A digital management and control system for concrete

Abstract

The present invention relates to the technical field of concrete, and discloses a digital management and control system for concrete. By analyzing the historical inventory data of concrete raw materials, the inventory trend of concrete raw materials is predicted to manage the inventory of concrete raw materials; the PSO algorithm is used to optimize the SVM model to classify the quality of concrete raw materials to monitor the quality of concrete raw materials; when a concrete production instruction is received, a production parameter scheme is inferred by using a fuzzy control algorithm according to the production plan information; key feature variables are extracted to evaluate the performance indexes of concrete, and the initial production scheme is adjusted according to the concrete test results to obtain a target production scheme. It can quickly and accurately identify the quality status of raw materials. It helps to timely discover raw materials with unqualified quality and prevent them from entering the production link, thus ensuring the quality of concrete.

Description

Technical Field

[0001] The present invention relates to the technical field of concrete, and particularly to a digital management and control system for concrete. Background Art

[0002] There are many problems in the traditional concrete production process. In production management, due to the lack of rigorous processes and scientific overall planning, it is often in a mess. The connection between each process is loose, the raw material procurement is highly arbitrary, without accurate planning, often resulting in supply imbalance. In terms of quality control, the raw material ratio depends on manual work, with frequent errors, inaccurate usage of cement, aggregates, etc., resulting in drastic fluctuations in the performance of concrete. The mixing time is only controlled by experience and is difficult to be accurate. Insufficient time makes the material mixing uneven and the workability deteriorates; too long will damage the aggregates and weaken the strength and durability. Secondly, the phenomenon of resource waste is extremely prominent. Due to the difficulty in real-time monitoring of production data and the lack of key information, enterprises cannot adjust and control in a timely manner. Once there is a deviation in the raw material ratio or abnormal mixing, it cannot be quickly detected and corrected, and the whole batch of concrete may be scrapped, wasting a large amount of raw materials. At the same time, due to unstable quality, the number of defective products increases, and the rework or waste treatment consumes more manpower, material resources and financial resources, seriously hindering the concrete production industry from moving towards efficient and high-quality development. Therefore, how to conduct precise intelligent control for the production of concrete and improve the quality of concrete production finished products is a technical problem that needs to be solved urgently at present. Summary of the Invention

[0003] The purpose of the present invention is to solve the above problems and design a digital management and control system for concrete.

[0004] The present invention provides a digital management and control system for concrete. The digital management and control system for concrete includes a raw material inventory management module, a raw material quality monitoring module, a production plan acquisition module, a ratio optimization module, a production parameter inference module, a production plan determination module and a data management module. Among them,

[0005] The raw material inventory management module predicts the inventory trend of concrete raw materials by analyzing the historical inventory data of concrete raw materials, obtains the inventory prediction result, and manages the inventory of concrete raw materials;

[0006] The raw material quality monitoring module is used to classify the quality of concrete raw materials by optimizing the SVM model with the PSO algorithm, obtain the quality result classification, and monitor the quality of concrete raw materials;

[0007] The production plan acquisition module is used to obtain production plan information according to the production instruction when receiving the concrete production instruction;

[0008] The ratio optimization module is used to determine the optimization objective according to the production plan information, optimize the ratio of raw materials by using the genetic algorithm, and obtain the raw material ratio plan through continuous iteration and optimization;

[0009] The production parameter inference module is used to take the factors affecting the concrete production parameters as input variables, take the production parameters as output variables, and infer the production parameter plan by using the fuzzy control algorithm according to the production plan information;

[0010] The production plan determination module is used to determine the initial production plan by combining the raw material ratio plan and the production parameter plan, and execute the initial production plan;

[0011] The data management module is used to receive the real-time data in the concrete production process, analyze and process it by using the DVAE-WAFFN-WGAN-GP-LSTM model according to the real-time data, extract the key feature variables, evaluate the performance indicators of the concrete, obtain the concrete detection results, and adjust the initial production plan according to the concrete detection results to obtain the target production plan.

[0012] Optionally, in the first implementation manner of the present invention, the raw material management module includes a data concatenation sub-module, a component reconstruction sub-module, a grabbing sub-module, a feature vector extraction sub-module, and a weight assignment sub-module, where,

[0013] The data concatenation sub-module is used to obtain the historical inventory data of the concrete raw materials, concatenate the inventory feature data in the historical inventory data at each moment by using word vectors to form a new time series, where the inventory feature data at least includes the purchase quantity, the consumption quantity, the available days, and the inventory quantity;

[0014] The component reconstruction sub-module is used to decompose the historical inventory data to obtain 1 residual component and 8 IMF components, obtain 9 components, and then reconstruct the components into 3 components with different frequency characteristics by using the ZCR coefficient;

[0015] The grabbing sub-module is used to grab the long-term dependence information of the time series x of the components with different frequency characteristics through the TCN module:

[0016]

[0017] Where, F(t) is the output after one dilated convolution operation, d is the sampling rate, that is, the window coefficient, k is the convolution kernel size, f(i) is the i-th element in the convolution kernel, and x t-d·i is only used for convolution operation on past data;

[0018] The feature vector extraction sub-module is used to input the extracted dependency information into the BiGRU prediction module to extract feature vectors in both forward and reverse directions and analyze time series data;

[0019] The weight assignment sub-module is used to assign weights to the input information of the BiGRU prediction module by using the Attention mechanism to highlight the contribution degree of important information, and then make a prediction. The prediction results of each frequency component are superimposed to obtain the inventory prediction result. According to the inventory prediction result, a purchase warning is automatically issued to remind the management personnel to make a purchase.

[0020] Optionally, in the second implementation manner of the present invention, the raw material quality monitoring module includes a first initialization sub-module, a segmentation sub-module, an update sub-module, a comparison sub-module, a first judgment sub-module, and a classification sub-module, where,

[0021] The first initialization sub-module is used to initialize the particle swarm capacity, the maximum number of iterations, the acceleration coefficient, and the inertia parameter of the PSO algorithm, initialize the initial values and search ranges of the penalty factor parameter and the kernel function parameter of the SVM model, and collect the physical and chemical property data of the concrete raw materials to obtain quality feature samples;

[0022] The segmentation sub-module is used to segment the quality feature samples into multiple non-overlapping subsets to obtain quality feature subsets, initialize the iteration counter to 1, select the first n pairs of quality feature subsets as the training set, and other subsets as the test set;

[0023] The update sub-module is used to perform a search within the search range by using the PSO algorithm, assign the initial value to the SVM model, calculate the initial fitness value, perform the iterative process of the PSO algorithm, and update the velocity and position of the particle individual:

[0024]

[0025] In the formula, k is the current iteration number of the algorithm; ω is the inertia parameter; and are the velocity and position of the i-th particle at the iteration number k, d is the space dimension; is the value of the personal best position of the i-th particle in the space dimension d at the k-th iteration, is the value of the global best position in the space dimension d at the k-th iteration, and are the velocity and position of the i-th particle at the iteration number k + 1, c1 and c2 are both acceleration coefficients, r1 and r2 are random functions, r1, r2 ∈ [0, 1], η is the contraction factor, is the total acceleration coefficient, K is the contraction coefficient, 0 < K < 1;

[0026] A comparison sub-module, which is used to calculate the new fitness values of each particle every iteration, compare the current optimal fitness value with the new fitness values of each particle, and if the new fitness value of a particle is better than the optimal fitness value, replace the current optimal fitness value with the new fitness value of this particle;

[0027] A first judgment sub-module, which is used to judge whether the current optimal fitness value reaches the maximum number of iterations. If not, the number of iterations is incremented by 1 for the next round of iteration. If so, the penalty factor parameter and the kernel function parameter are updated to obtain the optimal SVM model;

[0028] A classification sub-module, which is used to classify the quality of concrete raw materials by using the optimal SVM model, classify the raw materials into three categories: qualified, unqualified and to be tested, and obtain the quality result classification to process the unqualified raw materials in time.

[0029] Optionally, in the third implementation manner of the present invention, the ratio optimization module includes a second initialization sub-module, a construction sub-module, a selection operation sub-module, a crossover operation sub-module, a mutation operation sub-module and a second judgment sub-module, wherein,

[0030] The second initialization sub-module is used to initialize the population size, the maximum crossover probability and mutation probability, the minimum crossover probability and mutation probability, and the maximum number of iterations G;

[0031] The construction sub-module is used to determine the optimization goal according to the production plan information and construct the optimal ratio objective function of concrete raw materials:

[0032]

[0033] In the formula, f is the value of the optimal ratio objective function, n is the upper limit of summation, indicating the summation of n items of requirements, i is the type of concrete mix raw materials, x i are various raw materials used in concrete, y i are the unit prices corresponding to various raw materials;

[0034] The selection operation sub-module is used to encode the ratio of raw materials. After encoding, it is represented as an individual in the genetic algorithm. Set the iteration variable g = 0, calculate the fitness value of each individual in the initial population, sort the calculation results from large to small, and divide all individuals in the population into 6 equal parts according to the sorting results. Double the number of the 1 / 6 individuals ranked at the front, delete the 1 / 6 individuals ranked at the end, and do not enter the next generation. Copy the 2 / 3 individuals ranked in the middle and directly inherit them to the next generation;

[0035] The crossover operation sub-module is used to perform a crossover operation on each individual in the population after the selection operation. The crossover probability Pcn is:

[0036]

[0037] In the formula, f max is the maximum fitness value of all individuals in the current population, P cn is the crossover probability of individual n in the population, P cnmax is the maximum crossover probability, which is the upper limit value of the crossover probability, P cnmin is the minimum crossover probability, which is the lower limit value of the crossover probability, G is the total number of iterations, g is the current number of iterations, f avg is the average fitness value of all individuals in the population, f n is the fitness value of individual n in the population, P cnmin = 0.6;

[0038] The mutation operation sub-module is used to perform a mutation operation on each individual in the population after the crossover operation. The mutation probability P mn is:

[0039]

[0040] In the formula, P mn is the mutation probability of individual n, P mnmax is the maximum mutation probability, which is the upper limit value of the mutation probability, P mnmin is the minimum mutation probability, which is the lower limit value of the mutation probability, P mnmin = 0.5;

[0041] The second judgment sub-module is used to judge whether g is less than G. If not, let g = g + 1 and repeat the genetic operation. If so, judge whether the output result meets the termination condition. If it meets, the algorithm terminates. In the last generation population of the genetic algorithm, select the individual with the highest fitness value as the optimal raw material mixing ratio plan, where the termination condition is to reach the maximum number of iterations.

[0042] Optionally, in the fourth implementation manner of the present invention, the production parameter inference module includes an as sub-module, a fuzzification processing sub-module, and a fuzzy inference sub-module, where,

[0043] The as sub-module is used to use the factors affecting the concrete production parameters as input variables and the production parameters as output variables, where the input variables at least include the performance of raw materials and the production environment, and the output variables at least include the mixing time, mixing speed, and water-cement ratio;

[0044] The fuzzification processing sub-module is used to perform fuzzification processing on the input variable X and the output variable Y to convert them into the membership degrees μ(X i ) and μ(Y i) Suppose there are 3 fuzzy sets for both the input variables and the output variables:

[0045] μ(X i ) = {PS, PM, PB}

[0046] μ(Y i ) = {PS, PM, PB}

[0047] where PS, PM, and PB are the three fuzzy sets of small, medium, and large respectively;

[0048] The fuzzy inference sub-module is used to perform fuzzy inference based on the fuzzy sets and a preset fuzzy rule base, calculate the fuzzy sets corresponding to each output variable, defuzzify the fuzzy sets of the output variables using the maximum membership degree method to obtain the production parameter values, and generate a production parameter plan, where the preset fuzzy rule base includes the configuration relationship between the input variables and the output variables.

[0049] Optionally, in the fifth implementation manner of the present invention, the data management module includes a collection sub-module, an encoding and decoding sub-module, an evaluation sub-module, an analysis sub-module, and an adjustment sub-module, where,

[0050] The collection sub-module is used to collect real-time data during the concrete production process in real time through sensors installed on the production equipment, and input the collected real-time data into the DVAE-WAFFN-WGAN-GP-LSTM model;

[0051] The encoding and decoding sub-module is used to extract production feature variables through the encoding and decoding processes of the DVAE-WAFFN-WGAN-GP-LSTM model, and perform dimensionality reduction processing on the extracted production feature variables using the principal component analysis algorithm to obtain key feature variables, where the key feature variables at least include mixing time, mixing speed, water-cement ratio, ambient temperature, temperature of concrete, and ambient humidity;

[0052] The evaluation sub-module is used to evaluate the performance indicators of the concrete during the production process according to the extracted key feature variables, and output the concrete detection results, where the performance indicators at least include strength, fluidity, and durability;

[0053] The analysis sub-module is used to analyze the concrete detection results, determine the performance indicators that deviate from the concrete production target, and analyze the real-time data generated during the production process to determine the material ratio and production parameters to be adjusted;

[0054] The adjustment sub-module is used to adjust the initial production plan according to the material ratio and production parameters to be adjusted to obtain the target production plan.

[0055] Optionally, in the sixth implementation manner of the present invention, the encoding and decoding sub-module specifically includes the following processes:

[0056] Input the collected real-time data into the variational autoencoder DVAE. Through the DVAE, compress the high-dimensional input data into a low-dimensional latent representation, and then input it into the WAFFN. Concatenate the output layer features F1 and F2 of the variational autoencoder DVAE in terms of dimensions to generate spatio-temporal feature information f 12 , and input the spatio-temporal feature information into an activation function for non-linear transformation to obtain feature f′ 12 :

[0057] f 12 =[F1,F2]

[0058] f′ 12 =ReLU(W·f 12 +b)

[0059] In the formula, W and b represent the weight matrix and the bias vector, and ReLU is the activation function;

[0060] Perform a softmax operation on the feature f′ 12 to generate the weight vectors w1, w2 of the output layer features of each variational autoencoder:

[0061]

[0062] Perform weighted fusion on the calculated weight vectors and the variational encoder output features:

[0063]

[0064] In the formula, F U is the fusion feature, is the tensor multiplication operation, and w j represents the jth parameter vector for calculating the weight;

[0065] Decode the latent representation encoded by the WAFFN through the WGAN-GP and LSTM to reconstruct an approximate representation of the input real-time data. At the beginning of the encoding, the LSTM layer processes the time series data. At the last stage of the decoding, the LSTM layer will adjust the reconstructed data in the time series dimension according to the time dependence relationship. Through the encoding and decoding processes, extract the production feature variables that can reflect the concrete production process.

[0066] In the technical solution provided by the present invention, the method for implementing the concrete digital management control system includes the following steps:

[0067] Predict the inventory trend of concrete raw materials by analyzing the historical inventory data of concrete raw materials, obtain the inventory prediction results, and conduct inventory management on concrete raw materials;

[0068] Use the PSO algorithm to optimize the SVM model to classify the quality of concrete raw materials, obtain the quality result classification, and conduct quality monitoring on concrete raw materials;

[0069] When receiving a concrete production instruction, obtain production plan information according to the production instruction;

[0070] Determine the optimization objective according to the production plan information, use the genetic algorithm to optimize the raw material ratio, and through continuous iteration and optimization, obtain the raw material ratio plan;

[0071] Take the factors affecting concrete production parameters as input variables, take production parameters as output variables, and use the fuzzy control algorithm to infer the production parameter plan according to the production plan information;

[0072] Combine the raw material ratio plan and the production parameter plan to determine the initial production plan, and execute the initial production plan;

[0073] Receive real-time data during the concrete production process, analyze and process it using the DVAE-WAFFN-WGAN-GP-LSTM model according to the real-time data, extract key feature variables, evaluate the performance indicators of concrete, obtain the concrete test results, and adjust the initial production plan according to the concrete test results to obtain the target production plan.

[0074] Optionally, in the implementation manner of this method, the predicting the inventory trend of concrete raw materials by analyzing the historical inventory data of concrete raw materials, obtaining the inventory prediction results, and conducting inventory management on concrete raw materials includes:

[0075] Decompose the historical inventory data to obtain 1 residual component and 8 IMF components, getting 9 components, and then reconstruct the components into 3 components with different frequency characteristics through the ZCR coefficient;

[0076] Use the TCN module to capture long-term dependency information for the time series x of components with different frequency characteristics:

[0077]

[0078] Among them, F(t) is the output after one dilated convolution operation, d is the sampling rate, i.e., the window coefficient, k is the convolution kernel size, f(i) is the i-th element in the convolution kernel, and x t-d·i Only performs convolution operations on past data;

[0079] The extracted dependency information is input into the BiGRU prediction module to extract feature vectors in both forward and reverse directions and analyze time series data;

[0080] The Attention mechanism is used to allocate weights to the input information of the BiGRU prediction module to highlight the contribution degree of important information, and then prediction is carried out. The prediction results of each frequency component are superimposed to obtain the inventory prediction result, and a purchase warning is automatically issued according to the inventory prediction result to remind the management staff to make purchases.

[0081] Optionally, in the implementation manner of this method, the SVM model is optimized by the PSO algorithm to classify the quality of concrete raw materials to obtain a quality result classification for quality monitoring of concrete raw materials, including:

[0082] Initialize the particle swarm capacity, maximum number of iterations, acceleration coefficient, and inertia parameter of the PSO algorithm, initialize the initial values and search ranges of the penalty factor parameter and kernel function parameter of the SVM model, and collect the physical and chemical property data of concrete raw materials to obtain quality feature samples;

[0083] The quality feature samples are divided into multiple non-overlapping subsets to obtain quality feature subsets. Initialize the iteration counter to 1, select the first n pairs of quality feature subsets as the training set, and other subsets as the test set;

[0084] The PSO algorithm is used to perform a search within the search range, assign the initial value to the SVM model, calculate the initial fitness value, and perform the iterative process of the PSO algorithm to update the velocity and position of the particle individual:

[0085]

[0086] In the formula, k is the current iteration number of the algorithm; ω is the inertia parameter; and are the velocity and position of the i-th particle at the k-th iteration, d is the spatial dimension; is the value of the personal best position of the i-th particle in the spatial dimension d at the k-th iteration, is the value of the global best position in the spatial dimension d at the k-th iteration, and are the velocity and position of the i-th particle at the (k + 1)-th iteration, is the current position of the i-th particle in the spatial dimension d at the k-th iteration; c2 are both acceleration coefficients, r1, r2 are random functions, r1, r2 ∈ [0, 1], η is the contraction factor, is the total acceleration coefficient, K is the contraction coefficient, 0 < K < 1;

[0087] Calculate the new fitness value of each particle in each iteration, and compare the current optimal fitness value with the new fitness values of each particle. If the new fitness value of a particle is better than the optimal fitness value, replace the current optimal fitness value with the new fitness value of this particle;

[0088] Judge whether the current optimal fitness value reaches the maximum number of iterations. If not, increase the number of iterations by 1 and perform the next round of iteration. If so, update the penalty factor parameter and the kernel function parameter to obtain the optimal SVM model;

[0089] Use the optimal SVM model to classify the quality of concrete raw materials, divide the raw materials into three categories: qualified, unqualified and to be detected, and obtain the quality result classification to promptly process the unqualified raw materials.

[0090] Its beneficial effects are as follows: By analyzing the historical inventory data of concrete raw materials, predicting the inventory trend of concrete raw materials, obtaining the inventory prediction result, and conducting inventory management on concrete raw materials; using the PSO algorithm to optimize the SVM model to classify the quality of concrete raw materials, obtaining the quality result classification, and conducting quality monitoring on concrete raw materials; when receiving a concrete production instruction, obtaining production plan information according to the production instruction; determining an optimization goal according to the production plan information, using the genetic algorithm to optimize the raw material ratio, and obtaining a raw material ratio plan through continuous iteration and optimization; taking the factors affecting concrete production parameters as input variables, taking production parameters as output variables, and using the fuzzy control algorithm to infer a production parameter plan according to the production plan information; combining the raw material ratio plan and the production parameter plan to determine an initial production plan, and executing the initial production plan; receiving real-time data during the concrete production process, analyzing and processing it using the DVAE-WAFFN-WGAN-GP-LSTM model, extracting key feature variables, evaluating the performance indicators of concrete, obtaining the concrete detection result, and adjusting the initial production plan according to the concrete detection result to obtain the target production plan. 1. By analyzing the historical inventory data of concrete raw materials to predict the inventory trend, it is possible to understand the inventory situation of raw materials in advance. This helps to avoid production stagnation caused by raw material shortages, ensure the continuity of concrete production, and reasonable inventory prediction can reduce unnecessary inventory backlogs and lower inventory costs, including warehousing costs and capital occupation costs. 2. It can quickly and accurately identify the quality status of raw materials. This helps to promptly discover raw materials with unqualified quality and prevent them from entering the production process, thus ensuring the quality of concrete. 3. Using the genetic algorithm to optimize the raw material ratio can comprehensively consider various factors and find the optimal raw material ratio plan through continuous iteration. It can ensure that while meeting the quality requirements of concrete, the cost is minimized to the greatest extent, making the raw material ratio more in line with the actual production needs and improving production efficiency. 4. By inferring a production parameter plan using the fuzzy control algorithm, it is possible to fully consider the influence of various complex factors on production. This helps to maintain stable production conditions during the production process and improve the consistency and reliability of concrete production. 5. Timely extracting key feature variables and evaluating the performance indicators of concrete can ensure that the quality of the produced concrete always meets the requirements and avoid quality fluctuations. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] By reading the detailed description of the preferred embodiments below, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention.

[0092] Figure 1Schematic diagram of the first embodiment of a concrete digital management control system provided by an embodiment of the present invention;

[0093] Figure 2 Schematic diagram of the second embodiment of a concrete digital management control system provided by an embodiment of the present invention;

[0094] Figure 3 Schematic diagram of the third embodiment of a concrete digital management control system provided by an embodiment of the present invention. Detailed implementation manners

[0095] The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and above-mentioned drawings of the present invention are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data used can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order different from that illustrated or described herein. In addition, the terms "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0096] For ease of understanding, the specific processes of the embodiments of the present invention are described below. Please refer to Figure 1 Schematic diagram of the first embodiment of a concrete digital management control system provided by an embodiment of the present invention. The method specifically includes a raw material inventory management module, a raw material quality monitoring module, a production plan acquisition module, a ratio optimization module, a production parameter inference module, a production plan determination module and a data management module. Among them,

[0097] The raw material inventory management module predicts the inventory trend of concrete raw materials by analyzing the historical inventory data of concrete raw materials, obtains the inventory prediction result, and manages the inventory of concrete raw materials;

[0098] Specifically, this embodiment also includes: a data concatenation sub-module, which is used to obtain the historical inventory data of concrete raw materials, and concatenate the inventory feature data at each moment with the historical inventory data by using word vectors to form a new time series, where the inventory feature data at least includes the purchase quantity, consumption quantity, available days and inventory quantity;

[0099] The component reconstruction sub-module is used to decompose the historical inventory data to obtain 1 residual component and 8 IMF components, obtaining 9 components, and then reconstructing the components into 3 components with different frequency characteristics through the ZCR coefficient;

[0100] The grabbing sub-module is used to grab long-term dependency information of the time series x of components with different frequency characteristics through the TCN module:

[0101]

[0102] where F(t) is the output after one dilated convolution operation, d is the sampling rate i.e., the window coefficient, k is the convolution kernel size, f(i) is the i-th element in the convolution kernel, and x t-d·i performs convolution operations only on past data;

[0103] The feature vector extraction sub-module is used to input the extracted dependency information into the BiGRU prediction module to extract feature vectors in both forward and backward directions and analyze time series data;

[0104] The weight assignment sub-module is used to assign weights to the input information of the BiGRU prediction module using the Attention mechanism to highlight the contribution degree of important information and then make predictions, and superimpose the prediction results of each frequency component to obtain the inventory prediction result, and automatically issue a purchase warning according to the inventory prediction result to remind the management to make purchases.

[0105] The raw material quality monitoring module is used to optimize the SVM model using the PSO algorithm to classify the quality of concrete raw materials and obtain the quality result classification to monitor the quality of concrete raw materials;

[0106] Specifically, this embodiment further includes: the first initialization sub-module, which is used to initialize the particle swarm capacity, maximum iteration number, acceleration coefficient, and inertia parameter of the PSO algorithm, initialize the initial values and search ranges of the penalty factor parameter and kernel function parameter of the SVM model, and collect the physical and chemical property data of concrete raw materials to obtain quality feature samples;

[0107] The segmentation sub-module is used to segment the quality feature samples into multiple non-overlapping subsets to obtain quality feature subsets, initialize the iteration counter to 1, select the first n pairs of quality feature subsets as the training set, and other subsets as the test set;

[0108] The update sub-module is used to perform a search within the search range using the PSO algorithm, assign the initial value to the SVM model, calculate the initial fitness value, perform the iterative process of the PSO algorithm, and update the velocity and position of the particle individuals:

[0109]

[0110] In the formula, k is the current iteration number of the algorithm; ω is the inertia parameter; and are the velocity and position of the i-th particle when the iteration number is k, and d is the spatial dimension; is the value of the personal best position of i particles in the spatial dimension d at the k-th iteration, is the value of the global best position in the spatial dimension d at the k-th iteration, and are the velocity and position of the i-th particle at the (k + 1)-th iteration, c1 and c2 are both acceleration coefficients, r1 and r2 are random functions, r1, r2 ∈ [0, 1], and η is a contraction factor, is the total acceleration coefficient, K is the contraction coefficient, 0 < K < 1;

[0111] A comparison sub-module, which is used to calculate the new fitness value of each particle every iteration, compare the current optimal fitness value with the new fitness values of each particle, and if the new fitness value of a particle is better than the optimal fitness value, then replace the new fitness value of this particle with the current optimal fitness value;

[0112] A first judgment sub-module, which is used to judge whether the current optimal fitness value reaches the maximum number of iterations. If not, the number of iterations is incremented by 1 for the next round of iteration. If so, the penalty factor parameter and the kernel function parameter are updated to obtain the optimal SVM model;

[0113] A classification sub-module, which is used to classify the quality of concrete raw materials using the optimal SVM model, classify the raw materials into three categories: qualified, unqualified, and to be tested, and obtain the quality result classification to process the unqualified raw materials in a timely manner.

[0114] A production plan acquisition module, which is used to obtain production plan information according to the production instruction when receiving a concrete production instruction;

[0115] Specifically, this embodiment further includes: when receiving a concrete production instruction, this instruction often contains key information such as the type of concrete to be produced (for example, high-strength concrete for infrastructure construction or ordinary concrete for indoor floors), the estimated output, and the delivery time.

[0116] First, after extracting the information about the type of concrete from the production instruction, the production plan information can be further determined. For different types of concrete, their production plans are significantly different. For example, high-strength concrete may require specific raw materials, such as high-quality cement and specifically graded aggregates, and the production process may require more strict quality control and a longer mixing time. According to these requirements, the production plan information will include the procurement plan for these raw materials to ensure that all raw materials are in place before production begins.

[0117] Secondly, the predicted production information is a core part of the production plan. Based on this, the quantities of various raw materials required can be calculated. At the same time, combined with the delivery time, a reasonable production schedule can be arranged. For example, if the predicted production volume is large and the delivery time is tight, the production plan needs to consider whether to increase the running time of the production line, or whether to temporarily allocate more production equipment and personnel to ensure the completion of the production task on time.

[0118] In addition, the delivery time in the production instruction also involves logistics and storage plans. If the delivery time is late, it may be necessary to plan sufficient storage areas to store the produced concrete, and at the same time, consider how to keep the performance of the concrete unaffected during storage, such as by controlling the temperature and humidity of the storage environment, etc.

[0119] The ratio optimization module is used to determine the optimization goal according to the production plan information, optimize the ratio of raw materials by using the genetic algorithm, and through continuous iteration and optimization, obtain the raw material ratio plan;

[0120] Specifically, this embodiment also includes: a second initialization sub-module, which is used to initialize the population size, the maximum crossover probability and mutation probability, the minimum crossover probability and mutation probability, and the maximum number of iterations G;

[0121] The construction sub-module is used to determine the optimization goal according to the production plan information and construct the optimal ratio objective function of concrete raw materials:

[0122]

[0123] In the formula, f is the value of the optimal ratio objective function, n is the upper limit of summation, indicating the summation of n items of requirements, i is the type of raw materials for the concrete mix, x i are the various raw materials used in the concrete, y i are the unit prices corresponding to the various raw materials;

[0124] The selection operation sub-module is used to encode the ratio of raw materials. After encoding, it is represented as an individual in the genetic algorithm. Set the iteration variable g = 0, calculate the fitness value of each individual in the initial population, and sort the calculation results from large to small. According to the sorting results, divide all individuals in the population into 6 equal parts, double the number of the top 1 / 6 individuals, delete the last 1 / 6 individuals and do not enter the next generation, and copy the middle 2 / 3 individuals and directly inherit them to the next generation;

[0125] The crossover operation sub-module is used to perform a crossover operation on each individual in the population after the selection operation. The crossover probability P cn is:

[0126]

[0127] In the formula, f max is the maximum fitness value of all individuals in the current population, P cn is the crossover probability of individual n in the population, P cnmax is the maximum crossover probability, which is the upper limit value of the crossover probability, P cnmin is the minimum crossover probability, which is the lower limit value of the crossover probability, G is the total number of iterations, g is the current number of iterations, f avg is the average fitness value of all individuals in the population, f n is the fitness value of individual n in the population, P cnmin = 0.6;

[0128] The mutation operation sub-module is used to perform mutation operations on each individual in the population after the crossover operation. The mutation probability P mn is:

[0129]

[0130] In the formula, P mn is the mutation probability of individual n, P mnmax is the maximum mutation probability, which is the upper limit value of the mutation probability, P mnmin is the minimum mutation probability, which is the lower limit value of the mutation probability, P mnmin = 0.5;

[0131] The second judgment sub-module is used to judge whether g is less than G. If not, let g = g + 1 and repeat the genetic operation. If so, judge whether the output result meets the termination condition. If it does, the algorithm terminates. In the last generation population of the genetic algorithm, select the individual with the highest fitness value as the optimal raw material ratio plan, where the termination condition is to reach the maximum number of iterations.

[0132] The production parameter inference module is used to use the factors affecting the concrete production parameters as input variables, the production parameters as output variables, and infer the production parameter plan according to the production plan information by using the fuzzy control algorithm;

[0133] Specifically, this embodiment further includes: As a sub-module, it is used to use the factors affecting the concrete production parameters as input variables and the production parameters as output variables, where the input variables at least include the performance of raw materials and the production environment, and the output variables at least include the mixing time, mixing speed, and water-cement ratio;

[0134] The fuzzification processing sub-module is used to perform fuzzification processing on the input variable X and the output variable Y to convert them into the membership degrees μ(X i ), μ(Y i ) of the corresponding fuzzy sets. It is assumed that there are 3 fuzzy sets for both the input variable and the output variable:

[0135] μ(X i ) = {PS, PM, PB}

[0136] μ(Y i ) = {PS, PM, PB}

[0137] Wherein, PS, PM, and PB are respectively the small, medium, and large fuzzy sets;

[0138] The fuzzy inference sub-module is used to perform fuzzy inference based on the fuzzy sets and a preset fuzzy rule base, calculate the fuzzy sets corresponding to each output variable, defuzzify the fuzzy sets of the output variables by using the maximum membership degree method, obtain the production parameter values, and generate the production parameter scheme, wherein the preset fuzzy rule base includes the configuration relationship between the input variables and the output variables.

[0139] The production plan determination module is used to determine the initial production plan by combining the raw material ratio plan and the production parameter plan, and execute the initial production plan;

[0140] Specifically, this embodiment further includes:

[0141] I. Raw material ratio plan

[0142] Analysis of raw material characteristics: When determining the raw material ratio plan, it is first necessary to conduct a detailed analysis of the characteristics of various concrete raw materials. This includes detecting the grade, setting time, and strength development curve of cement; measuring the particle size distribution, mud content, and crushing index of aggregates; and evaluating the water reduction rate, setting retardation time, air-entraining performance, etc. of admixtures. For example, for cement, if it is Portland cement, it has high early strength but high heat of hydration. When used in large-volume concrete structures, it is necessary to comprehensively consider its dosage and heat dissipation measures. For aggregates, the maximum particle size of coarse aggregates affects the fluidity and pumpability of concrete, while the fineness modulus of fine aggregates affects the workability of concrete.

[0143] Ratio calculation and optimization: According to the design strength grade, durability requirements, and workability requirements of concrete, use professional ratio calculation software or empirical formulas to calculate the preliminary raw material ratio. This process usually uses the mass method or the volume method for calculation. For example, when calculating the water-cement ratio, based on the design strength of concrete and the actual strength of cement, calculate the theoretical water-cement ratio through the Bolomey formula. Then, in combination with the selection of sand ratio (usually determined through experiments), calculate the dosages of cement, water, sand, and stone. Further, use advanced optimization algorithms such as genetic algorithms to optimize the preliminary ratio. These algorithms consider multi-objective factors such as raw material cost and environmental impact, and find the optimal raw material ratio plan on the premise of meeting the performance requirements of concrete.

[0144] (2) Production parameter plan

[0145] Determination of mixing parameters: The mixing parameters in the production parameter plan are crucial. It includes determining the type of mixing equipment (such as a forced mixer or a self-falling mixer) and the mixing time. For different types and strength grades of concrete, the mixing time varies. For example, due to the complexity of its raw materials and high requirements for uniformity, high-strength concrete usually requires a longer mixing time, generally 90 - 120 seconds, to ensure the full mixing of cement, aggregates, and admixtures. At the same time, the rotation speed of the mixing equipment also needs to be adjusted according to the working performance of the concrete. Too fast rotation speed may cause aggregate breakage and affect the performance of the concrete; too slow rotation speed may result in uneven mixing.

[0146] Transportation and pouring parameters: The transportation parameters involve selecting appropriate transportation equipment and determining the transportation route. For ready-mixed concrete, the transportation time and mixing speed of the mixer truck need to be strictly controlled to prevent excessive slump loss of the concrete during transportation. The pouring parameters include determining the pouring method (such as pumping, chute, tower crane hoisting, etc.) and the pouring speed. When pouring large-volume concrete, a layered pouring plan needs to be formulated to control the pouring thickness and interval time of each layer to prevent cold joints in the concrete.

[0147] (3) Determination of the initial production plan

[0148] Scheme integration: Integrate the optimized raw material proportioning scheme and the determined production parameter scheme. This process needs to consider the mutual influence and synergy between the two. For example, the type and dosage of admixtures in the raw material proportioning will affect the initial setting time of the concrete, which in turn affects the setting of transportation and pouring parameters. If the admixture has a retarding effect, the transportation and pouring time arrangements can be relatively loose, but at the same time, it is necessary to consider the problems such as slow strength development caused by too long retarding time.

[0149] II. Execution of the initial production plan

[0150] (1) Preparation of raw materials

[0151] Procurement and inspection: According to the raw material proportioning scheme, procure the required raw materials such as cement, aggregates, and admixtures. During the procurement process, strictly select suppliers according to quality standards and inspect the incoming raw materials. For example, the inspection of cement includes checking its factory certificate, test report, and conducting sampling re-tests. The test items include strength, soundness, setting time, etc. For aggregates, check indicators such as particle size distribution and mud content to ensure that the quality of the raw materials meets the requirements of the production plan.

[0152] Storage and batching: After the raw materials enter the site, they are stored according to the specified storage conditions. Cement should be stored in a dry and well-ventilated warehouse to prevent moisture absorption and caking; aggregates should be stacked separately according to different particle sizes, and rain-proof facilities should be set up. In the batching stage, precise metering equipment is used to weigh the raw materials according to the mixing ratio plan. For example, an electronic scale is used to accurately measure cement and admixtures, and a loader and belt scale are used to measure aggregates to ensure the accuracy of batching.

[0153] (2) Production operation

[0154] Mixing operation: Start the mixing equipment according to the mixing parameters in the production parameter plan. First, put the aggregates into the mixer. After mixing for a certain time, then add cement and admixtures, and finally add water for mixing.

[0155] During the mixing process, the operator should closely observe the operation of the mixer, such as parameters like current and rotation speed, to ensure uniform mixing and compliance with the specified mixing time.

[0156] The data management module is used to receive real-time data during the concrete production process, and analyze and process it using the DVAE-WAFFN-WGAN-GP-LSTM model according to the real-time data, extract key feature variables, evaluate the performance indicators of the concrete, obtain the concrete test results, and adjust the initial production plan according to the concrete test results to obtain the target production plan.

[0157] Specifically, this embodiment also includes: a collection sub-module, which is used to collect real-time data during the concrete production process through sensors installed on the production equipment, and input the collected real-time data into the DVAE-WAFFN-WGAN-GP-LSTM model;

[0158] The encoding and decoding sub-module is used to extract production feature variables through the encoding and decoding process of the DVAE-WAFFN-WGAN-GP-LSTM model, and perform dimensionality reduction processing on the extracted production feature variables using the principal component analysis algorithm to obtain key feature variables, where the key feature variables at least include mixing time, mixing speed, water-cement ratio, ambient temperature and concrete temperature, ambient humidity;

[0159] The evaluation sub-module is used to evaluate the performance indicators of the concrete during the production process according to the extracted key feature variables, and output the concrete test results, where the performance indicators at least include strength, fluidity and durability;

[0160] For strength assessment, the trained regression model is used to predict the strength value of concrete based on key characteristic variables. Support vector regression (SVR), random forest regression (RFR) and other algorithms can be used for strength prediction. The predicted strength value is compared with the set strength grade requirements to determine whether the strength of the concrete meets the standard.

[0161] For fluidity assessment, the fluidity is assessed by analyzing the parameters related to fluidity in the key characteristic variables, such as slump, expansion, etc., using fuzzy logic or neural network methods. The assessment results are divided into different levels, such as good, medium, poor, etc., so as to intuitively understand the fluidity of concrete.

[0162] For durability assessment, the parameters reflecting the concrete's impermeability, frost resistance, corrosion resistance, etc. are considered in the key characteristic variables. A durability assessment index system can be established, and the weight of each index can be determined by using methods such as the analytic hierarchy process (AHP), and then the durability score of the concrete can be calculated comprehensively. According to the score, it can be judged whether the durability of the concrete meets the requirements.

[0163] The analysis submodule is used to analyze the concrete test results, determine the performance indicators that are different from the concrete production goals, and analyze the real-time data generated during the production process to determine the material ratio and production parameters to be adjusted;

[0164] The adjustment submodule is used to adjust the initial production plan according to the material ratio and production parameters to be adjusted to obtain the target production plan.

[0165] Generate transportation and pouring measures according to the target production plan;

[0166] Transportation management: Load the mixed concrete into transportation equipment, such as a mixer truck. During transportation, keep the mixer truck tank rotating to prevent concrete segregation. According to the transportation distance and traffic conditions, reasonably plan the transportation route to ensure that the concrete reaches the construction site before initial setting. At the same time, use GPS and other technical means to monitor the transportation vehicles in real time to keep track of the transportation situation.

[0167] Pouring implementation: After the concrete is delivered to the construction site, it is poured according to the pouring parameters in the production parameter plan. Construction personnel must strictly follow the operating specifications such as layered pouring and vibrating and compacting. During the pouring process, pay attention to the fluidity and filling of the concrete, and deal with any abnormalities in a timely manner. For example, if it is found that the concrete does not flow smoothly in the formwork, it can be solved by adjusting the vibration method or increasing the vibration time.

[0168] The encoding and decoding submodules specifically include the following processes:

[0169] Input the collected real-time data into the variational autoencoder DVAE. Through the DVAE, compress the high-dimensional input data into a low-dimensional latent representation, and then input it into the WAFFN. Concatenate the output layer features F1 and F2 of the variational autoencoder DVAE in terms of dimensions to generate spatio-temporal feature information f 12 Input the spatio-temporal feature information into the activation function for non-linear transformation to obtain the feature f′ 12 :

[0170] f 12 =[F1,F2]

[0171] f′ 12 =ReLU(W·f 12 +b)

[0172] In the formula, W and b represent the weight matrix and the bias vector, and ReLU is the activation function;

[0173] Perform the softmax operation on the feature f′ 12 to generate the weight vectors w1, w2 of each output layer feature of the variational autoencoder:

[0174]

[0175] Perform weighted fusion on the calculated weight vectors and the variational encoder output features:

[0176]

[0177] In the formula, F U is the fusion feature, is the tensor multiplication operation, w j represents the jth parameter vector for calculating the weight, is the normalization term to ensure that the sum of the weights is 1;

[0178] Decode the latent representation encoded by the WAFFN through the WGAN-GP and the LSTM to reconstruct an approximate representation of the input real-time data. At the beginning of the encoding, the LSTM layer processes the time series data. At the last stage of the decoding, the LSTM layer will adjust the reconstructed data in the time series dimension according to the time dependence relationship. Through the encoding and decoding processes, extract the production feature variables that can reflect the concrete production process.

[0179] 1. Regarding the basic principle of WGAN-GP

[0180] Generative Adversarial Network (GAN): A GAN consists of a Generator and a Discriminator. The goal of the Generator is to generate data that is as realistic as possible, while the task of the Discriminator is to distinguish between real data and the data generated by the Generator. During the training process, the two play against each other, and ultimately the Generator can generate data with a distribution similar to that of the real data.

[0181] WGAN: Wasserstein GAN is an improvement over the traditional GAN. It uses the Wasserstein distance (also known as the Earth-Mover distance) to measure the difference between the distribution of the generated data and the real data. Compared with the Jensen-Shannon divergence used in traditional GANs, the Wasserstein distance has better properties during the optimization process.

[0182] WGAN-GP: WGAN-GP adds a Gradient Penalty term on the basis of WGAN. The role of the gradient penalty is to further constrain the behavior of the Discriminator, making the Discriminator satisfy the Lipschitz continuity condition, thereby ensuring a more stable training process and avoiding problems such as mode collapse.

[0183] 2. Interaction between WGAN-GP and the encoding process - The role of the Discriminator

[0184] Evaluating the authenticity of latent variables: When interacting with the encoding process, the Discriminator receives the latent variables output by the DVAE. Under the concept of generative adversarial, the Discriminator attempts to determine whether the information represented by this latent variable comes from the encoding of real data (ideally, after encoding, real data should have a certain specific distribution in the latent space) or is forged by the Generator. For example, if the latent variables are properly encoded, their distribution in the latent space should match the distribution characteristics of the real data encoding learned by the Discriminator, and the Discriminator will output a relatively high probability indicating that it is "real"; conversely, if the latent variables do not conform to this distribution characteristic, the Discriminator will judge it as "false".

[0185] Influencing the latent space representation: The judgment result of the Discriminator is fed back to the entire model. If the Discriminator believes that some latent variables do not conform to the characteristics of real data, it means that the encoding process may need to be adjusted. This feedback can guide the model to improve the encoding method, such as adjusting the encoder parameters of the DVAE to make the distribution of the latent variables more in line with the expected distribution of the real data encoding.

[0186] 3. Interaction between WGAN-GP and the encoding process - The role of the Generator

[0187] Optimizing the latent space representation: The generator also plays a crucial role in this interaction process. The generator attempts to generate latent variables (in the latent space) that conform to the real data distribution. It can adjust its generation strategy based on the discriminator's feedback. For example, if the discriminator indicates that there are differences between the currently generated latent variables and the latent variables encoded from real data (such as differences in distribution), the generator will change its parameters to produce latent variables that are closer to the latent variable encoding results of real data.

[0188] Making the generated samples conform to the real data distribution: In this way, the interaction between the generator and the discriminator not only affects the generation of latent variables but also indirectly affects the entire encoding-decoding process. The ultimate goal is to make the samples reconstructed from the optimized latent variables (in this scenario, the data related to production feature variables) during decoding more conform to the distribution of real production data, thereby improving the accuracy and effectiveness of the model in extracting production feature variables. For example, during the production process, if the real data has specific patterns (such as different distribution patterns of feature variables during normal production and faulty production), this interaction can help the model better learn and capture these patterns.

[0189] The beneficial effects are as follows. By analyzing the historical inventory data of concrete raw materials, predicting the inventory trend of concrete raw materials, and obtaining the inventory prediction result, inventory management of concrete raw materials can be carried out; the PSO algorithm is used to optimize the SVM model to classify the quality of concrete raw materials, and the quality result classification is obtained to monitor the quality of concrete raw materials; when a concrete production instruction is received, production plan information is obtained according to the production instruction; an optimization target is determined according to the production plan information, and the genetic algorithm is used to optimize the raw material ratio. Through continuous iteration and optimization, a raw material ratio plan is obtained; the factors affecting concrete production parameters are used as input variables, the production parameters are used as output variables, and a production parameter plan is inferred by using the fuzzy control algorithm according to the production plan information; an initial production plan is determined by combining the raw material ratio plan and the production parameter plan, and the initial production plan is executed; real-time data during the concrete production process is received, and the DVAE-WAFFN-WGAN-GP-LSTM model is used for analysis and processing according to the real-time data to extract key feature variables, evaluate the performance indexes of concrete, obtain the concrete test result, and adjust the initial production plan according to the concrete test result to obtain the target production plan. 1. By analyzing the historical inventory data of concrete raw materials to predict the inventory trend, the inventory situation of raw materials can be understood in advance. It helps to avoid production stagnation caused by raw material shortages, ensure the continuity of concrete production, and reasonable inventory prediction can reduce unnecessary inventory backlogs and lower inventory costs, including warehousing costs and capital occupation costs. 2. The quality status of raw materials can be quickly and accurately identified. It helps to promptly discover raw materials with unqualified quality and prevent them from entering the production process, thus ensuring the quality of concrete. 3. Using the genetic algorithm to optimize the raw material ratio can comprehensively consider various factors and find the optimal raw material ratio plan through continuous iteration. It can ensure that while meeting the quality requirements of concrete, the cost is minimized to make the raw material ratio more in line with the actual production needs and improve production efficiency. 4. By inferring the production parameter plan through the fuzzy control algorithm, various complex factors affecting production can be fully considered. It helps to maintain stable production conditions during the production process and improve the consistency and reliability of concrete production. 5. Timely extracting key feature variables and evaluating the performance indexes of concrete can ensure that the quality of the produced concrete always meets the requirements and avoid quality fluctuations.

[0190] Please refer to Figure 2 , the schematic diagram of the third embodiment of a concrete digital management and control system provided by an embodiment of the present invention. In this system, the raw material quality monitoring module includes a first initialization sub-module, a segmentation sub-module, an update sub-module, a comparison sub-module, a first judgment sub-module, and a classification sub-module, where:

[0191] The first initialization sub-module is used to initialize the particle swarm capacity, maximum number of iterations, acceleration coefficients, and inertia parameters of the PSO algorithm, initialize the initial values and search ranges of the penalty factor parameter and kernel function parameter of the SVM model, and collect the physical and chemical property data of the concrete raw materials to obtain quality characteristic samples;

[0192] The segmentation sub-module is used to segment the quality characteristic samples into multiple non-overlapping subsets to obtain quality characteristic subsets, initialize the iteration counter to 1, select the first n pairs of quality characteristic subsets as the training set, and other subsets as the test set;

[0193] The update sub-module is used to perform a search within the search range using the PSO algorithm, assign the initial value to the SVM model, calculate the initial fitness value, perform the iterative process of the PSO algorithm, and update the velocity and position of the particle individuals:

[0194]

[0195] In the formula, k is the current iteration number of the algorithm; ω is the inertia parameter; and are the velocity and position of the i-th particle at the k-th iteration, and d is the space dimension; is the value of the personal best position of the i-th particle in the space dimension d at the k-th iteration, is the value of the global best position in the space dimension d at the k-th iteration, and are the velocity and position of the i-th particle at the (k + 1)-th iteration, c1 and c2 are both acceleration coefficients, r1 and r2 are random functions, r1, r2 ∈ [0, 1], η is the contraction factor, is the total acceleration coefficient, K is the contraction coefficient, 0 < K < 1;

[0196] The comparison sub-module is used to calculate the new fitness value of each particle every iteration, compare the current optimal fitness value with the new fitness values of each particle, and if the new fitness value of the particle is better than the optimal fitness value, then replace the current optimal fitness value with the new fitness value of the particle;

[0197] The first judgment sub-module is used to judge whether the current optimal fitness value reaches the maximum number of iterations. If not, the iteration number is incremented by 1 for the next round of iteration. If so, the penalty factor parameter and kernel function parameter are updated to obtain the optimal SVM model;

[0198] The classification sub-module is used to classify the quality of the concrete raw materials using the optimal SVM model, divide the raw materials into three categories: qualified, unqualified, and to be detected, to obtain the quality result classification, so as to process the unqualified raw materials in time.

[0199] The beneficial effects are as follows. By implementing the above solution, the PSO algorithm is used to optimize the SVM model to classify the quality of concrete raw materials, which can quickly and accurately identify the quality status of raw materials. It helps to promptly detect raw materials with unqualified quality and prevent them from entering the production process, thereby ensuring the quality of concrete. Efficient quality monitoring can reduce product defects caused by raw material quality problems, lower the rework and scrap rates, and save production costs.

[0200] Please refer to Figure 2 , the schematic diagram of the second embodiment of a concrete digital management and control system provided by the embodiment of the present invention. In this system, the ratio optimization module includes a second initialization sub-module, a construction sub-module, a selection operation sub-module, a crossover operation sub-module, a mutation operation sub-module, and a second judgment sub-module, where:

[0201] The second initialization sub-module is used to initialize the population size, the maximum crossover probability and mutation probability, the minimum crossover probability and mutation probability, and the maximum number of iterations G;

[0202] The construction sub-module is used to determine the optimization goal according to the production plan information and construct the optimal ratio objective function of concrete raw materials:

[0203]

[0204] In the formula, f is the value of the optimal ratio objective function, n is the upper limit of summation, indicating the summation of n items of requirements, i is the type of raw materials for the concrete mix, x i are various raw materials used in the concrete, and y i are the unit prices corresponding to various raw materials;

[0205] The selection operation sub-module is used to encode the ratio of raw materials. After encoding, it is represented as an individual in the genetic algorithm. Set the iteration variable to g = 0, calculate the fitness value of each individual in the initial population, sort the calculation results from large to small, and divide all individuals in the population into 6 equal parts according to the sorting results. Double the number of the top 1 / 6 individuals, delete the last 1 / 6 individuals, which do not enter the next generation, and copy the middle 2 / 3 individuals and directly inherit them to the next generation;

[0206] The crossover operation sub-module is used to perform a crossover operation on each individual in the population after the selection operation. The crossover probability P cn is:

[0207]

[0208] In the formula, f max is the maximum fitness value of all individuals in the current population, and P cnis the crossover probability of individual n in the population, P cbmax is the maximum crossover probability, which is the upper limit value of the crossover probability, P cnmin is the minimum crossover probability, which is the lower limit value of the crossover probability, G is the total number of iterations, g is the current number of iterations, f avg is the average fitness value of all individuals in the population, f n is the fitness value of individual n in the population, P cnmin = 0.6;

[0209] The mutation operation sub-module is used to perform mutation operations on each individual in the population after the crossover operation. The mutation probability P mn is:

[0210]

[0211] In the formula, P mn is the mutation probability of individual n, P mnmax is the maximum mutation probability, which is the upper limit value of the mutation probability, P mnmin is the minimum mutation probability, which is the lower limit value of the mutation probability, P mnmin = 0.5;

[0212] The second judgment sub-module is used to judge whether g is less than G. If not, let g = g + 1 and repeat the genetic operation. If so, judge whether the output result meets the termination condition. If it does, the algorithm terminates. In the last generation population of the genetic algorithm, select the individual with the highest fitness value as the optimal raw material ratio plan, where the termination condition is to reach the maximum number of iterations.

[0213] Its beneficial effects are as follows. By implementing the above scheme, using the genetic algorithm to optimize the raw material ratio can comprehensively consider various factors and find the optimal raw material ratio plan through continuous iteration. It can ensure that while meeting the quality requirements of concrete, the cost is minimized, for example, reducing the usage of expensive raw materials. Determining the optimization goal based on the production plan information makes the raw material ratio more in line with the actual production requirements and improves production efficiency.

[0214] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A digital management and control system for concrete, characterized in that, The concrete digital management control system includes: A data concatenation sub-module, which is used to obtain the historical inventory data of concrete raw materials, and concatenate the inventory feature data in the historical inventory data at each moment using word vectors to form a new time series, where the inventory feature data at least includes the purchase quantity, consumption quantity, available days, and inventory quantity; A component reconstruction sub-module, which is used to decompose the historical inventory data to obtain 1 residual component and 8 IMF components, obtaining 9 components, and then reconstructing the components into 3 components with different frequency characteristics through the ZCR coefficient; A capture sub-module, which is used to capture long-term dependence information on the time series x of components with different frequency characteristics through the TCN module; Among them, F(t) is the output after one dilated convolution operation, d is the sampling rate, that is, the window coefficient, k is the convolution kernel size, f(i) is the i-th element in the convolution kernel, and x t-d·i is for convolution operation only on past data; A feature vector extraction sub-module, which is used to input the extracted dependence information into the BiGRU prediction module to extract feature vectors in both forward and reverse directions and analyze the time series data; A weight assignment sub-module, which is used to use the Attention mechanism to assign weights to the input information of the BiGRU prediction module to highlight the contribution degree of important information and then make a prediction, and superimpose the prediction results of each frequency component to obtain an inventory prediction result, and automatically issue a purchase warning according to the inventory prediction result to remind the management personnel to make a purchase; A raw material quality monitoring module, which is used to optimize the SVM model using the PSO algorithm to classify the quality of concrete raw materials to obtain a quality result classification for quality monitoring of concrete raw materials; A production plan acquisition module, which is used to obtain production plan information according to the production instruction when receiving a concrete production instruction; A second initialization sub-module, which is used to initialize the population size, the maximum crossover probability and mutation probability, the minimum crossover probability and mutation probability, and the maximum number of iterations G; A construction sub-module, which is used to determine an optimization target according to the production plan information and construct an optimal mixture ratio objective function for concrete raw materials; where f is the value of the optimal proportion objective function, n is the upper limit of summation, indicating the summation of n demands, i is the type of raw materials for the concrete mix proportion, x i are the various raw materials used in the concrete, and y i are the unit prices corresponding to the various raw materials; A selection operation sub-module, which is used to encode the mixture ratio of raw materials, and after encoding, it is represented as an individual in the genetic algorithm. Set the iteration variable g = 0, calculate the fitness value of each individual in the initial population, and sort the calculation results from largest to smallest. According to the sorting results, divide all individuals in the population into 6 equal parts on average, double the number of the top 1 / 6 individuals, delete the last 1 / 6 individuals and do not enter the next generation, and copy the middle 2 / 3 individuals and directly inherit them to the next generation; The crossover operation sub-module is used to perform a crossover operation on each individual in the population after the selection operation, and the crossover probability P cn is as follows: where f max is the maximum fitness value of all individuals in the current population, P cn is the crossover probability of individual n in the population, P cnmax is the maximum crossover probability and the upper limit value of the crossover probability, P cnmin is the minimum crossover probability and the lower limit value of the crossover probability, G is the total number of iterations, g is the current number of iterations, f avg is the average fitness value of all individuals in the population, f n is the fitness value of individual n in the population, P cnmin = 0.6; Mutation operator sub-module, which is used to perform mutation operations on each individual in the population after crossover operations, with a mutation probability P mn being: where P mn is the mutation probability of individual n, P mnmax is the maximum mutation probability, which is the upper limit value of the mutation probability, and P mnmin is the minimum mutation probability, which is the lower limit value of the mutation probability, and P mnmin = 0.5; A second judgment sub-module, which is used to judge whether g is less than G. If not, let g = g + 1 and repeat the genetic operation. If so, judge whether the output result meets the termination condition, and the algorithm terminates. In the last generation population of the genetic algorithm, select the individual with the highest fitness value as the optimal raw material mixture ratio scheme, where the termination condition is to reach the maximum number of iterations; A production parameter inference module, which is used to use the factors affecting concrete production parameters as input variables and the production parameters as output variables, and infer a production parameter scheme using the fuzzy control algorithm according to the production plan information; The production plan determination module is used to determine the initial production plan in combination with the raw material ratio plan and the production parameter plan, and execute the initial production plan; The acquisition sub-module is used to collect real-time data during the concrete production process in real time through sensors installed on the production equipment, and input the collected real-time data into the DVAE-WAFFN-WGAN-GP-LSTM model; The encoding and decoding sub-module is used to extract production feature variables through the encoding and decoding processes of the DVAE-WAFFN-WGAN-GP-LSTM model, and perform dimensionality reduction processing on the extracted production feature variables using the principal component analysis algorithm to obtain key feature variables, where the key feature variables at least include mixing time, mixing speed, water-cement ratio, ambient temperature, temperature of concrete, and ambient humidity; The evaluation sub-module is used to evaluate the performance indicators of the concrete during the production process according to the extracted key feature variables, and output the concrete detection results, where the performance indicators at least include strength, fluidity, and durability; The analysis sub-module is used to analyze the concrete detection results, determine the performance indicators that are different from the concrete production target, and analyze the real-time data generated during the production process to determine the material ratio and production parameters to be adjusted; The adjustment sub-module is used to adjust the initial production plan according to the material ratio and production parameters to be adjusted to obtain the target production plan.

2. The digital management and control system for concrete according to claim 1, wherein The raw material quality monitoring module includes a first initialization sub-module, a segmentation sub-module, an update sub-module, a comparison sub-module, a first judgment sub-module, and a classification sub-module, where, The first initialization sub-module is used to initialize the particle swarm capacity, maximum number of iterations, acceleration coefficient, and inertia parameter of the PSO algorithm, initialize the initial values and search ranges of the penalty factor parameter and kernel function parameter of the SVM model, and collect the physical and chemical property data of the concrete raw materials to obtain quality feature samples; The segmentation sub-module is used to segment the quality feature samples into multiple non-overlapping subsets to obtain quality feature subsets, initialize the iteration counter to 1, select the first n pairs of quality feature subsets as the training set, and other subsets as the test set; The update sub-module is used to perform a search within the search range using the PSO algorithm, assign the initial value to the SVM model, calculate the initial fitness value, perform the iterative process of the PSO algorithm, and update the velocity and position of the particle individuals: where k is the current iteration number of the algorithm; ω is the inertia parameter; and are the velocity and position of the i-th particle at iteration k, and d is the spatial dimension; is the value of the personal best position of the i-th particle in the spatial dimension d at iteration k, is the value of the global best position in the spatial dimension d at iteration k, and are the velocity and position of the i-th particle at iteration k + 1, c1 and c2 are both acceleration coefficients, r1 and r2 are random functions, r1, r2 ∈ [0, 1], η is the contraction factor, is the total acceleration coefficient, K is the contraction coefficient, 0 < K < 1; The comparison sub-module is used to calculate the new fitness value of each particle every iteration and compare the pros and cons of the current optimal fitness value and the new fitness values of each particle. If the new fitness value of the particle is better than the optimal fitness value, then replace the current optimal fitness value with the new fitness value of the particle; The first judgment sub-module is used to judge whether the current optimal fitness value reaches the maximum number of iterations. If not, the number of iterations is incremented by 1 for the next round of iteration. If so, the penalty factor parameter and kernel function parameter are updated to obtain the optimal SVM model; A classification sub-module is used to classify the quality of concrete raw materials using the optimal SVM model, classifying the raw materials into three categories: qualified, unqualified, and pending inspection, obtaining the quality result classification, so as to process the unqualified raw materials in a timely manner.

3. The digital management control system for concrete according to claim 1, wherein The production parameter inference module includes an as-is sub-module, a fuzzification processing sub-module, and a fuzzy inference sub-module, where The as-is sub-module is used to take the factors affecting concrete production parameters as input variables and the production parameters as output variables. The input variables at least include the performance of raw materials and the production environment, and the output variables at least include mixing time, mixing speed, and water-cement ratio; Fuzzification sub-module, which is used to perform fuzzification processing on the input variable X and the output variable Y to convert them into membership degrees μ(X i ) and μ(Y i ) of the corresponding fuzzy sets. It is assumed that there are 3 fuzzy sets for both the input variable and the output variable: μ(X i ) = {PS, PM, PB} μ(Y i ) = {PS, PM, PB} In the formula, PS, PM, and PB are the small, medium, and large fuzzy sets respectively; The fuzzy inference sub-module is used to perform fuzzy inference based on the fuzzy sets and a preset fuzzy rule base, calculate the fuzzy sets corresponding to each output variable, defuzzify the fuzzy sets of the output variables using the maximum membership degree method, obtain the production parameter values, and generate the production parameter plan. The preset fuzzy rule base includes the configuration relationship between the input variables and the output variables.

4. A concrete digital management control system according to claim 1, characterized in that, The encoding and decoding sub-module specifically includes the following process: The collected real-time data is input into the variational autoencoder DVAE. Through the DVAE, the high-dimensional input data is compressed into a low-dimensional latent representation, and then input into the WAFFN. The output layer features F1 and F2 of the variational autoencoder DVAE are concatenated in dimension to generate spatio-temporal feature information f 12 The spatio-temporal feature information is input into an activation function for non-linear transformation to obtain feature f′ 12 : f 12 = [F1,F2] f′ 12 = ReLU(W·f 12 + b) In the formula, W and b represent the weight matrix and the bias vector, and ReLU is the activation function; Perform a normalization exponential function operation on the feature f′ 12 to generate the weight vectors w1, w2 of the features of each variational autoencoder output layer: Perform weighted fusion on the calculated weight vector and the variational encoder output features: where F U is the fused feature, is the tensor multiplication operation, and w j represents the j-th parameter vector for calculating the weight; Decode the latent representation encoded by WAFFN through WGAN-GP and LSTM to reconstruct an approximate representation of the input real-time data. At the beginning of encoding, the LSTM layer processes time series data. At the final stage of decoding, the LSTM layer will adjust the reconstructed data in the time series dimension according to the time dependence relationship. Through the encoding and decoding processes, the production feature variables that can reflect the concrete production process are extracted.

5. A digital management control system for concrete as claimed in claim 1, wherein, The method for implementing the concrete digital management control system includes the following steps: By analyzing the historical inventory data of concrete raw materials, predict the inventory trend of concrete raw materials to obtain the inventory prediction result, so as to manage the inventory of concrete raw materials; Use the PSO algorithm to optimize the SVM model to classify the quality of concrete raw materials, obtain the quality result classification, so as to monitor the quality of concrete raw materials; When receiving a concrete production instruction, obtain the production plan information according to the production instruction; Determine the optimization goal according to the production plan information, use the genetic algorithm to optimize the raw material ratio, and through continuous iteration and optimization, obtain the raw material ratio plan; Take the factors affecting concrete production parameters as input variables and the production parameters as output variables, and use the fuzzy control algorithm to infer the production parameter plan according to the production plan information; Combine the raw material ratio plan and the production parameter plan to determine the initial production plan and execute the initial production plan; Receive real-time data during the concrete production process, and analyze and process it using the DVAE-WAFFN-WGAN-GP-LSTM model according to the real-time data, extract key feature variables, evaluate the performance indicators of the concrete, obtain the concrete test results, and adjust the initial production plan according to the concrete test results to obtain the target production plan.

6. A digital management and control system for concrete as described in claim 5, characterized in that, By analyzing the historical inventory data of concrete raw materials, predicting the inventory trend of concrete raw materials, and obtaining the inventory prediction results for inventory management of concrete raw materials, including: Decompose the historical inventory data to obtain 1 residual component and 8 IMF components, resulting in 9 components, and then reconstruct the components into 3 components with different frequency characteristics through the ZCR coefficient; Capture long-term dependency information for the time series x of components with different frequency characteristics through the TCN module: Among them, F(t) is the output after a single dilated convolution operation, d is the sampling rate, i.e., the window coefficient, k is the convolution kernel size, f(i) is the i-th element in the convolution kernel, and x t-d·i performs convolution operations only on past data; Input the extracted dependency information into the BiGRU prediction module to extract feature vectors in both forward and reverse directions and analyze the time series data; Use the Attention mechanism to assign weights to the input information of the BiGRU prediction module to highlight the contribution degree of important information and then make a prediction, and superimpose the prediction results of each frequency component to obtain the inventory prediction result. Automatically issue a procurement warning according to the inventory prediction result to remind the management personnel to make a purchase.

7. A concrete digital management control system as claimed in claim 5, wherein, Use the PSO algorithm to optimize the SVM model to classify the quality of concrete raw materials and obtain the quality result classification for quality monitoring of concrete raw materials, including: Initialize the particle swarm capacity, maximum iteration number, acceleration coefficient, and inertia parameter of the PSO algorithm, initialize the initial values and search ranges of the penalty factor parameter and kernel function parameter of the SVM model, and collect the physical and chemical property data of concrete raw materials to obtain the quality feature samples; Divide the quality feature samples into multiple non-overlapping subsets to obtain quality feature subsets, initialize the iteration counter to 1, select the first n pairs of quality feature subsets as the training set, and other subsets as the test set; Use the PSO algorithm to conduct a search within the search range, assign the initial value to the SVM model, calculate the initial fitness value, and perform the iterative process of the PSO algorithm to update the velocity and position of the particle individuals: where k is the current iteration number of the algorithm; ω is the inertia parameter; and are the velocity and position of the i-th particle at iteration number k, d is the spatial dimension; is the value of the personal best position of the i-th particle in the spatial dimension d at iteration k, is the value of the global best position in the spatial dimension d at iteration k, and are the velocity and position of the i-th particle at iteration number k + 1, is the current position of particle i in the spatial dimension d at iteration k; c2 are both acceleration coefficients, r1, r2 are random functions, r1, r2 ∈ [0, 1], η is the contraction factor, is the total acceleration coefficient, K is the contraction coefficient, 0 < K < 1; Calculate the new fitness value of each particle for each iteration and compare the current optimal fitness value with the new fitness values of each particle. If the new fitness value of a particle is better than the optimal fitness value, replace the current optimal fitness value with the new fitness value of this particle; Judge whether the current optimal fitness value reaches the maximum iteration number. If not, increase the iteration number by 1 for the next round of iteration. If so, update the penalty factor parameter and kernel function parameter to obtain the optimal SVM model; Use the optimal SVM model to classify the quality of concrete raw materials, divide the raw materials into three categories: qualified, unqualified, and to be tested, and obtain the quality result classification to process the unqualified raw materials in a timely manner.

Images