Concrete mix proportion optimization system and method based on industrial data analysis

Abstract

The invention relates to the technical field of industrial data analysis, in particular to a concrete mix proportion optimization system and method based on industrial data analysis, comprising a data acquisition and sensing module, a data storage and processing module, a core analysis module and an application output module; compared with the defects of long period, high cost and difficulty in coping with fluctuation of raw materials due to the fact that the prior art mainly depends on laboratory trial and matching and experiences of engineers, the method has the advantages that the performance prediction model based on machine learning and multi-source real-time industrial data fusion analysis are adopted, and the concrete performance can be rapidly and accurately predicted; the mixing proportion design efficiency and scientificity are remarkably improved, and fundamental conversion from experience driving to data driving is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial data analysis, and in particular to a concrete mix proportion optimization system and method based on industrial data analysis. BACKGROUND

[0002] Concrete mix proportion design is the core link of determining the performance, cost and environmental benefits of concrete, and its goal is to determine the optimal combination ratio of raw materials such as water, cement, aggregate, admixture and additive. The traditional method mainly relies on the experience of engineers, a large number of laboratory test mixing and reference to specification table, the whole process is not only cumbersome and time-consuming, but also seriously depends on subjective judgment. With the increasing requirements of the construction industry on the performance, cost control and green production of concrete, this traditional mode dominated by experience has been difficult to meet the needs of modern concrete intelligent manufacturing and fine management.

[0003] The prior art is usually limited to single objective optimization, such as only pursuing strength compliance or minimum cost, and lacks comprehensive trade-off of multi-objectives such as workability, long-term durability and carbon emissions. The historical data relied on are mostly static laboratory records, which fail to effectively integrate real-time fluctuation information in the production process, for example, the change of aggregate moisture content will directly cause the actual water-cement ratio to deviate from the design value, resulting in distorted performance prediction. In addition, the existing method almost completely ignores the impact of raw material supply chain fluctuations on cost optimization, and lacks scientific prediction means for the microstructure and long-term performance of concrete, resulting in that the mix proportion design scheme is often feasible in the short term but lacks long-term durability, or cannot achieve true global optimization, making it difficult to support the application requirements of high-performance concrete and complex engineering.

[0004] The present application aims to solve the above problems by constructing a concrete mix proportion optimization system based on industrial data analysis. By deploying online sensors to capture raw material characteristics and production process data in real time, and integrating supply chain information to build a high-quality data set, a machine learning model is used to accurately predict concrete performance, combined with digital twin technology to simulate its long-term durability. Finally, a multi-objective optimization algorithm is used to automatically generate an optimal mix proportion scheme under multiple constraints such as performance, cost, environmental protection, and output a recommended ranking, thereby realizing an innovative breakthrough from experience-driven to data-driven, from single objective to global optimization, and from short-term performance to full life cycle value in the mix proportion design. SUMMARY

[0005] In order to overcome the problems proposed in the background art, the present application proposes a concrete mix proportion optimization system and method based on industrial data analysis.

[0006] The technical solution of the present application is: a concrete mix proportion optimization system based on industrial data analysis, comprising: A data acquisition and perception module for acquiring multi-source industrial data of concrete production; A data storage and processing module for cleaning, data fusion and feature engineering of the multi-source industrial data collected by the data acquisition and perception module; A core analysis module for predicting concrete performance through a machine learning model and weighing cost, performance and environmental protection indicators using a multi-objective optimization engine to finally output an optimal mix proportion scheme; An application output module for outputting the optimized mix proportion scheme and control instructions.

[0007] As preferred, the data acquisition and perception module specifically includes: A11: a near-infrared spectrometer for real-time monitoring of the water content of aggregates; A12: a laser particle size analyzer for detecting the fineness of powdery raw materials; A13: a densitometer for detecting the concentration and solid content of water-reducing agents; A14: a machine vision system for real-time analysis of aggregate particle gradation and particle shape through image processing and deep learning algorithms; A15: a pH sensor for monitoring the pH of water-reducing agent solution; A16: an array of temperature sensors for monitoring raw material temperature; A17: an online viscometer for real-time monitoring of the rheological properties of the mixture.

[0008] The data acquisition and perception module also accesses macro data of external supply chain systems through a data interface, including futures market prices, inventory levels and logistics information of raw materials.

[0009] As preferred, the core analysis module specifically includes: A21: a machine learning-based performance prediction model for training the model using historical data to quickly and accurately predict key performance indicators such as strength and workability of concrete under a specified mix proportion; A22: a multi-objective optimization engine for weighing and searching under multiple objectives and constraints such as performance, cost, environmental protection, etc., to automatically generate an optimal mix proportion scheme set; A23: a digital twin simulation unit for simulating the hydration process and microstructure evolution of concrete based on physical mechanism models to predict its long-term durability and service performance.

[0010] As preferred, the data storage and processing module specifically includes: S11: receiving raw data stream from data collection and perception module, wherein the raw data stream includes real-time raw material characteristic data from online monitoring sensors, process time series data from production equipment, mix proportion and performance data from laboratory, and macro supply chain data accessed through data interface; S12: data cleaning on the raw data stream, including handling missing values, removing outliers due to sensor failure or transmission interference, and uniting and standardizing the format of data from different sources; S13: spatio-temporal alignment and fusion of the cleaned multi-source data, the spatio-temporal alignment including applying uniform timestamps to all data and associating based on formula number and production batch number to construct complete data records for analysis; S14: performing feature engineering operations on the fused data, including: A. calculating derived features including water-binder ratio, total amount of cementitious materials, and sand ratio; B. extracting statistical features from process time series data, including mean, variance of mixing current, and integral value in a specific time interval; C. preprocessing macro supply chain data to generate features reflecting future trends of raw material prices or inventory costs; S15: outputting the high-quality data set after feature engineering to the core analysis module for training and inference based on the machine learning-based performance prediction model.

[0011] As a preferred, the machine learning-based performance prediction model, when working, specifically includes: S21: receiving data set output from data storage and processing module, which contains historical mix proportion data, corresponding real-time raw material characteristic data, process data features, and macro features; S22: dividing the data set into training set and test set according to timestamp, and training the preset machine learning algorithm using the training set to fit the complex nonlinear mapping relationship between mix proportion features and concrete performance indicators, wherein when training the preset machine learning algorithm using the training set, the objective function is: ; wherein, is the optimal model parameter, is the model parameter, is the empirical loss term, represents the prediction result of the model M for the i-th input feature vector under given parameters , is the regularization term, is the regularization function, is a regularization coefficient; S23: deploying the trained model in a production environment, receiving new mix proportion schemes and corresponding real-time sensor data, and generating predicted values of key performance indicators of concrete; S24: outputting the predicted values to the multi-objective optimization engine as the core basis for scheme optimization and evaluation.

[0012] The performance prediction model based on machine learning is trained using a gradient boosting tree algorithm or a neural network algorithm, and the input features include water-binder ratio, cementitious material composition, real-time aggregate moisture content, real-time powder fineness, and environmental temperature and humidity. The output labels include the compressive strength and slump of the concrete.

[0013] The multi-objective optimization engine uses a genetic algorithm or a particle swarm algorithm to search for a Pareto optimal solution set under the conditions of meeting the constraints of national standard specifications and raw material availability, with the cost function, performance function, and sustainability function as multiple objectives. As a preferred embodiment, the multi-objective optimization engine, when in operation, specifically includes: S31: receiving performance prediction values for a set of candidate mix proportion schemes from the performance prediction model based on machine learning, and receiving long-term durability index prediction values from the digital twin simulation unit; S32: defining a plurality of optimization objective functions, wherein the defined optimization objective functions specifically include a cost objective function, a performance objective function, and an environmental protection objective function; S33: defining a set of constraint conditions, wherein the defined set of constraint conditions includes performance constraints, durability constraints, and formulation constraints; S34: using a multi-objective optimization algorithm to perform iterative search within the formulation solution space based on the objective functions and constraint conditions, generating a set of Pareto optimal solutions, wherein each solution in the set of Pareto optimal solutions represents an optimal trade-off scheme between multiple objectives; S35: using a weighted fusion method to fuse the cost and performance, obtaining a priority index for each solution in the optimal solution set, and sorting and outputting the Pareto optimal solution set according to the calculated priority index.

[0015] As a preferred embodiment, when defining a plurality of optimization objective functions, the defined optimization objective functions specifically include: A31: cost objective function: ; wherein, is the price of the jth raw material, is the unit consumption of the jth raw material, and M is the number of raw material types; A32: performance objective function: ; wherein, is the 28-day compressive strength prediction value of the performance prediction model for the scheme x, taking a negative value indicates pursuing the strength maximization; A33: environmental protection objective function: ; wherein, is the unit carbon emission factor of the jth raw material.

[0016] As preferred, when the cost and performance are fused in a weighted fusion manner to obtain the priority index of each solution in the optimal solution set, the principle formula is: ; wherein, is the pre-defined weight coefficient of the cost objective function, is the pre-defined objective function of the performance objective function, is the minimum value of the cost objective function value in all solutions in the Pareto optimal solution set, is the maximum value of the cost objective function value in all solutions in the Pareto optimal solution set, is the minimum value of the cost performance function value in all solutions in the Pareto optimal solution set, is the maximum value of the cost performance function value in all solutions in the Pareto optimal solution set, is the estimated total cost of raw materials required for producing one unit of concrete when the ith optimal mix proportion scheme is adopted, is the performance value of the ith optimal mix proportion scheme .

[0017] As preferred, the digital twin simulation unit, when working, specifically comprises: S41: receiving standardized mix proportion data from the data storage and processing module, including water-binder ratio, chemical composition and proportion of each binder material, aggregate characteristics and initial curing conditions; S42: calling and initializing a micro-model based on physical mechanism, setting the initial boundary conditions and material parameters of the model according to the input mix proportion data; S43: iteratively performing simulation calculation of the micro-model within a set time step, simulating the hydration reaction kinetics of cement particles, the evolution process of micro-pore structure, and the transport and diffusion of water and ions in the pore network; S44: calculating and outputting the long-term performance and durability index prediction value based on the simulation results; S45: The predicted values ​​of long-term performance and durability indicators are fed to the multi-objective optimization engine as durability constraints or optimization objectives when it performs multi-objective optimization.

[0018] The method for optimizing concrete mix proportions based on industrial data analysis includes the following steps: S51: Data Acquisition and Sensing, real-time acquisition of multi-source industrial data such as raw material characteristics, production process, laboratory data and supply chain information through sensors and interfaces; S52: Data preprocessing and fusion, cleaning, standardizing, spatiotemporally aligning and feature engineering multi-source data to build high-quality datasets for analysis; S53: Performance prediction, using machine learning models to quickly predict key performance indicators of concrete with a specified mix proportion based on processed data; S54: Durability simulation, which uses digital twin technology to simulate the hydration process and microstructure evolution of concrete and predict its long-term durability indicators. S55: Multi-objective optimization, under multiple constraints such as performance, cost, and environmental protection, uses optimization algorithms to automatically search and generate the set of optimal mix proportion schemes; S56: Scheme decision ranking. The optimized schemes are weighted and fused according to preset weights to generate priority indexes and rank them to provide support for decision-making. S57: Outputs and Applications. Outputs optimized mix design schemes, control instructions, and procurement strategies to guide production and operations.

[0019] The beneficial effects of this invention are: 1. Compared with existing technologies that mainly rely on laboratory testing and engineers' experience, which have the disadvantages of long cycle, high cost and difficulty in dealing with raw material fluctuations, this solution adopts a performance prediction model based on machine learning and multi-source real-time industrial data fusion analysis, which can quickly and accurately predict concrete performance, significantly improve the efficiency and scientificity of mix design, and realize a fundamental transformation from experience-driven to data-driven. 2. Compared with existing technologies that typically optimize only for strength or cost, making it difficult to balance multiple performance and economic benefits, this solution adopts a multi-objective optimization engine that simultaneously weighs multiple indicators such as cost, strength, workability, environmental protection and long-term durability, and automatically generates a Pareto optimal solution set, providing decision-makers with a comprehensive and balanced optimal solution selection. 3. Compared with the shortcomings of existing technologies, which lack scientific evaluation methods for the long-term performance of concrete and often lead to insufficient structural durability, this solution adopts digital twin simulation technology to simulate the cement hydration process and microstructure evolution through physical mechanism models, accurately predict the long-term durability index of concrete, and provide full life-cycle performance assurance for important projects. 4. Compared with the shortcomings of existing technologies where the formula design and production control are disconnected and cannot respond to production fluctuations in real time, this solution builds an intelligent production system with real-time perception, dynamic prediction and closed-loop control by deploying multi-dimensional detection sensors and integrating production process data. This enables adaptive adjustment of the mixing ratio during the production process and greatly improves the stability of product quality. 5. Compared with existing technologies that only focus on production technology optimization and fail to coordinate with the supply chain, this solution innovatively integrates macro supply chain data, incorporating raw material price trends, inventory and logistics information into the optimization objectives. This allows the blending ratio optimization results to have cost foresight, achieving deep synergy between technology optimization and business decision-making, and creating greater overall economic benefits. Attached Figure Description

[0020] Fig. 1 The diagram shown is a structural schematic of the concrete mix proportion optimization system based on industrial data analysis of the present invention. Fig. 2 The diagram shown is a flowchart of the concrete mix proportion optimization method based on industrial data analysis of the present invention. Detailed Implementation

[0021] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0022] Please see Figs. 1-2 The present invention provides an embodiment of a concrete mix proportion optimization system based on industrial data analysis, comprising: The data acquisition and sensing module is used to acquire multi-source industrial data on concrete production. The data storage and processing module is used to clean, fuse, and feature-engineer multi-source industrial data collected by the data acquisition and sensing module. The core analysis module is used to predict concrete performance through machine learning models and to use a multi-objective optimization engine to weigh cost, performance and environmental indicators, and finally output the optimal mix proportion scheme. The application output module is used to output the optimized mix design and control instructions.

[0023] In this embodiment, the present invention constructs an intelligent system that integrates data acquisition, processing, analysis and application, thereby realizing the multi-dimensional industrial data fusion analysis of concrete mix proportions. By using machine learning and multi-objective optimization technology to replace the traditional experience-based trial mix design, the scientificity, efficiency and economic benefits of mix proportion design are significantly improved. At the same time, it provides core technical support for the refined management and sustainable development of concrete production.

[0024] As a preferred option, the data acquisition and sensing module specifically includes: A11: Near-infrared spectrometer for real-time monitoring of the water content of aggregates; A12: Laser particle size analyzer for detecting the fineness of powdered raw materials; A13: Density meter for detecting the concentration and solid content of water-reducing agents; A14: Machine vision system for real-time analysis of the particle size distribution and particle shape of aggregates through image processing and deep learning algorithms; A15: pH sensor for monitoring the pH of the water-reducing agent solution; A16: Temperature sensor array for monitoring the temperature of raw materials; A17: Online viscometer for real-time monitoring of the rheological properties of the mixture.

[0025] The data acquisition and perception module also accesses macro data of the external supply chain system through a data interface, including futures market prices, inventory levels, and logistics information of raw materials.

[0026] In this embodiment, the present application realizes dynamic perception of key raw material characteristics such as aggregate water content and powder fineness by deploying detection sensors such as near-infrared spectrometers and laser particle size analyzers, enabling the performance prediction model to reason based on real-time data rather than static empirical values, greatly improving the adaptability and prediction accuracy of mix proportion optimization in the face of raw material fluctuations, and ensuring the stability of concrete production quality from the source.

[0027] As a preferred, the core analysis module specifically includes: A21: Machine learning-based performance prediction model for quickly and accurately predicting key performance indicators such as the strength and workability of concrete under a specified mix proportion by training the model with historical data; A22: Multi-objective optimization engine for balancing and searching under multiple objectives and constraints such as performance, cost, and environmental protection, automatically generating an optimal mix proportion scheme set; A23: Digital twin simulation unit for simulating the hydration process and microstructure evolution of concrete based on physical mechanism models, predicting its long-term durability and service performance.

[0028] In the present embodiment, the present application dynamically monitors multi-dimensional characteristic data of aggregate moisture content and particle shape, powder fineness, water-reducing agent concentration and pH value, raw material temperature, and the rheological properties of the mixture, etc. by deploying a comprehensive sensor network composed of near-infrared spectrometer, laser particle size analyzer, densitometer, machine vision system, pH sensor, temperature sensor array, and online viscosity meter, etc. and synchronously accesses the market price and inventory information of the supply chain system; thereby constructing a high-dimensional real-time industrial database, driving data fusion analysis of the machine learning model and multi-objective optimization engine, and finally realizing adaptive dynamic optimization of the concrete mix proportion in the production process and supply chain collaborative decision-making, so as to comprehensively improve the quality stability, cost-effectiveness, and adaptability to complex working conditions of concrete production.

[0029] Preferably, the data storage and processing module, when in operation, specifically comprises: S11: receiving the raw data stream from the data acquisition and perception module, wherein the raw data stream includes real-time raw material characteristic data from online monitoring sensors, process time series data from production equipment, mix proportion and performance data from the laboratory, and macro supply chain data accessed through a data interface; S12: data cleaning of the raw data stream, including handling missing values, removing abnormal values due to sensor failure or transmission interference, and uniting and standardizing the format of data from different sources; S13: spatio-temporal alignment and fusion of the cleaned multi-source data, the spatio-temporal alignment including applying a uniform timestamp to all data and associating based on the formula number and production batch number to construct complete data records for analysis; S14: performing feature engineering operations on the fused data, including: A. calculating derived features, including water-binder ratio, total amount of cementitious materials, and sand ratio; B. extracting statistical features from process time series data, including average, variance of stirring current, and integral value of specific time interval; C. preprocessing macro supply chain data to generate features reflecting future trends or inventory costs of raw material prices; S15: outputting the high-quality data set processed by feature engineering to the core analysis module for training and reasoning based on the performance prediction model of the machine learning.

[0030] In the embodiment, the application solves the fusion problem of multi-source industrial data caused by heterogeneity, missing, and asynchrony through systematic data cleaning, spatio-temporal alignment, and feature engineering process. The derived water-binder ratio, stirring current statistical features, and supply chain trend features provide high-value and interpretable inputs for machine learning models, lay a data foundation for high-precision prediction and optimization, and improve the reliability and automation level of the system.

[0031] As preferred, the performance prediction model based on machine learning, when working, specifically includes: S21: receiving a data set output from the data storage and processing module, the data set containing historical mix proportion data, corresponding real-time raw material characteristic data, process data features, and macro features; S22: dividing the data set into a training set and a test set according to the time stamp, and training the preset machine learning algorithm using the training set to fit the complex nonlinear mapping relationship between the mix proportion features and the concrete performance indicators, wherein when the training set is used to train the preset machine learning algorithm, the objective function is: ; wherein, is the optimal model parameter, is the model parameter, is the experience loss term, represents the prediction result of the model M under the given parameter for the i-th input feature vector , is the regularization term, is the regularization function, is the regularization coefficient; S23: deploying the trained model in the production environment, receiving new mix proportion schemes and corresponding real-time sensor data, and generating prediction values of the key performance indicators of concrete; S24: outputting the prediction values to the multi-objective optimization engine as the core basis for scheme optimization and evaluation.

[0032] wherein the performance prediction model based on machine learning is trained using gradient boosting tree algorithm or neural network algorithm, the input features include water-binder ratio, cementitious material composition, real-time aggregate moisture content, real-time powder fineness, and environmental temperature and humidity, and the output labels include the compressive strength and slump of concrete.

[0033] In the embodiment, the application trains the machine learning model by defining an objective function comprising an experience loss term and a regularization term, ensures that the model learns the complex nonlinear mapping relationship between the efficient learning matching ratio and performance while having good generalization ability, effectively avoids overfitting, and makes the model maintain stable and reliable prediction performance when facing unknown new formulations, thereby providing an accurate evaluation basis for the optimization engine.

[0034] The multi-objective optimization engine adopts a genetic algorithm or a particle swarm algorithm to search for a Pareto optimal solution set under the conditions of meeting the national standard specification constraints and raw material availability constraints.

[0035] Preferably, the multi-objective optimization engine comprises the following steps when working: S31: receiving performance prediction values of a set of candidate matching ratio schemes from the machine learning-based performance prediction model and long-term durability index prediction values from the digital twin simulation unit; S32: defining a plurality of optimization objective functions, wherein the defined optimization objective functions comprise a cost objective function, a performance objective function, and an environmental protection objective function; S33: defining a set of constraint conditions, wherein the defined set of constraint conditions comprises performance constraints, durability constraints, and formulation constraints; S34: using a multi-objective optimization algorithm to perform iterative search in the formulation solution space based on the objective functions and constraint conditions, to generate a set of Pareto optimal solutions, wherein each solution in the set of Pareto optimal solutions represents an optimal trade-off scheme between multiple objectives; S35: using a weighted fusion method to fuse the cost and performance, to obtain a priority index of each solution in the set of optimal solutions, and to sort and output the set of Pareto optimal solutions according to the calculated priority index.

[0036] In the embodiment, the application automatically searches for a set of Pareto optimal solutions by using a multi-objective optimization algorithm, and innovatively uses a weighted fusion algorithm to calculate a priority index for sorting, thereby converting subjective decision preferences into objective quantitative indicators, providing clear and scientific decision support for quickly recommending a comprehensive optimal solution from a plurality of optimal trade-off schemes, simplifying the decision-making process, and improving the system practicality.

[0037] Preferably, when defining a plurality of optimization objective functions, the defined optimization objective functions comprise: A31: a cost objective function: ; wherein, is the price of the jth raw material, Let M be the unit usage of the j-th raw material, and M be the quantity of different types of raw materials. A32: Performance objective function: ; in, This is the predicted 28-day compressive strength value of scheme x by the performance prediction model. A negative value indicates the pursuit of maximizing strength. A33: Environmental protection objective function: ; in, Let be the unit carbon emission factor of the j-th raw material.

[0038] In this embodiment, the present invention precisely defines three objective functions: cost, performance, and environmental protection. It incorporates cost control, strength requirements, and low-carbon sustainability goals into a quantitative optimization framework, enabling the system's output mix design to not only meet basic performance requirements but also proactively pursue maximum economic benefits and minimum environmental impact, aligning with the development trends of green building and intelligent manufacturing.

[0039] As a preferred approach, when using a weighted fusion method to combine cost and performance to obtain the priority index of each solution in the optimal solution set, the principle formula is as follows: ; in, These are predefined weighting coefficients for the cost objective function. For the predefined objective function of the performance objective function, It represents the minimum value of the cost objective function among all solutions in the Pareto optimal solution set. It represents the maximum value of the cost objective function among all solutions in the Pareto optimal solution set. It represents the minimum cost-performance function value among all solutions in the Pareto optimal solution set. It represents the maximum value of the cost-performance function among all solutions in the Pareto optimal solution set. To adopt the i-th optimal mix proportion scheme At that time, the estimated total cost of raw materials required to produce one unit of concrete, For the i-th optimal mix design Performance values.

[0040] In this embodiment, the present invention cleverly solves the problem of inconsistent dimensions and inability to directly compare multiple objectives by calculating the priority index through normalization processing and weighted fusion formula. It transforms the user's preference for cost and performance into a single calculable index, realizes the scientific ranking of Pareto solution set, and makes the system output result both mathematically optimal and engineering practical.

[0041] As preferred, the digital twin simulation unit, when in operation, specifically comprises: S41: receiving standardized mix proportion data from the data storage and processing module, including water-binder ratio, chemical composition and proportion of each binder, aggregate characteristics and initial curing conditions; S42: calling and initializing a micro-model based on physical mechanism, setting the initial boundary conditions and material parameters of the model according to the input mix proportion data; S43: iteratively performing simulation calculation of the micro-model within a set time step, simulating the hydration reaction kinetics of cement particles, the evolution process of micro-pore structure, and the transport and diffusion of water and ions in the pore network; S44: based on the simulation results, calculating and outputting the long-term performance and durability index prediction values; S45: delivering the long-term performance and durability index prediction values to the multi-objective optimization engine as the durability constraint condition or optimization target when performing multi-objective optimization.

[0042] In the present embodiment, the present application simulates the hydration process and microstructure evolution by calling the physical mechanism model through the digital twin simulation unit, realizes the accurate prediction of the long-term durability of concrete, breaks through the optimization limitation of traditional methods which only rely on short-term strength, provides a forward-looking scientific basis for the design of concrete mix proportion of major projects, and guarantees the long service life and high durability of the structure.

[0043] The method for analyzing concrete mix proportion optimization based on industrial data comprises the following steps: S51: data acquisition and perception, real-time acquisition of multi-source industrial data such as raw material characteristics, production process, laboratory data and supply chain information through sensors and interfaces; S52: data preprocessing and fusion, cleaning, standardizing, spatio-temporal aligning and feature engineering of multi-source data, and construction of high-quality data set for analysis; S53: performance prediction, using machine learning model to quickly predict key performance indicators of concrete under specified mix proportion based on processed data; S54: durability simulation, simulating the hydration process and microstructure evolution of concrete through digital twin technology to predict its long-term durability index; S55: multi-objective optimization, under multiple constraints such as performance, cost, environmental protection, etc., using optimization algorithm to automatically search and generate optimal mix proportion scheme set; S56: scheme decision ranking, according to the preset weight, the optimization scheme is weighted and fused to calculate the priority index and sort, providing support for decision-making; S57: output and application, outputting the optimized mix proportion scheme, control instruction and procurement strategy to guide production and operation.

[0044] In this embodiment, the present application realizes the full-process automation and intelligence from data perception to decision output through the method steps, deeply integrates industrial data analysis technology into concrete mix proportion design and production management, significantly improves design efficiency, reduces trial and error cost, guarantees material performance, and finally realizes the deep integration of technology and production and operation through output control instructions and procurement strategies.

[0045] Embodiment one: daily production optimization of commercial concrete mixing station Scenario: A large commercial concrete mixing station, daily production of C20 to C50 various strength grade ordinary commercial concrete, supply to multiple residential and commercial building projects. The station is facing the challenges of large raw material fluctuation, significant cost control pressure, and different project requirements for concrete workability.

[0046] Application process: The mixing station deploys the system described in the present application. In the data acquisition and perception layer, near-infrared moisture meters are installed above the sand and gravel aggregate conveying belt, and laser particle size analyzers are installed on the conveying pipelines of fly ash and cement warehouses to monitor the water content and fineness of each batch of raw materials in real time. At the same time, the control system of the mixing station opens the data interface, and the system can collect the main motor current curve and mixing time of the mixer in real time. In addition, the system also accesses the enterprise's ERP system through API to obtain real-time inventory and the latest purchase price of raw materials.

[0047] In the data storage and processing layer, the system cleans, aligns and fuses real-time data from sensors, production process data from PLC, historical sample mix proportion and strength data from the laboratory, and inventory cost data from ERP. Feature engineering calculates the actual water-binder ratio of each batch of concrete (dynamically adjusts water consumption according to real-time water content), total amount of cementitious materials, and extracts features such as average torque and current stability from the mixing current curve to build a high-quality data set.

[0048] In the core analysis layer, the system uses historical data to train an XGBoost machine learning model that can accurately predict slump and 28-day compressive strength based on dynamically changing mix proportions and real-time raw material characteristics. Subsequently, a multi-objective optimization engine is triggered, with the objective function set as: a, minimum cost; b, strength meets requirements with certain surplus; c, use as much fly ash as possible to reduce carbon emissions. The constraint conditions include slump range, maximum water-binder ratio, etc. The optimization engine uses genetic algorithm to quickly search and generate dozens of Pareto optimal formulations. At the same time, the digital twin simulation unit will quickly microscopically simulate these candidate formulations to ensure that their long-term carbonation resistance performance meets the basic requirements.

[0049] Finally, at the application output level, the system sorts the optimal solution set according to the current cost weight, pushes the highest priority mix ratio to the control room, and automatically issues it to the automated mixing station control system. After the operator confirms, the system starts production. At the same time, the system also generates a procurement suggestion, suggesting that the current fly ash has a high cost performance and that the production volume of this batch should be increased.

[0050] Beneficial effects: The system successfully realizes the dynamic real-time optimization of the mix ratio design of the mixing station. The cost of each cubic meter of concrete is reduced by about yuan on average, and the phenomenon of returning goods due to substandard work is basically eliminated. At the same time, through the optimization of the use of industrial waste, the environmental load is reduced, achieving a win-win of economic benefits and environmental benefits.

[0051] Example Two: High-performance concrete design for major infrastructure projects Scenario: A certain cross-sea bridge engineering project requires the preparation of C60 high-strength marine concrete, which must have extremely high resistance to chloride ion penetration, low heat to reduce the risk of cracking, and a design life of 100 years in harsh marine corrosion environment. The traditional trial method has a long cycle, high cost, and is difficult to accurately quantify long-term durability.

[0052] Application process: The project command center introduces the system of the application to conduct scientific research and design of concrete mix ratio. In the data collection and perception layer, in addition to deploying conventional real-time sensors, the laboratory also provides detailed chemical composition analysis reports of all cementitious materials such as cement, silica fume, and mineral powder (such as C3A content, SiO2 content, etc.). These high-dimensional data are also sent to the system.

[0053] In the data storage and processing stage, the system integrates the chemical properties of all raw materials, information about the planned use of various special admixtures, and detailed performance data of hundreds of small sample trial mixes conducted by the laboratory in the early stage (including durability indicators such as electric flux and chloride ion diffusion coefficient), constructing an extremely detailed and high-quality data set.

[0054] In the core analysis stage, the system's powerful capabilities are fully demonstrated. First, the machine learning performance prediction model can accurately predict the 7-day and 28-day strength and initial slump of concrete under different mix ratios based on extensive experimental data. Subsequently, the digital twin simulation unit becomes the core role. It calls the DuCOM model based on physical mechanisms to conduct in-depth simulation of the candidate mix ratios generated by the optimization engine. The simulation process accurately calculates the hydration reaction heat and hydration process of cement and admixtures, and simulates the evolution process of the micro-pore structure. Finally, the simulation unit outputs the key long-term performance indicators: the chloride ion penetration depth after 100 years and the dry shrinkage value at 180 days of age.

[0055] The multi-objective optimization engine iteratively searches with the hard constraints of "100-year chloride penetration depth no more than 5mm" and "180-day shrinkage value no more than 400 micro-strain", and the optimization objectives of "lowest cost" and "lowest hydration heat". Due to the huge variable space (involving multiple special materials), the optimization process is complex, but the system successfully generates a set of Pareto optimal solutions.

[0056] When the application is output, the system provides three optimal schemes for the engineer. The first scheme has the lowest cost and meets the requirement of hydration heat; the second scheme has the lowest hydration heat but higher cost; and the third scheme achieves the best balance between cost and thermal performance. The engineer combines the construction environment temperature and finally selects the third scheme. The system then outputs the accurate mixing ratio and detailed long-term performance prediction report of the scheme, providing a crucial scientific basis for the design and construction of the bridge.

[0057] Beneficial effects: The system provides a scientific and reliable high-performance concrete solution for the cross-sea bridge project, shortening the original several months of mixing ratio research and development period to several weeks and greatly reducing the research and development cost. More importantly, the durability of the concrete is quantitatively predicted through digital twin technology, ensuring the long-term safety and durability of this major infrastructure project, which is fundamentally impossible with traditional test methods.

[0058] The embodiments of the present application are described in detail above in combination with the drawings, but the present application is not limited to the above-described embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the purpose of the present application.

Claims

1. A system for optimizing concrete mix based on industrial data analysis, characterized in that: The application comprises: a data acquisition and perception module for obtaining multi-source industrial data of concrete production; a data storage and processing module for cleaning, data fusion and feature engineering of the multi-source industrial data collected by the data acquisition and perception module; a core analysis module for predicting concrete performance through a machine learning model and weighing cost, performance and environmental protection indicators using a multi-objective optimization engine to finally output an optimal mix proportion scheme; an application output module for outputting the optimized mix proportion scheme and control instructions.

2. The system for optimizing concrete mix proportion based on industrial data analysis according to claim 1, characterized in that: The data acquisition and perception module specifically comprises: A11: a near-infrared spectrometer for real-time monitoring of the water content of aggregates; A12: a laser particle size analyzer for detecting the fineness of powdery raw materials; A13: a densitometer for detecting the concentration and solid content of water reducing agents; A14: a machine vision system for real-time analysis of aggregate particle gradation and particle shape through image processing and deep learning algorithms; A15: a pH sensor for monitoring the pH of water reducing agent solution; A16: an array of temperature sensors for monitoring raw material temperature; A17: an online viscometer for real-time monitoring of the rheological properties of the mixture.

3. The system for optimizing concrete mix proportion based on industrial data analysis according to claim 2, characterized in that: The core analysis module specifically comprises: A21: a machine learning-based performance prediction model for training the model using historical data to quickly and accurately predict key performance indicators such as strength and workability of concrete under a specified mix proportion; A22: a multi-objective optimization engine for weighing and searching under multiple objectives and constraints such as performance, cost, and environmental protection to automatically generate an optimal mix proportion scheme set; A23: a digital twin simulation unit for simulating the hydration process and microstructure evolution of concrete based on physical mechanism models to predict its long-term durability and service performance.

4. The system for optimizing concrete mix proportion based on industrial data analysis according to claim 3, characterized in that: The data storage and processing module, when in operation, specifically comprises: S11: receiving raw data streams from the data acquisition and perception module, wherein the raw data streams include real-time raw material property data from online monitoring sensors, process time series data from production equipment, mix proportion and performance data from the laboratory, and macro supply chain data accessed through a data interface; S12: cleaning the raw data stream, including handling missing values, removing abnormal values due to sensor failure or transmission interference, and unitizing and formatting different sources of data; S13: aligning and fusing the cleaned multi-source data in space and time, the space-time alignment includes applying a uniform timestamp to all data and associating based on formula number and production batch number to construct complete data records for analysis; S14: performing feature engineering operations on the fused data, including: A. calculating derived features including water-binder ratio, total amount of cementitious materials and sand ratio; B. extracting statistical features from process time series data, including average, variance and integral value of stirring current in a specific time interval; C. preprocessing macro supply chain data to generate features reflecting future trends or inventory costs of raw material prices; S15: Output the high-quality data set after feature engineering to the core analysis module for training and inference based on the machine learning-based performance prediction model.

5. The system for optimizing concrete mix proportion based on industrial data analysis according to claim 4, characterized in that: The machine learning-based performance prediction model in operation specifically includes: S21: Receive the data set output from the data storage and processing module, which contains historical mix proportion data, corresponding real-time raw material characteristic data, process data features, and macro features; S22: Divide the data set into a training set and a test set according to the timestamp, and use the training set to train the pre-set machine learning algorithm to fit the complex nonlinear mapping relationship between the mix proportion features and the concrete performance indicators, wherein the objective function is: ； in, For optimal model parameters, For model parameters, For experience loss items, This indicates that model M, given parameters Next, for the i-th input feature vector The prediction results For regularization terms, For regularization functions, The regularization coefficient is used. S23: Deploy the trained model in the production environment, receive new mix proportion schemes and corresponding real-time sensor data, and generate predicted values of key concrete performance indicators; S24: Output the predicted values to the multi-objective optimization engine as the core basis for scheme optimization and evaluation.

6. The system for optimizing concrete mix proportion based on industrial data analysis according to claim 5, characterized in that: The multi-objective optimization engine in operation specifically includes: S31: Receive performance prediction values for a set of candidate mix proportion schemes from the machine learning-based performance prediction model, and receive long-term durability indicator prediction values from the digital twin simulation unit; S32: Define multiple optimization objective functions, wherein the defined optimization objective functions specifically include cost objective functions, performance objective functions, and environmental protection objective functions; S33: Define a set of constraint conditions, wherein the defined set of constraint conditions includes performance constraints, durability constraints, and formula constraints; S34: Use a multi-objective optimization algorithm to perform iterative search in the formula solution space based on the objective functions and constraint conditions, generating a set of Pareto optimal solutions, wherein each solution in the Pareto optimal solution set represents an optimal trade-off scheme between multiple objectives; S35: Use a weighted fusion method to fuse the cost and performance to obtain a priority index for each solution in the optimal solution set, and sort and output the Pareto optimal solution set according to the calculated priority index.

7. The system for optimizing concrete mix proportion based on industrial data analysis according to claim 6, characterized in that: When defining multiple optimization objective functions, the defined optimization objective functions specifically include: A31: Cost objective function: ； wherein, Pj is the price of the jth raw material, M is the number of raw material types; A32: Performance objective function: ； wherein, is the 28-day compressive strength prediction value of the performance prediction model for the scheme x, and a negative value indicates a pursuit of maximum strength; A33: Environmental protection objective function: ； wherein, Cj is the unit carbon emission factor for the jth raw material.

8. The system for optimizing concrete mix proportion based on industrial data analysis according to claim 7, characterized in that: When using a weighted fusion method to fuse the cost and performance to obtain a priority index for each solution in the optimal solution set, the principle formula is: ； wherein, is a predefined weight coefficient of the cost objective function, is a predefined objective function of the performance objective function, is the minimum value of the cost objective function value in all solutions in the Pareto optimal solution set, is the maximum value of the cost objective function value in all solutions in the Pareto optimal solution set, is the minimum value of the cost performance function value in all solutions in the Pareto optimal solution set, is the maximum value of the cost performance function value in all solutions in the Pareto optimal solution set, is the estimated total cost of raw materials needed for producing one unit of concrete when the i-th optimal mix proportion scheme is adopted, is the performance value of the i-th optimal mix proportion scheme .

9. The system for optimizing concrete mix proportion based on industrial data analysis according to claim 8, characterized in that: The digital twin simulation unit in operation specifically includes: S41: Receive standardized mix proportion data from the data storage and processing module, including water-binder ratio, chemical composition and proportion of each binder material, aggregate characteristics, and initial curing conditions; S42: Call and initialize the micro-model based on physical mechanism, set the initial boundary conditions and material parameters of the model according to the input mix proportion data; S43: Iteratively execute simulation calculation of the micro-model within a set time step, simulate the hydration reaction kinetics of cement particles, the evolution process of micro-pore structure, and the transport and diffusion of water and ions in the pore network; S44: Based on the simulation results, calculate and output the long-term performance and durability index prediction value; S45: Deliver the long-term performance and durability index prediction value to the multi-objective optimization engine as its durability constraint condition or optimization target when performing multi-objective optimization.

10. A method for optimizing concrete mix proportion based on industrial data analysis, characterized in that: Comprising the following steps: S51: Data acquisition and perception, real-time acquisition of raw material characteristics, production process, laboratory data and supply chain information and other multi-source industrial data through sensors and interfaces; S52: Data preprocessing and fusion, cleaning, standardization, spatio-temporal alignment and feature engineering of multi-source data, and construction of high-quality data set for analysis; S53: Performance prediction, using machine learning model to quickly predict key performance indicators of concrete under specified mix proportion based on processed data; S54: Durability simulation, simulating the hydration process and microstructure evolution of concrete through digital twin technology to predict its long-term durability index; S55: Multi-objective optimization, under the multiple constraints of performance, cost, environmental protection, etc., using optimization algorithm to automatically search and generate the optimal mix proportion scheme set; S56: Scheme decision ranking, according to the preset weight, the optimization scheme is weighted and fused to calculate the priority index and sort, providing support for decision-making; S57: Output and application, output the optimized mix proportion scheme, control instruction and procurement strategy to guide production and operation.

Images