Foam concrete construction process monitoring method and system based on multi-source data fusion

Abstract

The application discloses a multi-source data fusion foam concrete construction process monitoring method and system, relates to the technical field of concrete construction, and comprises the following steps: obtaining a standard response data set of foam concrete, including associated storage of raw material proportioning parameters and performance index parameters; constructing and training a mixed layered performance prediction model, including a shared prediction layer based on integrated learning and a plurality of independent output layers connected with the shared prediction layer; combining a multi-objective optimization algorithm, taking the optimization of compressive strength and cost as the target, constructing an evaluation function for global optimization, and obtaining a target mix proportion scheme; collecting real-time process data associated with the target mix proportion scheme, inputting the real-time process data into the mixed layered performance prediction model, and obtaining a prediction result; comparing the prediction result with a preset design target, and generating an adjustment instruction for adjusting the proportioning parameters of subsequent mixing batches. The method solves the problem that the construction situation cannot be dynamically adjusted in real time in the prior art, and the performance of different batches of foam concrete fluctuates greatly.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of concrete construction, in particular to a foamed concrete construction process monitoring method and system based on multi-source data fusion. BACKGROUND

[0002] With the development of the construction industry, foamed concrete is increasingly widely used in various projects. However, the traditional construction monitoring method mainly relies on manual experience and limited parameter monitoring, which is difficult to comprehensively and accurately reflect the actual performance and construction quality of foamed concrete, and cannot dynamically adjust according to real-time construction conditions, resulting in large fluctuations in the performance of foamed concrete of different batches. SUMMARY

[0003] The foamed concrete construction process monitoring method and system based on multi-source data fusion provided by the embodiments of the present application solve the technical problem that the construction conditions cannot be dynamically adjusted in real time in the prior art, resulting in large fluctuations in the performance of foamed concrete of different batches.

[0004] The technical solution of the present application to solve the above technical problem is as follows:

[0005] In a first aspect, the present application provides a foamed concrete construction process monitoring method based on multi-source data fusion, which comprises:

[0006] obtaining a standard response data set of foamed concrete, wherein the standard response data set comprises associatedly stored raw material proportioning parameters and performance index parameters;

[0007] based on the standard response data set, constructing and training a mixed hierarchical performance prediction model, wherein the mixed hierarchical performance prediction model comprises a shared prediction layer based on ensemble learning and a plurality of independent output layers connected to the shared prediction layer;

[0008] combining a multi-objective optimization algorithm, taking the optimization of compressive strength and cost as the target, and combining the mixed hierarchical performance prediction model to construct an evaluation function for global optimization, thereby obtaining a target mix proportion scheme;

[0009] collecting real-time process data associated with the target mix proportion scheme in a real-time construction process, and inputting the real-time process data into the mixed hierarchical performance prediction model to obtain a performance prediction result;

[0010] comparing the performance prediction result with a preset design target, and generating an adjustment instruction for adjusting the mix proportioning parameters of the subsequent mixing batch according to the comparison result.

[0011] In a second aspect, the present application provides a foamed concrete construction process monitoring system based on multi-source data fusion, which comprises:

[0012] The data acquisition module is configured to obtain a standard response data set of the foam concrete, wherein the standard response data set comprises stored raw material proportioning parameters and performance index parameters;

[0013] The model training module is configured to construct and train a mixed hierarchical performance prediction model based on the standard response data set, wherein the mixed hierarchical performance prediction model comprises a shared prediction layer based on ensemble learning and a plurality of independent output layers connected to the shared prediction layer.

[0014] The scheme acquisition module is configured to combine a multi-objective optimization algorithm to optimize the compressive strength and cost as the target, and to combine the mixed hierarchical performance prediction model to construct an evaluation function for global optimization to obtain a target mix proportioning scheme.

[0015] The performance prediction module is configured to collect real-time process data associated with the target mix proportioning scheme in a real-time construction process, and to input the real-time process data into the mixed hierarchical performance prediction model to obtain a performance prediction result.

[0016] The instruction adjustment module is configured to compare the performance prediction result with a preset design target, and to generate an adjustment instruction for adjusting the proportioning parameters of a subsequent mixing batch according to the comparison result.

[0017] The present application provides one or more technical solutions, at least having the following technical effects or advantages:

[0018] The present application provides a multi-source data fusion foam concrete construction process monitoring method and system, which first uses a multi-source data fusion method to fully integrate raw material proportioning parameters and performance index parameters and other information to construct and train a mixed hierarchical performance prediction model. According to the real-time collected process data, the performance of the foam concrete can be accurately predicted to provide a scientific basis for the proportioning adjustment in the construction process. By collecting and analyzing the real-time process data associated with the target mix proportioning scheme, possible problems in the construction process can be found. When the performance prediction result deviates from the preset design target, the system generates an adjustment instruction to optimize the proportioning parameters of the subsequent mixing batch, thereby effectively reducing the performance fluctuation of different batches of foam concrete. In addition, the evaluation function is constructed by combining a multi-objective optimization algorithm for global optimization, which not only ensures that the foam concrete reaches the ideal compressive strength, but also achieves good results in cost control. The construction process ensures the quality while maximizing the economic benefits.

[0019] Through the above technical solutions, the multi-source data fusion foam concrete construction process monitoring method and system provide a scientific, efficient and reliable solution for the construction of foam concrete in the construction industry, which helps to improve the construction quality and efficiency and promotes the sustainable development of the construction industry. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort on the basis of these drawings.

[0021] Figure 1 is a flow diagram of a multi-source data fusion foam concrete construction process monitoring method provided by the embodiments of the present application;

[0022] Figure 2 is a structural diagram of a multi-source data fusion foam concrete construction process monitoring system provided by the embodiments of the present application.

[0023] In the drawings, the components represented by the respective reference numerals are described as follows:

[0024] The data acquisition module 11, the model training module 12, the scheme acquisition module 13, the performance prediction module 14, and the instruction adjustment module 15. DETAILED DESCRIPTION

[0025] The embodiments of the present application provide a multi-source data fusion foam concrete construction process monitoring method and system, which are used to solve the technical problem that the construction situation cannot be dynamically adjusted in real time in the prior art, resulting in large performance fluctuation of different batches of foam concrete.

[0026] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort fall within the scope of protection of the present application.

[0027] In the description of the present application, the terms "first", "second" are used only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0028] In the description of the present application, the term "for example" is used to indicate "serving as an example, instance, or illustration." Any embodiment described as "for example" in the present application is not necessarily to be construed as preferred or advantageous over other embodiments. The following description is presented to enable any person skilled in the art to make and use the present application. In the following description, for the purpose of explanation, details are set forth. It is apparent to those skilled in the art that the present application can be practiced without the use of these specific details. In other instances, well-known structures and processes are not elaborated in detail in order not to obscure the description of the present application. Thus, the present application is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.

[0029] Embodiment one, as shown in the present application, provides a multi-source data fusion foam concrete construction process monitoring method, comprising: Figure 1

[0030] S10: obtaining a standard response data set of foam concrete, wherein the standard response data set comprises associatedly stored raw material proportioning parameters and performance index parameters;

[0031] In the present application, by collecting historical foam concrete construction project records, a standard response data set of foam concrete is obtained, which includes raw material proportioning parameters used in different projects and corresponding performance index parameters. The raw material proportioning parameters include various key factors affecting the performance of foam concrete, and the performance index parameters reflect the quality and characteristics of foam concrete.

[0032] The raw material proportioning parameters at least include cementitious material composition, gas-forming agent dosage, water-material ratio, and gas-forming regulator dosage. The performance index at least includes compressive strength, dry density, fluidity, and expansion height.

[0033] In the present application, the obtained standard response data set comprises associatedly stored raw material proportioning parameters and performance index parameters, and the raw material proportioning parameters at least include cementitious material composition, gas-forming agent dosage, water-material ratio, and gas-forming regulator dosage. The cementitious material composition determines the setting and hardening performance of foam concrete, and different cementitious materials will produce different strength and durability in the hydration reaction process; the gas-forming agent dosage affects the number and size of pores in foam concrete, and appropriate gas-forming agent dosage can make foam concrete have good heat preservation and sound insulation performance; the water-material ratio affects the fluidity and strength of foam concrete, and too large water-material ratio may reduce the strength of concrete, and too small water-material ratio will affect the construction performance. The gas-forming regulator dosage can adjust the speed and stability of the gas-forming process, and ensure the uniformity of foam concrete.

[0034] ​The performance indicators include at least compressive strength, dry density, fluidity and expansion height. The compressive strength is an indicator for measuring the load-bearing capacity of the foam concrete structure, and is related to the safety and reliability of the building; the dry density reflects the porosity and quality of the foam concrete, and has a direct impact on the heat preservation and sound insulation effect; the fluidity reflects the operability of the foam concrete in the construction process, and appropriate fluidity can ensure that the concrete uniformly fills the formwork; and the expansion height reflects the effect of the gas evolution process and affects the volume stability of the foam concrete.

[0035] Through the associated storage and analysis of the raw material ratio parameters and the performance indicator parameters, a rich and accurate data basis is provided for subsequent construction and training of a mixed layered performance prediction model, so that the monitoring and quality control of the foam concrete construction process are better achieved.

[0036] Specifically, step S10 in the method includes:

[0037] M raw material ratio parameters and N performance indicator parameters are obtained, and a multi-factor orthogonal test is designed correspondingly to obtain an orthogonal test table, wherein M and N are positive integers greater than or equal to 1;

[0038] A multi-factor orthogonal test is performed based on the orthogonal test table to obtain a multi-factor orthogonal test result, wherein the multi-factor orthogonal test result includes a first raw material ratio parameter set and a first performance indicator parameter set;

[0039] Based on the multi-factor orthogonal test result, the performance sensitivity of the M raw material ratio parameters to the N performance indicator parameters is analyzed and determined;

[0040] According to the performance sensitivity, a preliminary value range of the M raw material ratio parameters is obtained by extrapolation.

[0041] In the embodiment of the application, first, M raw material ratio parameters and N performance indicator parameters are obtained, and a multi-factor orthogonal test is designed to obtain an orthogonal test table, wherein M and N are positive integers greater than or equal to 1. After obtaining the orthogonal test table, a multi-factor orthogonal test is performed to obtain a test result including a first raw material ratio parameter set and a first performance indicator parameter set. The multi-factor orthogonal test result is analyzed to determine the performance sensitivity of each raw material ratio parameter to different performance indicator parameters, i.e., to determine the degree of influence of each parameter on each performance indicator.

[0042] Then, a preliminary value range of M raw material proportioning parameters is obtained according to performance sensitivity extrapolation. For example, if the cementitious material composition has a high performance sensitivity to the compressive strength and dry density of the foam concrete, a preliminary value range of the cementitious material composition is extrapolated according to the ideal range of the compressive strength and dry density and in combination with the test results, so as to provide a targeted parameter range for subsequent tests and optimization, avoid invalid attempts in a too large parameter space, and improve test efficiency and accuracy.

[0043] Further, a standard response data set of the foam concrete is obtained, wherein the standard response data set includes the stored raw material proportioning parameters and performance index parameters, and further includes:

[0044] In combination with the preliminary value range, a second raw material proportioning parameter set of the M raw material proportioning parameters is determined to perform a single-factor test, so as to obtain a second performance index parameter set.

[0045] Based on the single-factor test results, a paste thickening curve and a gas evolution curve are extracted, and a thickening-gas evolution matching degree is calculated correspondingly.

[0046] An association between the thickening-gas evolution matching degree and the second performance index parameter set is established.

[0047] The second raw material proportioning parameter set in the single-factor test results is taken as an input factor, the thickening-gas evolution matching degree is taken as a process parameter, and the second performance index parameter set is taken as a terminal performance, and the standard response data set is obtained by performing structured storage in combination with the association.

[0048] In the embodiments of the application, after the preliminary value range of the M raw material proportioning parameters is determined, a second raw material proportioning parameter set is determined in combination with the value range to perform a single-factor test. In the single-factor test, one raw material proportioning parameter is changed each time, and other parameters are kept unchanged, so as to test the individual influence of each parameter on the performance of the foam concrete, and further obtain a second performance index parameter set.

[0049] Secondly, based on the single-factor test results, a paste thickening curve and a gas evolution curve are extracted, the paste thickening curve reflects the change of the consistency of the foam concrete paste in the mixing and setting process, and the gas evolution curve reflects the amount of gas generated by the gas evolution agent at different times. By analyzing the two curves, the physical changes of the foam concrete in the molding process are understood. A thickening-gas evolution matching degree is calculated correspondingly, and the thickening-gas evolution matching degree is an index for measuring the coordination of the thickening speed and the gas evolution speed of the foam concrete in the molding process. If the thickening speed is too fast and the gas evolution speed is too slow, the air holes may not be formed sufficiently, which affects the heat preservation and sound insulation performance of the foam concrete; on the contrary, if the gas evolution speed is too fast and the thickening speed is too slow, the air holes may be broken, which reduces the strength of the concrete.

[0050] Again, the correlation between the thickening-gas generation matching degree and the second performance index parameter set is established, the second raw material ratio parameter set in the single factor test result is taken as the input factor, the thickening-gas generation matching degree is taken as the process parameter, the second performance index parameter set is taken as the terminal performance, the structured storage is performed combined with the correlation, so as to obtain the standard response data set. Based on the above structured storage mode, the data is more orderly, which provides a comprehensive and accurate data basis for subsequent construction and training of the mixed hierarchical performance prediction model, and helps to more accurately predict the performance of the foam concrete, realizes effective monitoring and quality control of the construction process.

[0051] S20: based on the standard response data set, a mixed hierarchical performance prediction model is constructed and trained, the mixed hierarchical performance prediction model includes a shared prediction layer based on ensemble learning and multiple independent output layers connected to the shared prediction layer;

[0052] In the embodiment of the application, first, based on the standard response data set, the mixed hierarchical performance prediction model is built and trained by artificial neural network, because the standard response data set contains the associated storage of raw material ratio parameters and performance index parameters. The mixed hierarchical performance prediction model includes a shared prediction layer based on ensemble learning and multiple independent output layers connected to the shared prediction layer.

[0053] Further, the shared prediction layer based on ensemble learning preliminarily predicts the performance of the foam concrete, integrates different algorithms and models to mine potential information in the data, and improves the accuracy and stability of the prediction. The multiple independent output layers can individually predict different performance indicators, each independent output layer focuses on a specific performance indicator, and analyzes the relationship between the indicator and the raw material ratio parameters in detail.

[0054] Specifically, step S20 in the method includes:

[0055] According to the performance sensitivity, the shared prediction layer is initialized and configured;

[0056] The second raw material ratio parameter set is taken as the training input, the thickening-gas generation matching degree is taken as the supervision, multiple base models in the shared prediction layer are trained respectively, and are integrated combined with a voting mechanism;

[0057] The thickening-gas generation matching degree is taken as the input, the second performance index parameter set is taken as the supervision, and N independent output layers are constructed and trained for N performance index parameters;

[0058] The input ends of the multiple independent output layers are connected to the output end of the shared prediction layer respectively, and the mixed hierarchical performance prediction model is generated.

[0059] In the embodiments of the present application, first, the shared prediction layer is initialized and configured according to the performance sensitivity, which reflects the degree of influence of each raw material ratio parameter on the performance index parameter, so as to set the initial parameters of the shared prediction layer.

[0060] Secondly, the second raw material ratio parameter set is determined after preliminary value range screening, and the second raw material ratio parameter set is taken as the training input, which contains parameters that affect the performance of foam concrete. The thickening-gas evolution matching degree is taken as the supervision to train multiple base models in the shared prediction layer. The thickening-gas evolution matching degree is a key indicator for measuring the coordination of the thickening speed and the gas evolution speed of the slurry in the foam concrete forming process, which provides accurate supervision information for the training of the base model. In the training process, each base model learns different features and rules, and then integrates through the voting mechanism. The voting mechanism integrates the prediction results of multiple base models to improve the accuracy and stability of the prediction of the shared prediction layer.

[0061] Among them, the voting mechanism makes decisions according to the prediction results of multiple base models, for example, the simple majority voting method, that is, the most frequently occurring prediction result is selected as the final prediction result. Through the voting mechanism to integrate multiple base models, the advantages of each base model are utilized, the limitations of a single model are reduced, and the reliability of the shared prediction layer for preliminary prediction of the performance of foam concrete is improved.

[0062] Then, N independent output layers are constructed and trained for N performance index parameters, taking the thickening-gas evolution matching degree as the input and the second performance index parameter set as the supervision. Each independent output layer focuses on a specific performance index to analyze the relationship between the index and the raw material ratio parameter.

[0063] Finally, the input ends of multiple independent output layers are connected with the output end of the shared prediction layer to generate a hybrid layered performance prediction model. The output of the shared prediction layer provides preliminary prediction information for the independent output layer, and the independent output layer makes more detailed prediction on this basis, so that the entire hybrid layered performance prediction model can accurately predict the performance indexes of foam concrete according to the raw material ratio parameters.

[0064] Exemplarily, the hybrid layered performance prediction model is built and trained based on an artificial neural network.

[0065] First, data collection, the second raw material ratio parameter set and the second performance index parameter set, and the thickening-gas evolution matching degree data are extracted from the standard response data set as training data.

[0066] Secondly, model building, design shared prediction layer and independent output layer structure. Shared prediction layer can adopt multiple different types of base model, such as decision tree, support vector machine, etc., through ensemble learning to integrate base model. Independent output layer is constructed according to the number of performance indicators, and then the parameters of the shared prediction layer are initialized according to the performance sensitivity.

[0067] Finally, model training, using the second raw material ratio parameter set as input, the thickening-gas generation matching degree as supervision signal, training multiple base models of shared prediction layer. Using Adam optimizer and mean square error (MSE) loss function to build training framework, setting batch size to 32, total training rounds to 50, and introducing early stopping mechanism (patience=5), when the validation set loss does not appear for 5 consecutive rounds, the training process is automatically terminated, and the trained multiple base models in the shared prediction layer are obtained. Each base model continuously adjusts its own weight and bias to minimize the error between the prediction result and the supervision signal. The voting mechanism is used to integrate the prediction results of multiple base models, for example, simple majority voting, to get the final output of the shared prediction layer.

[0068] With thickening-gas generation matching degree as input, the second performance indicator parameter set as supervision signal, N independent output layers are trained respectively. Each independent output layer is aimed at a specific performance indicator, and adjusts its own parameters through back propagation algorithm to improve the prediction accuracy of the performance indicator. Using Adam optimizer and mean square error (MSE) loss function to build training framework, setting batch size to 32, total training rounds to 50, and introducing early stopping mechanism (patience=5), when the validation set loss does not appear for 5 consecutive rounds, the training process is automatically terminated, and the trained multiple independent output layers are connected with the output end of the shared prediction layer to form a complete hybrid layered performance prediction model. In the connection process, ensure that data can be correctly transmitted and processed between different layers.

[0069] Among them, according to the performance sensitivity, the shared prediction layer is initialized and configured, including:

[0070] Obtain the performance sensitivity of M raw material ratio parameters, form M performance sensitivity groups, and each performance sensitivity group includes N performance sensitivities;

[0071] Based on the results of the multi-factor orthogonal experiment, range and variance analysis are performed on N performance indicator parameters, and corresponding normalization is performed.

[0072] Taking the normalized range as the first weight coefficient and the normalized variance as the second weight coefficient, traversing M performance sensitivity groups, weighting and fusing N performance sensitivities in each performance sensitivity group to obtain the ratio parameter weight coefficient.

[0073] According to the ratio parameter weight coefficient, a base model scale ratio corresponding to the M material ratio parameters is determined, and a plurality of base models of the shared prediction layer are initialized according to the base model scale ratio.

[0074] In the embodiments of the present application, first, the performance sensitivity of M material ratio parameters is obtained, and M performance sensitivity groups are formed, each group containing N performance sensitivities, which reflect the influence degree of each material ratio parameter on different performance indicators. Based on the results of the multi-factor orthogonal test, range and variance analysis is performed on the N performance indicator parameters.

[0075] Range analysis can find the primary and secondary order of the influence of each factor on the performance indicators, and variance analysis can judge the significance of the influence of each factor on the performance indicators. The analysis results are normalized to make the analysis results of different indicators comparable.

[0076] Secondly, the normalized range is taken as the first weight coefficient, and the normalized variance is taken as the second weight coefficient, and the N performance sensitivities in the M performance sensitivity groups are weighted and fused to obtain the ratio parameter weight coefficient. The ratio parameter weight coefficient comprehensively considers the influence of range and variance on performance sensitivity and reflects the importance of each material ratio parameter.

[0077] Then, according to the ratio parameter weight coefficient, a base model scale ratio corresponding to the M material ratio parameters is determined. The base model scale ratio reflects the importance and proportion of each material ratio parameter in the shared prediction layer, and the base model scale corresponding to the parameter with high importance is relatively large. According to the base model scale ratio, a plurality of base models of the shared prediction layer are initialized to improve the accuracy and stability of prediction.

[0078] For example, assuming that the value of M is 3, representing three material ratio parameters of cementitious material amount, foaming agent amount and water-binder ratio; and N is 2, representing two performance indicator parameters of compressive strength and dry density.

[0079] The performance sensitivity of three material ratio parameters is obtained, and three performance sensitivity groups are formed, each group containing two performance sensitivities. For example, the performance sensitivities of cementitious material amount on compressive strength and dry density are 0.6 and 0.4 respectively; the performance sensitivities of foaming agent amount on the two performance indicators are 0.3 and 0.7 respectively; and the performance sensitivities of water-binder ratio on the two performance indicators are 0.5 and 0.5 respectively.

[0080] Assuming that the normalized range is 0.6 and 0.4 respectively, and the normalized variance is 0.7 and 0.3 respectively. The normalized range is taken as the first weight coefficient, and the normalized variance is taken as the second weight coefficient, and the performance sensitivities in each performance sensitivity group are weighted and fused.

[0081] For the performance sensitivity group of the amount of cementitious material, the weight coefficient of the proportioning parameter obtained after weighted fusion is (0.6*0.6+0.7*0.4)=0.64;

[0082] For the performance sensitivity group of the amount of foaming agent, the weight coefficient of the proportioning parameter obtained is (0.3*0.6+0.7*0.4)=0.46;

[0083] For the performance sensitivity group of the water-binder ratio, the weight coefficient of the proportioning parameter obtained is (0.5*0.6+0.5*0.7)=0.65.

[0084] According to the proportioning parameter weight coefficient, the base model scale ratio corresponding to the amount of cementitious material, the amount of foaming agent and the water-binder ratio is determined. For example, assuming that the calculated base model scale ratios are 0.3, 0.2 and 0.5 respectively. Then, the multiple base models in the shared prediction layer are initialized according to the base model scale ratios. For example, there are three base models in the shared prediction layer, i.e., decision tree, support vector machine and neural network. Different base models are allocated resources and parameters are adjusted according to the base model scale ratios. The base model corresponding to the water-binder ratio occupies a larger proportion in the shared prediction layer, so as to improve the accuracy and stability of the performance prediction of foam concrete.

[0085] Further, the multiple base models in the shared prediction layer are trained respectively with the second raw material proportioning parameter set as the training input and the thickening-gas evolution matching degree as the supervision, including:

[0086] Based on the single-factor categories of the second raw material proportioning parameter set, the second raw material proportioning parameter set is divided into M single parameter sets;

[0087] The corresponding thickening-gas evolution matching degrees are matched for the M single parameter sets to form M single parameter sample sets;

[0088] The iterative random exchange between the M single parameter sample sets is performed with the proportioning parameter weight coefficient as the exchange probability and the reciprocal of the proportioning parameter weight coefficient as the exchange ratio, and the multiple base models in the shared prediction layer are trained based on the iterative random exchange result.

[0089] In the embodiments of the present application, first, the second raw material proportioning parameter set is divided into M single parameter sets based on the single-factor categories of the second raw material proportioning parameter set. Each single parameter set corresponds to one raw material proportioning parameter. The corresponding thickening-gas evolution matching degrees are matched for the M single parameter sets to form M single parameter sample sets. The thickening-gas evolution matching degree is used as supervision information to provide feedback for each single parameter sample set, so that the training process is more targeted.

[0090] Secondly, the M single-parameter sample sets are iteratively and randomly exchanged with each other, taking the matching parameter gravity coefficient as the exchange probability and taking the reciprocal of the matching parameter gravity coefficient as the exchange ratio. In the process of iterative and random exchange, each single-parameter sample set has a certain probability of data exchange with other sample sets, and the exchange ratio is determined by the reciprocal of the matching parameter gravity coefficient.

[0091] Finally, based on the results of the iterative and random exchange, the multiple base models in the training shared prediction layer are divided into different numbers of M groups, corresponding to M raw material matching parameters. During the training process, each base model learns and adjusts according to the exchanged sample set, constantly optimizes its own weights and biases, and improves the prediction ability of the foam concrete performance. Different base models can learn different features and rules, integrate the base models through ensemble learning, fully exert the advantages of each base model, reduce the limitations of a single model, and thus improve the reliability and accuracy of the shared prediction layer in the preliminary prediction of the foam concrete performance.

[0092] Illustratively, assuming that M takes the value of 4, representing four different raw material matching parameters, four single-parameter sample sets are formed. In the process of iterative and random exchange, a certain single-parameter sample set may exchange data with other sample sets with a certain exchange probability, and the exchange ratio is determined according to the reciprocal of the matching parameter gravity coefficient. After multiple iterations and random exchanges, the multiple base models in the shared prediction layer are trained using the exchanged sample set, each base model constantly adjusts its own parameters to adapt to new sample data, and finally improves the prediction performance of the entire shared prediction layer.

[0093] S30: In combination with a multi-objective optimization algorithm, taking the optimization of the compressive strength and the cost as the target, an evaluation function is constructed by combining the mixed hierarchical performance prediction model to perform global optimization, and a target matching scheme is obtained;

[0094] In the embodiments of the present application, in combination with a multi-objective optimization algorithm, taking the compressive strength and the cost as the optimization target, an evaluation function is constructed by using the mixed hierarchical performance prediction model to perform global optimization, and the optimal solution is found under the condition of meeting various constraint conditions.

[0095] The evaluation function is constructed according to the compressive strength and the cost, the output result of the mixed hierarchical performance prediction model is compared with the target value, and the comprehensive evaluation value is calculated. Through the way of global optimization, the scheme that makes the evaluation function value optimal is found in all possible raw material matching combinations.

[0096] Specifically, the step S30 in the method comprises:

[0097] initializing a population, wherein each individual in the population represents a candidate matching scheme;

[0098] A multi-objective evaluation function is constructed, a first objective term of the multi-objective evaluation function being a performance prediction value output based on the mixed hierarchical performance prediction model, and a second objective term being a material cost calculated based on the alternative mix proportion scheme;

[0099] Constraint conditions are configured based on the foam concrete demand information of the target scenario.

[0100] The multi-objective optimization algorithm is run for iterative optimization, the fitness of each individual in the population is calculated based on the multi-objective evaluation function and the constraint conditions, and the population is updated through selection, crossover and mutation operations until a termination condition is met, and the one with the optimal multi-objective evaluation function value is output as the target mix proportion scheme.

[0101] In the embodiments of the application, first, an initial population is generated, a certain number of individuals are randomly generated, each individual representing an alternative mix proportion scheme, covering different raw material ratio combinations.

[0102] Secondly, a multi-objective evaluation function is constructed, a first objective term being a performance prediction value output by the mixed hierarchical performance prediction model, such as a performance indicator of compressive strength, and a second objective term being a material cost calculated based on the alternative mix proportion scheme.

[0103] Then, constraint conditions are configured based on the foam concrete demand information of the target scenario. The constraint conditions include a minimum compressive strength requirement of the foam concrete, a maximum cost limit, specific working performance requirements, etc. By setting the constraint conditions, it is ensured that the scheme obtained in the optimization process meets the requirements of actual applications.

[0104] After that, the multi-objective optimization algorithm is run for iterative optimization, in each iteration, the fitness of each individual in the population is calculated based on the multi-objective evaluation function and the constraint conditions. The fitness value reflects the degree of excellence of the individual in meeting the objective function and the constraint conditions.

[0105] Through the selection operation, individuals with higher fitness are retained and individuals with lower fitness are eliminated, so that the population evolves in a better direction; the crossover operation simulates gene exchange in biological genetics, exchanges part of the information of two excellent individuals to generate new individuals and increase the diversity of the population; the mutation operation randomly changes some genes of the individual to prevent the algorithm from falling into a local optimal solution.

[0106] In the iteration process, the population is updated by selection, crossover and mutation operations until the termination condition is met. The termination condition can be that the maximum number of iterations is reached, or the fitness value of the population no longer improves significantly within a certain number of iterations. When the termination condition is met, the individual with the optimal multi-objective evaluation function value is output as the target mix ratio scheme. The target mix ratio scheme achieves a good balance between the performance and cost of foam concrete, meeting the performance requirements of foam concrete for the target scene and ensuring the reasonableness of the cost, which can provide scientific and effective mix ratio guidance for actual foam concrete construction, improving construction quality and economic benefits.

[0107] S40: Collect real-time process data associated with the target mix ratio scheme during real-time construction, and input the real-time process data into the mixed layered performance prediction model to obtain a performance prediction result;

[0108] In the embodiments of the present application, during real-time construction, real-time process data associated with the target mix ratio scheme is collected by sensors and data acquisition equipment, including the actual input amount of raw materials, mixing time, environmental temperature and humidity, etc. The real-time process data is input into the mixed layered performance prediction model, and the model will predict the performance of the foam concrete according to the learned rules and characteristics during previous training, to obtain a performance prediction result.

[0109] The performance prediction result includes multiple performance indicators of the foam concrete, such as compressive strength, dry density, and consistency-gas matching degree, etc. By predicting the performance indicators, the performance status of the foam concrete under the current construction conditions can be understood in a timely manner, and it can be judged whether it meets the expected quality requirements. If the prediction result shows that some performance indicators do not meet the requirements, adjustment measures can be taken in a timely manner. For example, if the compressive strength is insufficient, the amount of cementitious material can be appropriately increased; if the consistency-gas matching degree is not ideal, the amount of foaming agent or the mixing process can be adjusted.

[0110] S50: Compare the performance prediction result with the preset design target, and generate an adjustment instruction for adjusting the mix ratio parameters of the subsequent mixing batch according to the comparison result.

[0111] In the embodiments of the present application, the performance prediction result is compared with the preset design target in detail, and the preset design target is usually determined according to specific engineering requirements and quality standards.

[0112] If there is a deviation between the performance prediction result and the preset design target, the system will generate a corresponding adjustment instruction according to the type and degree of the deviation. The adjustment instruction has clear pertinence for different performance indicator deviations.

[0113] Exemplarily, when the predicted compressive strength is lower than the design target, the adjustment instruction can require an increase in the amount of cementitious material, while also possibly involving fine-tuning of the water-cement ratio to ensure that the working performance of the foamed concrete is maintained while the strength is improved. If the density is not up to standard, the amount of foaming agent can need to be adjusted or the grading of the raw materials can need to be optimized.

[0114] When dealing with deviations in the consistency-gas generation match, the adjustment instruction will pay more attention to the adjustment of the mixing process and time. If the match is not ideal, the mixing time can be appropriately extended, or the speed and manner of mixing can be changed to promote better coordination of the consistency and gas generation processes.

[0115] In summary, compared with the prior art, the present application realizes the multi-source data fusion of the foamed concrete construction process monitoring through the mechanism data layer, the intelligent core layer and the monitoring application layer. First, the raw material ratio parameters and performance index parameters are obtained, a multi-factor orthogonal experiment is designed, the results of the multi-factor orthogonal experiment are obtained, the performance sensitivity of the raw material ratio parameters to the performance index parameters is analyzed, a second raw material ratio parameter set of the raw material ratio parameters is determined based on the obtained preliminary value range, a single-factor experiment is performed, and a refined database is established based on the experimental results. Then, the artificial neural network is used for accurate performance prediction, a multi-objective optimization algorithm is combined, a hybrid layered performance prediction model is combined to construct an evaluation function for global optimization. Finally, the prediction result is compared with the preset design target, and an adjustment instruction for adjusting the subsequent mixing batch ratio parameters is generated. The uncertainty in the construction process is effectively reduced, and the quality stability of the foamed concrete is improved.

[0116] In summary, the embodiments of the present application have at least the following technical effects:

[0117] The foam concrete construction process monitoring method based on multi-source data fusion provided by the embodiment of the application first uses the multi-source data fusion method to fully integrate various information such as raw material proportioning parameters and performance index parameters, and constructs and trains a mixed hierarchical performance prediction model. According to the real-time process data collected, the performance of the foam concrete is accurately predicted, and a scientific basis is provided for the proportioning adjustment in the construction process. By collecting and analyzing the real-time process data associated with the target mixing proportion scheme, possible problems in the construction process can be found. When the performance prediction result deviates from the preset design target, the system generates an adjustment instruction to optimize the proportioning parameters of the subsequent mixing batch, thereby effectively reducing the performance fluctuation of different batches of foam concrete. In addition, the evaluation function is constructed by combining the multi-objective optimization algorithm for global optimization, which not only ensures that the foam concrete reaches the ideal compressive strength, but also achieves good results in cost control. The construction process maximizes economic benefits while ensuring quality. Through the above technical solutions, the foam concrete construction process monitoring method based on multi-source data fusion provides a scientific, efficient and reliable solution for the construction of foam concrete in the construction industry, which helps to improve the construction quality and efficiency and promotes the sustainable development of the construction industry.

[0118] Embodiment two, as shown in Figure 2 based on the same inventive concept of the foam concrete construction process monitoring method based on multi-source data fusion provided by embodiment one, the embodiment of the application also provides a foam concrete construction process monitoring system based on multi-source data fusion, comprising:

[0119] The data acquisition module 11 is used to obtain a standard response data set of foam concrete, wherein the standard response data set comprises associated stored raw material proportioning parameters and performance index parameters.

[0120] The model training module 12 is used to construct and train a mixed hierarchical performance prediction model based on the standard response data set, wherein the mixed hierarchical performance prediction model comprises a shared prediction layer based on ensemble learning and a plurality of independent output layers connected to the shared prediction layer.

[0121] The scheme acquisition module 13 is used to combine a multi-objective optimization algorithm to optimize the compressive strength and cost as the target, combine the mixed hierarchical performance prediction model to construct an evaluation function for global optimization, and obtain a target mixing proportion scheme.

[0122] The performance prediction module 14 is used to collect real-time process data associated with the target mixing proportion scheme in the real-time construction process, and input the real-time process data into the mixed hierarchical performance prediction model to obtain a performance prediction result.

[0123] The instruction adjustment module 15 is configured to compare the performance prediction result with a preset design target, and generate an adjustment instruction for adjusting a subsequent mixing batch proportioning parameter according to a comparison result.

[0124] In one embodiment, the data acquisition module 11 is specifically configured to:

[0125] Obtain M raw material proportioning parameters and N performance index parameters, and design a multi-factor orthogonal test corresponding thereto to obtain an orthogonal test table, wherein M and N are positive integers greater than or equal to 1;

[0126] Perform a multi-factor orthogonal test based on the orthogonal test table to obtain a multi-factor orthogonal test result, wherein the multi-factor orthogonal test result includes a first raw material proportioning parameter set and a first performance index parameter set;

[0127] Based on the multi-factor orthogonal test result, analyze and determine the performance sensitivity of the M raw material proportioning parameters to the N performance index parameters;

[0128] According to the performance sensitivity, extrapolate to obtain a preliminary value range of the M raw material proportioning parameters.

[0129] The raw material proportioning parameters at least include cementitious material composition, gas generating agent content, water material ratio, and gas generating regulator content. The performance index at least includes compressive strength, dry density, fluidity, and expansion height.

[0130] Further, in one application embodiment, a standard response data set of the foam concrete is obtained, wherein the standard response data set includes the associated stored raw material proportioning parameters and performance index parameters, and further includes:

[0131] Determine a second raw material proportioning parameter set of the M raw material proportioning parameters in combination with the preliminary value range to perform a single-factor test, and obtain a second performance index parameter set;

[0132] Based on the single-factor test result, extract a paste thickening curve and a gas generation curve, and correspondingly calculate a thickening-gas generation matching degree;

[0133] Establish an association between the thickening-gas generation matching degree and the second performance index parameter set;

[0134] Take the second raw material proportioning parameter set in the single-factor test result as an input factor, the thickening-gas generation matching degree as a process parameter, and the second performance index parameter set as a terminal performance, and combine the association to perform structured storage, thereby obtaining the standard response data set.

[0135] In one embodiment, the model training module 12 is specifically configured to:

[0136] initialize and configure the shared prediction layer according to the performance sensitivities;

[0137] train multiple base models in the shared prediction layer respectively with the second set of raw material proportioning parameters as training input and the densification-gas evolution matching degree as supervision, and integrate in combination with a voting mechanism;

[0138] construct and train N independent output layers respectively for N performance index parameters with the densification-gas evolution matching degree as input and the second set of performance index parameters as supervision;

[0139] connect the input ends of the multiple independent output layers respectively with the output end of the shared prediction layer to generate the hybrid layered performance prediction model.

[0140] Further, in one application embodiment, initializing and configuring the shared prediction layer according to the performance sensitivities comprises:

[0141] obtain the performance sensitivities of M raw material proportioning parameters to form M performance sensitivity groups, each of which includes N performance sensitivities;

[0142] based on the multiple-factor orthogonal test results, perform range and variance analysis on N performance index parameters and normalize them correspondingly;

[0143] take the normalized range as a first weight coefficient and the normalized variance as a second weight coefficient, perform weighted fusion on N performance sensitivities in each of the M performance sensitivity groups to obtain proportioning parameter weight coefficients;

[0144] determine the base model scale ratio corresponding to the M raw material proportioning parameters according to the proportioning parameter weight coefficients, and initialize multiple base models of the shared prediction layer according to the base model scale ratio.

[0145] Further, in one application embodiment, training multiple base models in the shared prediction layer respectively with the second set of raw material proportioning parameters as training input and the densification-gas evolution matching degree as supervision comprises:

[0146] based on single-factor categories of the second set of raw material proportioning parameters, divide the second set of raw material proportioning parameters into M single parameter sets;

[0147] match the densification-gas evolution matching degrees corresponding to the M single parameter sets to form M single parameter sample sets;

[0148] The iteration random exchange is performed between M single-parameter sample sets according to an exchange probability of the matching parameter gravity coefficient and an exchange ratio of an inverse of the matching parameter gravity coefficient, and a plurality of base models in the shared prediction layer are trained based on a result of the iteration random exchange.

[0149] In one embodiment, the scheme obtaining module 13 is specifically configured to:

[0150] Initialize a population, wherein each individual in the population represents an alternative mix proportion scheme;

[0151] Construct a multi-objective evaluation function, wherein a first objective term of the multi-objective evaluation function is a performance prediction value output by the hybrid hierarchical performance prediction model, and a second objective term is a material cost calculated based on the alternative mix proportion scheme;

[0152] Configure a constraint condition based on demand information of the target scenario for foam concrete;

[0153] Run the multi-objective optimization algorithm for iteration optimization, calculate the fitness of each individual in the population based on the multi-objective evaluation function and the constraint condition, and update the population through selection, crossover and mutation operations until a termination condition is met, and output a one with an optimal multi-objective evaluation function value as the target mix proportion scheme.

[0154] It should be noted that the above sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. Moreover, the above describes specific embodiments of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0155] The above only describes the preferred embodiments of the present application and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0156] The present application is only an exemplary description of the present application, and should be considered as covering any and all modifications, changes, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the scope of the present application. Thus, if these modifications and changes of the present application belong to the scope of the present application and its equivalents, the present application intends to include these modifications and changes.

Claims

1. A method for monitoring the construction process of foamed concrete using multi-source data fusion, characterized in that, include: Obtain M raw material ratio parameters and N performance index parameters, and design a multi-factor orthogonal experiment accordingly to obtain an orthogonal experiment table, where M and N are both positive integers greater than or equal to 1; A multi-factor orthogonal experiment is performed based on the orthogonal experimental table to obtain the multi-factor orthogonal experimental results, wherein the multi-factor orthogonal experimental results include a first raw material ratio parameter set and a first performance index parameter set; Based on the results of the multi-factor orthogonal experiment, the performance sensitivity of the M raw material ratio parameters to the N performance index parameters was analyzed and determined. Based on the performance sensitivity, the preliminary value range of the raw material ratio parameters of item M is obtained by extrapolation; Based on the preliminary value range, a second set of raw material ratio parameters for the raw material ratio parameters described in item M is determined, and a single-factor experiment is conducted to obtain the second set of performance index parameters. Based on the results of single-factor experiments, the slurry thickening curve and gas generation curve were extracted, and the thickening-gas generation matching degree was calculated accordingly. Establish the correlation between the thickening-gas generation matching degree and the second set of performance index parameters; Using the second raw material ratio parameter set in the single-factor test results as input factors, the thickening-gas generation matching degree as process parameters, and the second performance index parameter set as terminal performance, the data is structured and stored in combination with the correlation relationship to obtain a standard response dataset; Based on the standard response dataset, a hybrid hierarchical performance prediction model is constructed and trained. The hybrid hierarchical performance prediction model includes a shared prediction layer based on ensemble learning and multiple independent output layers connected to the shared prediction layer. By combining a multi-objective optimization algorithm with the goal of optimizing compressive strength and cost, and by constructing an evaluation function based on the hybrid stratified performance prediction model, global optimization is performed to obtain the target mix design. Real-time process data associated with the target mix design is collected during real-time construction, and the real-time process data is input into the hybrid stratified performance prediction model to obtain performance prediction results; The performance prediction results are compared with the preset design target, and adjustment instructions are generated based on the comparison results to adjust the mixing ratio parameters of subsequent batches. The process of constructing and training a hybrid hierarchical performance prediction model includes: Obtain the performance sensitivity of the raw material ratio parameters mentioned in item M, and form M performance sensitivity groups, each of which includes N performance sensitivity items; Based on the results of the multi-factor orthogonal experiment, range and variance analysis were performed on all N performance index parameters, and the corresponding normalization was performed. Using the normalized range as the first weighting coefficient and the normalized variance as the second weighting coefficient, the M performance sensitivity groups are traversed, and the N performance sensitivities in each performance sensitivity group are weighted and fused to obtain the ratio parameter intensity coefficient. Based on the density coefficient of the ratio parameter, determine the base model scale ratio corresponding to the raw material ratio parameter of item M, and initialize multiple base models of the shared prediction layer according to the base model scale ratio, and allocate resources and adjust parameters for different base models according to the base model scale ratio; Using the second raw material ratio parameter set as training input and the thickening-gas generation matching degree as supervision, multiple base models in the shared prediction layer are trained respectively, and integrated by combining a voting mechanism; Using the thickening-gas generation matching degree as input and the second performance index parameter set as supervision, N independent output layers are constructed and trained for N performance index parameters respectively; The hybrid hierarchical performance prediction model is generated by connecting the input terminals of the multiple independent output layers to the output terminal of the shared prediction layer.

2. The method for monitoring the construction process of foamed concrete by multi-source data fusion as described in claim 1, characterized in that, Using the second raw material ratio parameter set as training input and the thickening-gas generation matching degree as supervision, multiple base models in the shared prediction layer are trained respectively, including: Based on the single-factor categories of the second raw material ratio parameter set, the second raw material ratio parameter set is divided into M single-parameter sets; Match the corresponding thickening-gas generation matching degree for each of the M single parameter sets to form M single parameter sample sets; Using the intensity coefficient of the ratio parameter as the exchange probability and the reciprocal of the intensity coefficient of the ratio parameter as the exchange ratio, perform iterative random exchanges among the M single-parameter sample sets, and train multiple base models in the shared prediction layer based on the results of the iterative random exchanges.

3. The method for monitoring the construction process of foamed concrete using multi-source data fusion as described in claim 2, characterized in that, Combining a multi-objective optimization algorithm, with the goal of optimizing compressive strength and cost, and using the aforementioned hybrid stratified performance prediction model to construct an evaluation function for global optimization, the target mix design is obtained, including: Initialize the population, where each individual in the population represents an alternative matching scheme; A multi-objective evaluation function is constructed, wherein the first objective term of the multi-objective evaluation function is the performance prediction value output based on the hybrid hierarchical performance prediction model, and the second objective term is the material cost calculated based on the alternative mix proportion scheme. Configure constraints based on the foamed concrete demand information of the target scenario; The multi-objective optimization algorithm is run iteratively to find the best fit. The fitness of each individual in the population is calculated based on the multi-objective evaluation function and the constraints. The population is updated through selection, crossover, and mutation operations until the termination condition is met. The optimal multi-objective evaluation function value is then output as the target fit ratio scheme.

4. The method for monitoring the construction process of foamed concrete by multi-source data fusion as described in claim 1, characterized in that, The raw material proportioning parameters include at least the composition of the cementitious material, the dosage of the gas-generating agent, the water-to-material ratio, and the dosage of the gas-generating regulator.

5. The method for monitoring the construction process of foamed concrete by multi-source data fusion as described in claim 1, characterized in that, The performance indicators include at least compressive strength, dry density, flowability, and expansion height.

6. A multi-source data fusion-based monitoring system for the construction process of foamed concrete, characterized in that, For performing the method according to any one of claims 1-5, comprising: The data acquisition module is used to acquire the standard response dataset of foamed concrete, wherein the standard response dataset includes associated and stored raw material proportioning parameters and performance index parameters; The model training module is used to construct and train a hybrid hierarchical performance prediction model based on the standard response dataset. The hybrid hierarchical performance prediction model includes a shared prediction layer based on ensemble learning and multiple independent output layers connected to the shared prediction layer. The scheme acquisition module is used to combine a multi-objective optimization algorithm with the goal of optimizing compressive strength and cost, and to construct an evaluation function based on the hybrid hierarchical performance prediction model to perform global optimization and obtain the target mix ratio scheme. The performance prediction module is used to collect real-time process data associated with the target mix design during real-time construction, and input the real-time process data into the hybrid layered performance prediction model to obtain the performance prediction results. The instruction adjustment module is used to compare the performance prediction results with the preset design target, and generate adjustment instructions for adjusting the mixing ratio parameters of subsequent batches based on the comparison results.

Images