A cement compressive strength prediction method and device based on an integrated semi-supervised width regression system

Abstract

The application discloses a cement compressive strength prediction method and device based on an integrated semi-supervised width regression system, and the method comprises the following steps: determining cement compressive strength influencing factors; preparing cement samples according to the determined influencing factor corresponding attribute values and carrying out hydration reactions under standard curing conditions; testing and recording the cement samples according to a cement compressive strength determination method; training an integrated self-training semi-supervised width regression method by using labeled data and unlabeled data training sets, and predicting test data; if the test average absolute error is less than an expected value, the corresponding predicted strength is obtained; otherwise, the model structure is adjusted until a satisfactory prediction error is obtained. The application effectively solves the problem of scarcity of labeled data, utilizes the potential of unlabeled data to enhance the estimation of cement compressive strength, and introduces an integrated learning technology to increase the stability and robustness of the model.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of cement compressive strength prediction, and particularly relates to a cement compressive strength prediction method and device based on an integrated semi-supervised width regression system. BACKGROUND

[0002] Cement plays a vital role in the construction industry, and its quality is crucial to ensuring safety in production and daily life. Therefore, its quality detection is crucial in cement production. Cement compressive strength is one of the key parameters for evaluating cement quality, and its accurate estimation is crucial for ensuring the integrity and durability of structures in quality inspection and the construction industry.

[0003] In cement manufacturing, the detection of cement compressive strength still relies on traditional, time-consuming, and destructive measurements, which require a high level of technical expertise. Although technicians can obtain some accurate data, a large portion of the data shows serious errors due to limited technical proficiency. Worse still, traditional measurement methods are destructive, and when measurement errors are large or records are not timely, a large number of specimens have no cement compressive strength records. This situation hinders the improvement of cement quality and the research of high-performance cement. Therefore, it is urgent to develop more efficient detection techniques.

[0004] Intelligent computing techniques, such as neural networks, have provided efficient tools for the prediction of cement compressive strength. However, current research faces some challenges. The limited learning ability of shallow neural networks pushes the model to deeper architectures, while deep networks introduce a large number of parameters that need to be optimized, resulting in a large amount of time consumption. To address this challenge, an efficient width-based compressive strength estimation system is proposed, which improves learning speed and estimation efficiency.

[0005] Despite significant progress in detection methods, traditional methods mainly rely on supervised learning techniques, which require a large amount of labeled data. However, obtaining cement compressive strength values is often challenging. During the collection process, some samples may not have corresponding strength values due to equipment failure, human error, or filtering, etc. For example, during testing, samples may be damaged or encounter other problems, resulting in failure. Therefore, labeled data is still scarce, while there is a considerable amount of unlabeled data. This limits the performance of existing machine learning-based methods. On the other hand, due to the uneven level of testing technicians, there is not a small difference in the measured compressive strength of cement samples under the same formula, which greatly interferes with model learning. SUMMARY

[0006] The main purpose of the present application is to overcome the shortcomings and deficiencies of the prior art, provide a cement compressive strength prediction method and device based on an integrated semi-supervised width regression system, which effectively solves the problem of scarcity of labeled data by using semi-supervised learning algorithm and enhances the estimation of cement compressive strength by using the potential of unlabeled data.

[0007] In order to achieve the above purpose, the present application adopts the following technical scheme:

[0008] In a first aspect, the present application provides a cement compressive strength prediction method based on an integrated semi-supervised width regression system, comprising the following steps:

[0009] S1, determining the cement compressive strength influencing factors and preparing the corresponding cement samples, and obtaining the sample data set; the sample data set includes labeled data and unlabeled data;

[0010] S2, initializing multiple BLS models: using existing small amount of labeled data to train BLS model, and obtaining multiple labeled regression models;

[0011] S3, predicting the cement compressive strength value by using the labeled regression model, assigning the cement compressive strength value to the unlabeled data, and obtaining the pre-labeled data;

[0012] S4, according to the pre-labeled data of multiple labeled regression models, statistical decision is made, the average value of prediction value is calculated, the average prediction value is obtained, the average prediction value is screened by using the set threshold value, and the screened average prediction value is used as the pseudo label of the corresponding sample, and the pseudo label data is obtained, and the remaining samples are still unlabeled data;

[0013] S5, jump back to S2, use the selected pseudo label data and labeled data to iteratively train and update the labeled regression model, and the process is repeated until the set condition is met, the training is terminated, and the final labeled data set is obtained;

[0014] S6, using the final labeled data set to train the final decision BLS regression model;

[0015] S7, using the trained final decision BLS model to predict the cement compressive strength of new cement samples.

[0016] As a preferred technical scheme, the preparation of the corresponding cement sample and the acquisition of the sample data set are as follows:

[0017] According to the determined cement compressive strength influencing factor corresponding attribute value, the cement sample is prepared, and the hydration reaction is carried out under standard curing conditions;

[0018] The cement sample is tested according to the cement compressive strength determination method, and if the test result meets the set standard, the compressive strength is taken as the label of the cement sample, and the labeled data is obtained, otherwise the cement sample has no corresponding cement compressive strength, and unlabeled data is obtained.

[0019] As a preferred technical solution, the calculation of the compressive strength value is as follows:

[0020]

[0021] wherein, represents a linear mapping function, ζ(·) represents a nonlinear mapping function, X u represents unlabeled data, W e and β e represents a mapping feature connection weight, β h represents an enhanced feature node connection weight, and W represents a connection weight.

[0022] As a preferred technical solution, the average prediction value and the deviation value of the prediction value are calculated, the deviation value is compared with the preset deviation threshold value, and the average prediction value lower than the deviation threshold value is recorded as an effective prediction; for a sample, if multiple labeled regression models predict the sample, the corresponding average prediction value is obtained, and the average prediction value meeting the statistical probability is taken as the pseudo label of the sample.

[0023] As a preferred technical solution, the average prediction value and the deviation value of the prediction value s are calculated, and the deviation value is compared with the preset deviation threshold value, as follows:

[0024]

[0025] wherein, the error threshold ∈ ∈ [0, 1], and the probability threshold δ ∈ [0.5, 1].

[0026] As a preferred technical solution, the pseudo label is determined by the following formula:

[0027]

[0028] wherein, represents an average prediction value, Pr() represents a screening function, and δ represents a probability threshold.

[0029] As a preferred technical solution, the BLS model further comprises a feature mapping node number and an enhanced node number.

[0030] The trained BLS model is used to predict the cement compressive strength of a new cement sample, specifically:

[0031] determining the cement compressive strength influencing factors;

[0032] forming mapping feature nodes Z n ≡ [Z1, Z2, …, Z n ], wherein the i-th group of mapping features Z i is calculated, and each group of features contains k mapping features, as follows:

[0033]

[0034] wherein X u represents unlabeled data, W ei and β ei both represent the connection weights of the i-th group of mapping features Z i , and n represents the total number of mapping nodes;

[0035] performing nonlinear mapping on the mapping feature nodes Z n to obtain enhanced feature nodes H m ≡ [H1, H2, …, H m ], wherein the j-th group of enhanced features H j , as follows:

[0036] H j ≡ ζ (Z n W hj + β hj ), j = 1, 2, …, m

[0037] wherein the connection weights W hj and β hj are randomly generated, ζ (·) is a nonlinear activation function, and m represents the total number of enhanced nodes;

[0038] combining the mapping features Z i and the enhanced features H j to form new cement compressive strength identification features, and connecting them to the model output layer S.

[0039] In a second aspect, the present application further provides a cement compressive strength prediction system based on an integrated semi-supervised width regression system, which is applied to the cement compressive strength prediction method based on the integrated semi-supervised width regression system and comprises

[0040] In a third aspect, the present application provides an electronic device, which comprises:

[0041] at least one processor; and

[0042] a memory in communication connection with the at least one processor; wherein

[0043] The memory stores computer program instructions executable by the at least one processor, and the computer program instructions are executed by the at least one processor to enable the at least one processor to execute the cement compressive strength prediction method based on the integrated semi-supervised width regression system.

[0044] In a fourth aspect, the present application provides a computer readable storage medium, which stores a program, and the program is executed by a processor to implement the cement compressive strength prediction method based on the integrated semi-supervised width regression system.

[0045] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0046] The present application designs an efficient non-destructive detection method based on machine learning, introduces unlabeled data to improve prediction effect and reduce prediction error, is more suitable for actual application scenarios, and avoids waste of data resources; at the same time, the present application uses an integrated learning technology to use multiple width learning regression models to vote and label pseudo labels of cement data, avoids the problem that different test results are directly different in the cement compressive strength scene, can improve the prediction effect of the model, and ensures the stability and robustness of the model. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed 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 labor.

[0048] Figure 1 The flowchart of the cement compressive strength prediction method based on the integrated semi-supervised width regression system of the embodiments of the present application is shown in the figure.

[0049] Figure 2 The structure diagram of the BLS model of the embodiments of the present application is shown in the figure.

[0050] Figure 3 The structure diagram of the cement compressive strength prediction system based on the integrated semi-supervised width regression system of the embodiments of the present application is shown in the figure.

[0051] Figure 4 The structure diagram of the electronic device of the embodiments of the present application is shown in the figure. DETAILED DESCRIPTION

[0052] In the following, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the drawings in the embodiments of the present application, so that those skilled in the art can better understand 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 work fall within the scope of protection of the present application.

[0053] Reference to "an embodiment" in this application means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of other embodiments. It is expressly understood that the described embodiments of the application are merely example embodiments and that one of ordinary skill in the art would be able to make other embodiments of the application without the exercise of inventive faculty.

[0054] Broad Learning System (BLS) is a single-layer incremental neural network proposed by Professor Chen Junlong of the University of Macau in 2017 based on Random Vector Functional-Link Neural Network (RVFLNN) and Single-Layer Feedforward Neural Network (SLFN). Compared with traditional deep network model, this model has the advantages of fast, simple, and supporting incremental online model update while ensuring a certain accuracy.

[0055] Please refer to Figure 1 The embodiment provides a cement compressive strength prediction method based on an integrated semi-supervised width regression system, including the following steps:

[0056] S1, determine the cement compressive strength influencing factors and prepare the corresponding cement samples, and obtain the sample data set; the sample data set includes labeled data and unlabeled data.

[0057] Firstly, the types of cement compressive strength influencing factors need to be determined. In the embodiment, the cement compressive strength influencing factors include chemical components and physical properties. Specifically, the following factors can be used:

[0058] Burned vector, silicon dioxide, iron oxide, aluminum oxide, calcium oxide, magnesium oxide, sulfur trioxide, free calcium oxide, liquid phase amount, lime saturation coefficient, saturation coefficient KH- of free calcium oxide, silicon rate SM, aluminum rate IM, tricalcium silicate C3S, dicalcium silicate C2S, tricalcium aluminate C3A, tetracalcium aluminoferrite C2AF, fly ash FA, and specific surface area SSA.

[0059] After the cement compressive strength influencing factors are determined, the cement needs to be pretreated, and the steps are as follows:

[0060] S11, prepare cement samples according to the determined attribute values of the influencing factors of cement compressive strength, and carry out hydration reaction under standard curing conditions. The hydration reaction of cement is a complex and important process, which has a crucial influence on the performance and strength of concrete. In the experiment, appropriate cement types, mix proportions and water quality conditions need to be selected according to the specific situation to ensure the quality and performance of concrete and simulate the real application environment.

[0061] S12, test the cement samples according to the cement compressive strength determination method; if the test result meets the laboratory standard, record the compressive strength of the cement sample to form the labeled data; otherwise, the sample has no corresponding cement compressive strength, which is regarded as unlabeled data.

[0062] Then, the integrated self-training semi-supervised width regression method is trained using the labeled data and unlabeled data, and the test data is predicted, which will be described in detail in steps S2-S3.

[0063] S2, initialize multiple BLS models: train BLS models using existing small amount of labeled data to obtain multiple labeled regression models.

[0064] Please refer to Figure 2 In step S2, the parameters and structure of the BLS model are first initialized, including the number of feature mapping nodes, the number of enhanced nodes, etc., and then a small amount of labeled data {X l ,S l} is used to train N bls labeled regression models F(θ) for pre-labeling and further voting decision. In particular, in this embodiment, the output data of the labeled regression model is a continuous regression value, so the output is one-dimensional.

[0065] S3, predict the cement compressive strength value using the labeled regression model, assign the cement compressive strength value to the unlabeled data, and obtain the pre-labeled data.

[0066] After the training is completed according to the current labeled data in step S3, all labeled regression models assign regression values to the unlabeled data. For unlabeled data X u , each width learning model F(θ) corresponds to a regression value S. The predicted regression value of the regression model is directly used as the "pre-label" y of the unlabeled data, as follows:

[0067]

[0068] wherein, represents a linear mapping function, ζ(·) represents a nonlinear mapping function, X u represents unlabeled data, W e and β e represent mapping feature connection weights, βh represents the enhanced feature node connection weight, and W represents the connection weight.

[0069] It is emphasized that if the data is normalized before use, the reverse normalization operation is required here.

[0070] S4, according to the pre-labeled data of multiple labeling regression models, statistical decision is made, the average value of the predicted value is calculated, the average predicted value is obtained, the average predicted value is screened by using the set threshold, and the screened average predicted value is used as the pseudo label of the corresponding sample, and the pseudo label data is obtained, and the remaining sample is still unlabeled data.

[0071] To avoid the performance degradation caused by the false pseudo-labeled samples to the labeled data set {X l , l}(t), the present embodiment adopts a strategy to comprehensively decide the voting results of multiple labeling regression models to ensure the accuracy of the auxiliary training data . The output of the regression task is a continuous real value, therefore, the average value is used for the result integration of multiple regression models. Not all average values can be used as the final pseudo label, therefore, a suitable threshold needs to be set to determine the reliability of the predicted value when generating the pseudo label. For a sample x i , first, the average predicted value is calculated according to the predicted output of all labeling regression models, as follows:

[0072]

[0073] Among them, represents the average predicted value, s i represents the predicted value.

[0074] This average value represents the collective prediction result of N bls basic models on the sample. Next, the deviation value of the average predicted value and the predicted value is calculated, the deviation value is compared with the preset deviation threshold, and the average predicted value lower than the deviation threshold is recorded as the effective prediction. For a sample, if multiple labeling regression models predict the sample and obtain the corresponding average predicted value, the average predicted value meeting the statistical probability is used as the pseudo label of the sample, otherwise, the sample still has no pseudo label. The generation of the pseudo label is as follows:

[0075]

[0076] Among them, the error threshold ∈ ∈ [0, 1], the probability threshold δ ∈ [0.5, 1], and Pr() represents the screening function.

[0077] S5, jump back to S2, update the tagging regression model by iteratively training with pseudo-label data and labeled data, repeat the process until the set conditions are met, the training is terminated, and the final labeled data set is obtained;

[0078] S6, train the final decision BLS regression model using the final labeled data set.

[0079] Repeat the iterative process, where each iteration step includes generating pseudo-labels using the current model, followed by training a new model combining these pseudo-labels and original labeled data. Iteration continues until certain stopping conditions are met, such as unlabelled data being exhausted or reaching a preset maximum number of iterations. Multiple base classifiers are only used to obtain effective pseudo-label data and do not participate in the final decision. Therefore, after the iteration is completed, a final BLS regression model is trained using all labeled data. In the testing or verification phase, this model will be used to predict regression values. S7, use the trained final decision BLS model to predict the compressive strength of new cement samples.

[0080] Finally, in the prediction stage, the following steps will be carried out:

[0081] Determine the factors affecting the compressive strength of cement;

[0082] Form mapping feature nodes Z through random mapping n ≡[Z1,Z2,…,Z n ], where the i-th group of mapping features Z i is calculated, and each group of features contains k mapping features, as follows:

[0083]

[0084] Nonlinearly map the mapping feature nodes Z n to obtain enhanced feature nodes H m ≡[H1,H2,…,H m ], where the j-th group of enhanced features H j is as follows:

[0085] H j ≡ζ(Z n W hj +β hj ),j=1,2,…,m

[0086] Where, the connection weight W hj and β hj are randomly generated, and ζ(·) is a nonlinear activation function;

[0087] Map the mapping features Z i and the enhanced features H jThe new cement compressive strength identification features are combined to form a new cement compressive strength identification feature, and are connected to the model output layer S.

[0088] Since the information of the model output layer S is known, that is, corresponding to the cement compressive strength, only the connection weight W of the feature layer and the output layer needs to be calculated, and the calculation formula is as follows:

[0089] W=A -1 S

[0090] In the calculation, the ridge regression is used to solve the weight matrix, that is, W is solved through the optimization problem of formula (3):

[0091]

[0092] Wherein, σ1=σ2=v=u=2, and the solution is:

[0093] W=(λI+AA T ) -1 A T S.

[0094] It should be noted that, for the foregoing method embodiments, in order to facilitate the description, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other order or simultaneously.

[0095] Based on the same idea as the cement compressive strength prediction method based on the integrated semi-supervised width regression system in the above embodiment, the present application also provides a cement compressive strength prediction system based on an integrated semi-supervised width regression system, which can be used to execute the cement compressive strength prediction method based on the integrated semi-supervised width regression system. For the convenience of description, in the structural schematic diagram of the cement compressive strength prediction system based on the integrated semi-supervised width regression system embodiment, only the part related to the embodiment of the present application is shown, and those skilled in the art can understand that the illustrated structure does not constitute a limitation on the device, and can include more or fewer components than the illustrated, or combine certain components, or different component arrangements.

[0096] Please refer to Figure 3 In another embodiment of the present application, a cement compressive strength prediction system based on an integrated semi-supervised width regression system 10 is provided, which comprises a preprocessing module 11, a pre-labeling module 12, an auxiliary training data selection module 13, an iterative updating module 14, a training decision model module 15 and a result evaluation module 16.

[0097] The preprocessing module 11 is used to determine the cement compressive strength influencing factors and pre-process the cement samples to obtain a sample data set; the sample data set includes labeled data and unlabeled data;

[0098] a pre-labeling module 12 configured to initialize a plurality of BLS models, train the BLS models using existing small amounts of labeled data, and obtain a plurality of labeling regression models;

[0099] an auxiliary training data selection module 13 configured to predict cement compressive strength values using the labeling regression models, assign the cement compressive strength values to unlabeled data to obtain pre-labeled data, statistically analyze the pre-labeled data of the plurality of labeling regression models, calculate an average value of the predicted values, obtain an average predicted value, filter the average predicted value using a set threshold, use the filtered average predicted value as a pseudo label of a corresponding sample, obtain pseudo-labeled data, and leave the rest of the samples as unlabeled data;

[0100] an iterative updating module 14 configured to jump back to S2, iteratively train and update the labeling regression models using the pseudo-labeled data and the labeled data, repeat the process in a loop until a set condition is met, terminate the training, and obtain a final labeled data set;

[0101] a training decision model module 15 configured to train a final decision BLS regression model using the final labeled data set;

[0102] a result evaluation module 16 configured to use the trained final decision BLS model to predict cement compressive strength of new cement samples.

[0103] It should be noted that the cement compressive strength prediction system based on the integrated semi-supervised width regression system of the present application corresponds to the cement compressive strength prediction method based on the integrated semi-supervised width regression system of the present application. The technical features and advantages described in the above embodiment of the cement compressive strength prediction method based on the integrated semi-supervised width regression system are applicable to the embodiment of the cement compressive strength prediction method based on the integrated semi-supervised width regression system. For specific content, please refer to the description in the method embodiment of the present application. Here, no further description is given, and this is hereby declared.

[0104] In addition, in the embodiment of the cement compressive strength prediction system based on the integrated semi-supervised width regression system of the above embodiment, the logical division of each program module is only an example. In actual application, the above functions can be completed by different program modules according to needs, for example, due to the configuration requirements of the corresponding hardware or the convenience of software implementation. That is, the internal structure of the cement compressive strength prediction system based on the integrated semi-supervised width regression system is divided into different program modules to complete all or part of the functions described above.

[0105] Please refer to Figure 4In one embodiment, an electronic device implementing a cement compressive strength prediction method based on an integrated semi-supervised width regression system is provided. The electronic device 20 can include a first processor 21, a first memory 22, and a bus. The electronic device 20 can also include a computer program stored in the first memory 22 and executable on the first processor 21, such as a cement compressive strength prediction program 23 based on an integrated semi-supervised width regression system.

[0106] The first memory 22 includes at least one type of readable storage medium, including a flash memory, a mobile hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. In some embodiments, the first memory 22 can be an internal storage unit of the electronic device 20, such as a mobile hard disk of the electronic device 20. In other embodiments, the first memory 22 can also be an external storage device of the electronic device 20, such as a plug-in mobile hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the first memory 22 can include both an internal storage unit and an external storage device of the electronic device 20. The first memory 22 can be used to store application software installed in the electronic device 20 and various data, such as the code of the cement compressive strength prediction program 23 based on an integrated semi-supervised width regression system, and can also be used to temporarily store data that has been output or will be output.

[0107] The first processor 21 can be composed of an integrated circuit in some embodiments, such as a single packaged integrated circuit, or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, combinations of various control chips, etc. The first processor 21 is the control core of the electronic device, which connects various components of the electronic device through various interfaces and lines, and executes various functions and processes data of the electronic device 20 by running or executing programs or modules stored in the first memory 22 and calling data stored in the first memory 22.

[0108] Figure 4 Only an electronic device with components is shown, and those skilled in the art can understand that, Figure 4 The structure shown does not constitute a limitation on the electronic device 20, and can include fewer or more components than shown, or combine certain components, or different component arrangements.

[0109] The cement compressive strength prediction program 23 stored in the first memory 22 in the electronic device 20 based on the integrated semi-supervised width regression system is a combination of a plurality of instructions, which, when running in the first processor 21, can realize:

[0110] S1, determining the cement compressive strength influencing factors and preparing the corresponding cement samples, and obtaining the sample data set; the sample data set includes labeled data and unlabeled data;

[0111] S2, initializing a plurality of BLS models: training BLS models using existing small amount of labeled data, and obtaining a plurality of labeled regression models;

[0112] S3, predicting the cement compressive strength value using the labeled regression model, assigning the cement compressive strength value to the unlabeled data, and obtaining the pre-labeled data;

[0113] S4, statistical decision is made according to the pre-labeled data of the plurality of labeled regression models, the average value of the predicted value is calculated, the average predicted value is obtained, the average predicted value is screened by using the set threshold, and the screened average predicted value is used as the pseudo label of the corresponding sample. The pseudo label data is obtained, and the remaining samples are still unlabeled data;

[0114] S5, jump back to S2, update the labeled regression model by iterative training using the selected pseudo label data and labeled data, repeat the process until the set condition is met, terminate the training, and obtain the final labeled data set;

[0115] S6, training the final decision BLS regression model using the final labeled data set;

[0116] S7, using the trained final decision BLS model to predict the cement compressive strength of new cement samples.

[0117] Further, the modules / units integrated in the electronic device 20, if realized in the form of software function units and sold or used as independent products, can be stored in a non-volatile computer readable storage medium. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory).

[0118] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing relevant hardware. The program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0119] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist, it should be considered as the scope of the present application.

[0120] The above embodiments are the preferred embodiments of the present application, but the embodiments of the present application are not limited to the above embodiments, and any changes, modifications, substitutions, combinations and simplifications made without departing from the spirit and principles of the present application should be equivalent replacement methods and should be included in the protection scope of the present application.

Claims

1. A method for estimating cement compressive strength based on an integrated semi-supervised width regression system, characterized in that: The steps include: S1. Determine the factors affecting cement compressive strength and prepare corresponding cement samples to obtain a sample data set; the sample data set includes labeled data and unlabeled data; S2. Initialize multiple BLS models: Use the existing small amount of labeled data to train the BLS model and obtain multiple labeling regression models; S3. Use the labeling regression model to predict the cement compressive strength value, assign the cement compressive strength value to the unlabeled data, and obtain pre-labeled data; the compressive strength value is calculated as follows: in, represents the linear mapping function, ζ(·) represents the nonlinear mapping function, X u represents unlabeled data, W e and β e represents the mapping feature connection weight, β h represents the enhanced feature node connection weight, and W represents the connection weight of the output layer; S4. Calculate the average of the predicted values ​​based on the pre-labeled data of multiple labeled regression models, obtain the average predicted value, filter the average predicted value using the set threshold, and use the filtered average predicted value as the pseudo label of the corresponding sample to obtain pseudo-label data. The remaining samples remain unlabeled data. S5, jump back to S2, use the selected pseudo-label data and labeled data to iteratively train and update the labeling regression model, and repeat the process until the set conditions are met, the training is terminated, and the final labeled data set is obtained; S6. Use the final labeled dataset to train the final decision BLS regression model; S7. Use the trained final decision BLS model to estimate the compressive strength of new cement samples.

2. The method for estimating cement compressive strength based on an integrated semi-supervised width regression system according to claim 1 is characterized in that: The preparation of the corresponding cement sample and the acquisition of the sample data set are specifically as follows: Cement samples were prepared according to the corresponding property values ​​of the factors affecting cement compressive strength, and hydration reaction was carried out under standard curing conditions; The cement samples are tested and recorded according to the cement compressive strength determination method. If the test results meet the set standards, the compressive strength is used as the label of the cement sample to obtain the labeled data. Otherwise, the cement sample has no corresponding cement compressive strength and the unlabeled data is obtained.

3. The method for estimating cement compressive strength based on an integrated semi-supervised width regression system according to claim 1 is characterized in that: Calculate the deviation between the average predicted value and the predicted value, compare the deviation value with the preset deviation threshold, and record the average predicted value below the deviation threshold as a valid prediction; For a certain sample, multiple labeling regression models predict the sample, obtain the corresponding average prediction value, and use the average prediction value that meets the statistical probability as the pseudo label of the sample.

4. The method for estimating cement compressive strength based on an integrated semi-supervised width regression system according to claim 3 is characterized in that: The calculated average predicted value The deviation value of the predicted value s is compared with the preset deviation threshold as follows: Among them, the error threshold ∈∈[0,1] and the probability threshold δ∈[0.5,1].

5. The method for estimating cement compressive strength based on an integrated semi-supervised width regression system according to claim 1, characterized in that: The pseudo label is determined by the following formula: in, represents the average predicted value, Pr() represents the screening function, and δ represents the probability threshold.

6. The method for estimating cement compressive strength based on an integrated semi-supervised width regression system according to claim 1, characterized in that: The BLS model also includes the number of feature mapping nodes and the number of enhancement nodes; The trained BLS model is used to estimate the compressive strength of new cement samples, specifically: Determine the factors affecting the compressive strength of cement; After random mapping, the mapping feature node Z is formed n ≡[Z1,Z2,…,Z n ], where the i-th group mapping feature Z i Calculation, each set of features contains k mapping features, as follows: Among them, X u represents unlabeled data, W ei and β ei Both represent the i-th group mapping feature Z i The connection weight of n represents the total number of mapping nodes; Map feature node Z n Perform nonlinear mapping to obtain enhanced feature node H m ≡[H1,H2,…,H m ], where the jth group of enhanced features H j , as follows: H j ≡ζ(Z n W hj +β hj ),j=1,2,…,m Among them, the connection weight W hj and β hj is randomly generated, ζ(·) is a nonlinear activation function, and m represents the total number of enhanced nodes; Map feature Z i and enhanced feature H j The new cement compressive strength identification features are merged and connected to the model output layer S.

7. A cement compressive strength prediction system based on an integrated semi-supervised width regression system, characterized in that: The method for estimating cement compressive strength based on an integrated semi-supervised width regression system applied to any one of claims 1 to 6 comprises a preprocessing module, a pre-marking module, an auxiliary training data selection module, an iterative update module, a training decision model module, and a result evaluation module; A preprocessing module is used to determine the factors affecting the compressive strength of cement and preprocess the cement samples to obtain a sample data set; the sample data set includes labeled data and unlabeled data; The pre-labeling module is used to initialize multiple BLS models: the BLS model is trained using a small amount of existing labeled data to obtain multiple labeled regression models; The auxiliary training data selection module is used to predict the cement compressive strength value using the labeling regression model, assign the cement compressive strength value to the unlabeled data, and obtain pre-labeled data; perform statistical decision-making based on the pre-labeled data of multiple labeling regression models, calculate the average of the predicted values, obtain the average predicted value, filter the average predicted value using the set threshold, and use the filtered average predicted value as the pseudo-label of the corresponding sample to obtain pseudo-labeled data; the remaining samples remain unlabeled data; The iterative update module is used to jump back to S2 and iteratively train and update the labeling regression model using pseudo-label data and labeled data. The process is repeated until the set conditions are met, the training is terminated, and the final labeled data set is obtained; The training decision model module uses the final labeled dataset to train the final decision BLS regression model; The result evaluation module uses the trained final decision BLS model to estimate the compressive strength of new cement samples.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores computer program instructions that can be executed by the at least one processor, and the computer program instructions are executed by the at least one processor so that the at least one processor can execute the cement compressive strength prediction method based on the integrated semi-supervised width regression system as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a program, characterized in that: When the program is executed by a processor, the method for estimating cement compressive strength based on an integrated semi-supervised width regression system according to any one of claims 1 to 6 is implemented.

Images