete inside an arch bridge pipe based on a parallel multi-chain DRAM model.
[0076] Example 3 A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the quantitative prediction method for ultrasonic wave velocity in concrete inside an arch bridge pipe based on a parallel multi-chain DRAM model as described in any of the preceding claims.
[0077] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for quantitative prediction of ultrasonic wave velocity in concrete inside arch bridge pipes based on a parallel multi-chain DRAM model, characterized in that, Includes the following steps: Step 1: Obtain the engineering parameters of the concrete inside the arch bridge pipe, including the water-cement ratio of the concrete inside the pipe. Aggregate particle size Age ; Step 2: Obtain measured ultrasonic test data of the matched cubic test block in the construction site laboratory. The test data is the ultrasonic wave velocity. ; Step 3: Input the engineering parameters and measured ultrasonic wave velocity into the pre-built parallel multi-chain DRAM model; the pre-built parallel multi-chain DRAM model is a model obtained by iteratively optimizing the DRAM model using a parallel multi-process algorithm; Step 4: The parallel multi-chain DRAM model outputs a prediction result of dense, relatively dense, or not dense based on the detection data; The specific steps for building the parallel multi-chain DRAM model are as follows: S1. Obtain sample data of engineering parameters for the concrete inside the arch bridge pipe, including the water-cement ratio of the concrete inside the pipe. Aggregate particle size Age ; S2. Obtain sample data of measured ultrasonic testing data of matched cubic test blocks in the construction site laboratory. The test data is the ultrasonic wave velocity. ; S3. Input the engineering parameters as deterministic parameters into the deterministic model to obtain the prior model parameter values, and determine the following model: In the formula, This is the predicted value of ultrasonic wave velocity. The water-cement ratio of the concrete inside the pipe. t For age period, For aggregate particle size, These are the prior model parameter values; S4. The measured ultrasonic test data of the matched cubic test block in the construction site laboratory is used as prior information and input into the parallel multi-chain DRAM model. The DRAM model is then iteratively trained using a parallel multi-process algorithm to determine the optimal parameter combination of the DRAM model, thus obtaining the DRAM model. The DRAM model includes: In the formula, This refers to the parameter vector of the uncertainty model that is updated based on the measured ultrasonic test data of matched cubic test blocks in the construction laboratory. The objective uncertainties are caused by geometric, material, and environmental parameters. To account for subjective uncertainty caused by incomplete factors, and To conform to a standard normal random distribution, and These are systematic errors caused by objective factors and subjective factors, respectively.
2. The method for quantitative prediction of ultrasonic wave velocity in concrete inside an arch bridge pipe based on a parallel multi-chain DRAM model according to claim 1, characterized in that, Based on the parallel multi-chain DRAM model, combined with confidence levels The compaction performance of the concrete inside the pipe was evaluated using values of 0.50 and 0.95, respectively, with the specific settings as follows: When the test data is within the 50% confidence interval, the compactness is considered very good. When the test data is within the 95% confidence interval, the density is considered good. When the test data is outside the 95% confidence interval, the density is considered poor.
3. The method for quantitative prediction of ultrasonic wave velocity in concrete inside an arch bridge pipe based on a parallel multi-chain DRAM model according to claim 1, characterized in that, Step S4 involves iterative analysis of the DRAM model using a parallel multi-process algorithm, specifically including the following steps: S41. Given the iteration parameters of the DRAM model, set the range of values for the iteration parameters; S42. Given the initial parameters of the parallel multi-process algorithm This place makes It follows a standard random normal distribution; S43. Given prior values, here we take the engineering parameters and the measured ultrasonic test data of the matched cubic test block in the construction site laboratory as the prior values for model calculation. S44. Follow the steps below until a Markov chain with a low rejection rate and a tendency to stabilize is generated. S441. Let the number of parallel chains be the number in the first iteration. If there are 8 groups, then according to step S42, the initial parameters with 8 different values are calculated; S442. During the iteration process, substitute the 8 sets of initial parameters from step S441 into the receiving probability respectively. In the process, the difference between the receiving probability and the standard normal random number is calculated and denoted as the Nth order receiving probability difference. If the Nth order receiving probability difference is greater than 0, the initial parameter is replaced with the initial parameter for the new round of iteration; if the Nth order receiving probability difference is less than 0, the receiving probability is increased to the N+1th order, and step S442 is continued, and so on, until the discrimination requirement of step S442 is met. The proposal function is to follow a log-normal distribution. This is the prior distribution function; S443. Proceed to the second iteration analysis. Based on the 8 sets of Markov chains calculated in step S442, select the Markov chain with a low rejection rate and a good steady-state distribution. Then proceed to step S442 to finally obtain a steady-state Markov chain that is infinitely close to ergodicity, stationarity, and aperiodicity.
4. The method for quantitative prediction of ultrasonic wave velocity in concrete inside an arch bridge pipe based on a parallel multi-chain DRAM model according to claim 3, characterized in that, During the parallel multi-process algorithm's parameter iteration process, the Markov chain that first reaches a steady state can be used as the initial value for the remaining 7 chains to be recalculated.
5. The method for quantitative prediction of ultrasonic wave velocity in concrete inside an arch bridge pipe based on a parallel multi-chain DRAM model according to claim 3, characterized in that, The iteration parameters mentioned in step S41 include: Adaptive iteration step size The value range is 10 to 100, and the increment is set to 10. Combustion period The value range is 1000~10000, and the specified increment is 1000; Acceptance probability order The value range is 4 to 8, and the specified increment is 2. initial error The value ranges from 0.01 to 0.10; Number of iterations The range is 50,000 to 100,000.
6. The method for quantitative prediction of ultrasonic wave velocity in concrete inside an arch bridge pipe based on a parallel multi-chain DRAM model according to claim 3, characterized in that, The rejection rate mentioned in step S44 is 0.4 for the first iteration analysis and 0.2 for the second iteration analysis.
7. The method for quantitative prediction of ultrasonic wave velocity in concrete inside an arch bridge pipe based on a parallel multi-chain DRAM model according to claim 3, characterized in that, The steady state described in step S443 is when the Markov chain iteratively tends to a certain stable value; The prior values include Markov chains with low rejection rates and tending towards stability, obtained from on-site laboratory matching cubic specimens.
8. An electronic device comprising at least one processor and a memory, and a number of processor cores matching parallel multiprocessing, and not less than eight, to implement the method for quantitative prediction of ultrasonic wave velocity in concrete inside arch bridge pipes based on a parallel multi-chain DRAM model as described in any one of claims 1-7.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the quantitative prediction method for ultrasonic wave velocity in concrete inside arch bridge pipes based on a parallel multi-chain DRAM model as described in any one of claims 1-7.
Citation Information
Patent Citations
Method for predicting service life of concrete after high temperature under multiple actions
CN114965967A
Method and equipment for detecting and evaluating internal compaction state of concrete-filled steel tube arch bridge
CN115341462A
