A full-automatic concrete batching optimization method based on multi-sensor fusion

Abstract

The application discloses a kind of full-automatic concrete batching optimization methods based on multi-sensor fusion, it is related to concrete production intelligent control technical field.The present application includes the following steps: step S1, system building, data acquisition and initialization;Step S2, ingredient influencing factor dimension reduction and key factor screening based on improved principal component analysis;Step S3, material effective utilization rate dynamic prediction based on fractional order grey-multi-sensor fusion BP neural network combination model;Step S4, batching scheme global optimization based on multi-constraint multi-objective dynamic programming;Step S5, multi-sensor data real-time fusion and batching parameter dynamic correction;Step S6, multi-system collaborative automation control and closed-loop iterative optimization.The present application is coupled by principal component analysis, fractional order grey-multi-sensor fusion BP neural network, multi-constraint multi-objective dynamic programming and multi-sensor fusion technology in depth, constructs full-process full-automatic batching optimization system, reduces artificial cost.

Description

Technical Field

[0001] This invention belongs to the field of intelligent control technology for concrete production, and in particular relates to a solution that addresses the problems of error-prone multi-sensor fusion, independent control of each subsystem, low batching accuracy, large fluctuations in material utilization, and high labor costs, thereby achieving unmanned, intelligent, and precise control of the entire concrete batching process. Background Technology

[0002] Currently, concrete mixing and batching systems work by having operators input the required concrete volume into a mix proportion formula to calculate the amount of materials needed for the concrete volume. The operators then input these material quantities into a batching device, which distributes the materials to various weighing hoppers. Data is read and written to a programmable logic controller (PLC) via a touchscreen, and the PLC controls the weighing hoppers, mixer, belt conveyor, and distributor. Existing technology suffers from the following core shortcomings: 1. The mixing and batching process requires many actions from the operator. Manual input is prone to data entry errors, which can lead to quality accidents such as scrapped concrete and poor production quality stability. 2. The mixing system and the inclined feeding system are independently controlled, requiring manual material distribution by the operator. If the weighing of the feeding system malfunctions, the material distribution will be inaccurate, resulting in insufficient weight of the electric pole and affecting the quality of the finished product. 3. The silica fume weighing system and the mixing system are independently controlled. Operators need to manually pump the silica fume into the residual slurry storage tank, weigh it a second time, and then discharge it into the mixing tank for use. The operation process is cumbersome, the labor cost is high, and the second weighing is prone to introducing errors. 4. Existing batching schemes mostly use fixed proportion formulas for calculation, without considering the impact of raw material property fluctuations, environmental conditions changes, and differences in material hydration efficiency on the batching effect. This results in large fluctuations in material utilization, low batching accuracy, and high material loss.

[0003] To address the aforementioned shortcomings, existing technologies mostly employ single-sensor data fusion or simple neural network fitting for optimization. These methods fail to effectively reduce the dimensionality of high-dimensional influencing factors, lack sufficient accuracy in predicting material utilization, and fail to achieve global optimization and multi-system collaborative control throughout the entire batching process. Consequently, they cannot fundamentally solve the pain points of high reliance on manual labor, low batching accuracy, and independent system operation. Summary of the Invention

[0004] The purpose of this invention is to provide a fully automated concrete batching optimization method based on multi-sensor fusion. It deeply couples principal component analysis, fractional gray-multi-sensor fusion BP neural network, multi-constraint multi-objective dynamic programming and multi-sensor fusion technology to construct a fully automated batching optimization system with "key factor screening, dynamic prediction, global optimization, real-time correction, collaborative control and closed-loop iteration". This solves the problems of insufficient prediction accuracy of material utilization rate, high dependence on manual labor and low batching accuracy in existing concrete batching methods.

[0005] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: This invention relates to a fully automated concrete batching optimization method based on multi-sensor fusion, comprising the following steps: Step S1, System Setup, Data Acquisition and Initialization: Build a batching control system including a core control unit, a multi-sensor acquisition unit, an execution unit, and a data storage unit; collect historical production data to build a database and complete preprocessing; and initialize and calibrate the system and sensors. Step S2: Dimensionality reduction and key factor screening of batching influencing factors based on improved principal component analysis: Standardize the total influencing indicators of concrete batching, construct the correlation coefficient matrix and solve the eigenvalues ​​and eigenvectors, screen the core principal components through the cumulative contribution rate, and determine the key factors affecting the effective utilization rate of materials and batching effect; Step S3: Dynamic prediction of material utilization rate based on fractional gray-multi-sensor fusion BP neural network combined model: Using the real-time data of multi-sensor corresponding to the key factors selected in step S2 as input, the basic predicted value and relative error sequence of material utilization rate are obtained through fractional gray model, and the relative error is predicted and compensated through BP neural network, and finally the dynamic predicted value of the utilization rate of each material is output. Step S4: Global optimization of batching scheme based on multi-constraint and multi-objective dynamic programming: Based on the material effective utilization rate predicted in step S3, the theoretical input of materials is calculated back, and a multi-constraint and multi-objective dynamic programming model is constructed with the objectives of minimizing batching error, minimizing material loss, optimizing production cost, achieving concrete strength standards, and balancing equipment operating load. The global optimal batching scheme and the whole process control parameters are solved. Step S5: Real-time fusion of multi-sensor data and dynamic correction of batching parameters: The real-time data collected by multiple sensors is denoised and preprocessed by Kalman filtering. An adaptive weighted fusion algorithm is used to obtain the comprehensive measurement value. Combined with the optimal batching scheme and the predicted value of material effective utilization rate, the material input and equipment operating parameters in the batching process are dynamically corrected in real time. Step S6, Multi-system collaborative automated control and closed-loop iterative optimization: Based on the optimal control parameters and real-time correction results, fully automatic collaborative control of the mixing system, tilting feeding system and silica fume weighing system is realized. After each batching is completed, the historical database is updated, and the model is iteratively optimized to form a closed-loop control of the entire process.

[0006] As a preferred technical solution, in step S1, the multi-sensor acquisition unit includes material property sensors, working environment sensors, and metering and equipment sensors; wherein, the material property sensors include sand and gravel moisture content sensors, cement fineness sensors, silica fume specific surface area sensors, and admixture solid content detectors; the working environment sensors include mixing tank temperature and humidity sensors, ambient temperature and humidity sensors, online concrete slump detectors, and mixing host current sensors; the metering and equipment sensors include high-precision weighing sensors for each weighing hopper, infrared displacement sensors for the feeding system, pressure sensors for the silica fume storage tank, equipment vibration sensors, valve response time detectors, and belt conveyor speed encoders; all sensors have a sampling frequency of not less than 10Hz and communicate bidirectionally with the core control unit in real time.

[0007] As a preferred technical solution, the specific implementation process of the improved principal component analysis in step S2 is as follows: Step S21: Standardize the total impact indicators. Assume there are m total impact indicators for concrete mix design, and N sets of historical samples. The first sample Each indicator takes a value of The standardized formula is: ; In the formula, The standardized index value, For the first The sample mean of each indicator, For the first The sample standard deviation of each indicator; Step S22: Construct the correlation coefficient matrix of the standardized indicators The formula is: ; In the formula, Let be the correlation coefficient between the i-th and j-th concrete mix design indices. This represents the total number of historical production samples. Step S23: Solve for the correlation coefficient matrix eigenvalues and the corresponding feature vectors, and construct from the feature vectors For each principal component, calculate the information contribution rate of each principal component. With cumulative contribution rate The calculation formula is: ; When the cumulative contribution rate When selecting the previous option Each principal component is used as the core principal component, and the absolute value of the loading in each principal component is selected. The indicators are used as key influencing factors.

[0008] As a preferred technical solution, in step S3, the effective utilization rate of materials is defined as: the ratio of the mass of material that actually participates in the hydration reaction and effectively contributes to the strength and workability of the hardened concrete to the total mass of the same material added during the batching process, specifically by the formula: ; In the formula, For the first Effective utilization rate of various materials For the first The effective quality of the material For the first The total mass of the input materials.

[0009] As a preferred technical solution, the specific implementation process of the fractional-order gray-multi-sensor fusion BP neural network combination model in step S3 is as follows: Step S31, Fractional Grey Model Preprocessing: Let the non-negative original sequence of the effective utilization rate of a certain material be... The accumulated sequence is obtained by using an r-order accumulation operator, as shown in the formula: ; In the formula, For fractional order, For combinations, construct The cumulative discrete grey model is used to estimate the model parameters using the least squares method, and the basic predicted value of the effective utilization rate is obtained. And calculate the relative error sequence. The formula is: ; Step S32: Construct a multi-sensor fusion BP neural network: The network has a 3-layer structure, and the number of neurons in the input layer is the same as the number of core principal components selected in step S2. The number of neurons in the output layer is equal to the number of core material types. The formula for optimizing the number of neurons in the hidden layer is: In the formula, This represents the number of neurons in the hidden layer. The range of values ​​for the adjustment constant is as follows: Using real-time multi-sensor data of key factors as input and the relative error sequence of a fractional-order grey model as the training target, the relative error compensation value is predicted after network training is completed. ; Step S33: Output of the final predicted value of the combined model: In the formula, For the first The final predicted value of the effective utilization rate of the material.

[0010] As a preferred technical solution, the specific implementation process of the multi-constraint, multi-objective dynamic programming model in step S4 is as follows: Step S41: Calculate the theoretical material input based on the predicted material utilization rate. The specific formula is as follows: ; In the formula, For the first Theoretical input quality of the material For the first The target effective quality of this material The standard density of concrete, For the total volume of concrete required, For the target strength grade of concrete, the first Standard mass percentage of each type of material; Step S42: Construct the overall objective function for multi-objective optimization: In the formula, To optimize the overall objective value for multiple objectives, The weight coefficients for each sub-objective satisfy the following conditions: ; To account for the overall batching error rate, To account for the overall material loss rate, For the overall cost of ingredients, This is the concrete strength deviation coefficient. The overall load rate of the equipment; You also need to set constraints, as follows: ; In the formula, For the first Optimize the input quality of various materials. , For material usage limits, To optimize the concrete compressive strength corresponding to the mix proportions, The target compressive strength of concrete. The slump of the concrete. The linear speed of the equipment; Step S43: Solve the dynamic programming model using the reverse order solution method to obtain the global optimal batching scheme and the whole process control parameters, including the optimal input amount of each material, belt conveyor speed, mixing speed, distributor opening, feeding speed, and silica fume addition amount.

[0011] As a preferred technical solution, the specific process of solving the dynamic programming model using the reverse solution method in step S43 is as follows: Step S431: Starting from the last stage, recursively solve for the optimal value function of each stage. and optimal decision : Step S432: In Stage given state In the feasible decision set Select ,calculate Take the minimum value and the corresponding optimal decision ; Step S433: In Stage given state In the feasible decision set Select The state transition equation is obtained as follows: ,calculate Take the minimum value and the corresponding optimal decision ; Step S434: Repeat the above process sequentially until... To obtain the initial state The optimal value function and optimal decision ; Step S435: Output the optimal solution sequentially: from Optimal decision Starting from this point, we obtain the state transition equation. Check again The optimal decision table is obtained Output the optimal decision for each stage in sequence. That is, the globally optimal ingredient ratio.

[0012] As a preferred technical solution, the specific process of real-time fusion of multi-sensor data and dynamic correction of ingredient parameters in step S5 is as follows: Step S51, Kalman filtering noise reduction preprocessing: Based on the optimal system state estimate and system control quantity at time k-1, the system state prediction, error covariance prediction, and Kalman gain calculation at time k are completed sequentially. Combined with the original measured values ​​of the sensors at time k, the optimal system state estimate that fits the actual working conditions on site is obtained. The error covariance matrix is ​​updated synchronously to provide a basis for the next acquisition cycle. Filtering parameters that match the measurement accuracy and equipment characteristics are configured for different types of sensors such as weighing, temperature and humidity, and displacement. This process is executed cyclically with the sensor acquisition frequency, effectively eliminating noise interference caused by equipment vibration, temperature drift, and electromagnetic interference on the production site, and obtaining the effective measurement data corresponding to each sensor. Step S52, Adaptive Weighted Data Fusion: For the effective sensor measurement data of the same measurement target in the concrete batching process after Kalman filtering, adaptive weighted data fusion processing is performed. First, the measurement fluctuation of each sensor in the current and consecutive previous acquisition cycles is statistically analyzed to calculate the real-time measurement variance. Then, dynamic weight coefficients are assigned to the effective measurement data of each sensor based on the measurement variance. The smaller the measurement variance, the larger the weight coefficient, and the sum of all weight coefficients is 1. Finally, all effective measurement data are weighted and fused according to the dynamic weight coefficient to obtain the unique, accurate and stable comprehensive measurement value of the measurement target at the current moment. The dynamic weight coefficient is updated synchronously with the real-time measurement variance of the sensor to realize the adaptive adjustment of the fusion process. Step S53, Real-time Dynamic Correction of Batching Parameters: Based on the comprehensive material measurement value obtained by adaptive weighted fusion, combined with the predicted material effective utilization rate obtained in step S3 and the real-time environmental temperature and humidity data, the input amount of each material in the batching process is dynamically corrected in real time. First, the comprehensive measurement value is compared with the target material weighing value to obtain the real-time deviation. Then, the basic correction is completed through the material effective utilization rate, and the secondary correction is completed through the temperature and humidity comprehensive correction coefficient. The corrected material input mass at the current moment is calculated by combining the results of multiple corrections. Then, it is converted into the execution unit control parameters and sent to the PLC core control unit. The PLC adjusts the opening of the weighing hopper valve, the speed of the material conveying equipment, and other execution actions in real time. This correction process is continuously executed throughout the entire process of material weighing and conveying. When the weighing is close to the target value, the feeding speed is adjusted to avoid overfeeding, so as to realize the closed-loop dynamic correction of batching parameters and the precise control of batching accuracy.

[0013] As a preferred technical solution, the specific implementation process of multi-system collaborative automation control in step S6 is as follows: Step S61: The mixing system and the inclined feeding system are controlled in a coordinated manner, and the coordination formula is as follows: ; In the formula, for Constantly monitor the mixer speed. For the synergy coefficient, for The feed quality flow rate after constant mixing for The total mass of materials in the mixing tank is monitored at all times; simultaneously, the material distribution rate is corrected using an aerial material compensation formula, which is: ; In the formula, Let t be the actual material distribution mass. To achieve the target material quality, To compensate for the mass of materials falling from the air. The feeding delay time is used to monitor the feeding system's operating status in real time. When the deviation continuously exceeds the threshold, it is judged as a fault and an alarm and emergency shutdown are triggered. In step S62, the silica fume weighing and stirring system is under fully automatic coordinated control. The silica fume inventory is calculated in real time using a pressure sensor, and the formula is as follows: ; In the formula, for The total mass of silica fume in the silica fume storage tank at any given time. for The pressure measurement value at the bottom of the storage tank at all times. The cross-sectional area of ​​the bottom of the storage tank. The acceleration due to gravity is used; based on the optimal batching scheme, the silica fume conveying pump is controlled to achieve automatic extraction and one-time weighing. After weighing, it is directly discharged into the mixing tank, and the silica fume dosage is adjusted based on the ambient humidity.

[0014] As a preferred technical solution, the specific process of closed-loop iterative optimization in step S6 is as follows: after each batching is completed, the system automatically collects the actual material usage, measured concrete strength, slump, material loss, and equipment operation data of this production, calculates the actual effective utilization rate of each material, and updates it to the historical database; after every 50 production batches are completed, the PCA model, combined prediction model, and dynamic programming model are retrained and their parameters are optimized to achieve adaptive iteration of the model.

[0015] The present invention has the following beneficial effects: (1) This invention is the first to deeply couple principal component analysis, fractional gray-multi-sensor fusion BP neural network, multi-constraint multi-objective dynamic programming and multi-sensor fusion technology to construct a fully automatic batching optimization system with "key factor screening, dynamic prediction, global optimization, real-time correction, collaborative control and closed-loop iteration". It completely eliminates the manual operation link, fundamentally solves the problem of concrete scrapping caused by manual input errors, and reduces labor costs.

[0016] (2) This invention achieves effective dimensionality reduction of high-dimensional influencing factors of concrete batching by improving the PCA model, eliminating redundant interference factors, and improving the computational efficiency and accuracy of subsequent prediction models. In view of the characteristics of small sample size, non-stationarity and large fluctuation of concrete batching data, a fractional gray-multi-sensor fusion BP neural network combination model is constructed to realize dynamic and accurate prediction of material utilization rate, improve prediction accuracy, and provide accurate core basis for batching optimization.

[0017] (3) The present invention constructs a multi-constraint, multi-objective dynamic programming model, which covers five core objectives: batching accuracy, material loss, production cost, concrete strength, and equipment load. It achieves the global optimal batching scheme and reduces material loss and overall production cost compared with the traditional fixed mix ratio scheme.

[0018] (4) This invention realizes real-time dynamic correction of the batching process and reduces batching error through Kalman filtering-adaptive weighted fusion multi-sensor data fusion technology; at the same time, it realizes fully automatic collaborative control of the stirring system, the inclined feeding system and the silica fume weighing system, which solves the pain points of inaccurate material distribution, insufficient pole feeding weight and cumbersome silica fume operation caused by independent control of each system in the prior art, and improves material distribution accuracy.

[0019] (5) The present invention constructs a closed-loop iterative optimization mechanism for the entire process. The more production batches there are, the better the prediction accuracy and optimization effect of the model. It has strong adaptability and can be widely applied to various concrete batching scenarios such as concrete mixing plants and pole production. It has significant economic benefits and industry promotion value.

[0020] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is an overall flowchart of a fully automated concrete batching optimization method based on multi-sensor fusion according to the present invention. Figure 2 This is an overall architecture diagram of the batching control system described in this invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0025] To make the purpose, technical solution, and advantages of this application clearer, the following description is provided in conjunction with the appendix. Figure 1-2 The present application will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the application.

[0026] Please see Figure 1 As shown, this invention is a fully automated concrete batching optimization method based on multi-sensor fusion, comprising the following steps: Step S1, System Setup, Data Acquisition and Initialization: Build a batching control system including a core control unit, a multi-sensor acquisition unit, an execution unit, and a data storage unit; collect historical production data to build a database and complete preprocessing; and initialize and calibrate the system and sensors. Step S2: Dimensionality reduction and key factor screening of batching influencing factors based on improved principal component analysis: Standardize the total influencing indicators of concrete batching, construct the correlation coefficient matrix and solve the eigenvalues ​​and eigenvectors, screen the core principal components through the cumulative contribution rate, and determine the key factors affecting the effective utilization rate of materials and batching effect; Step S3: Dynamic prediction of material utilization rate based on fractional gray-multi-sensor fusion BP neural network combined model: Using the real-time data of multi-sensor corresponding to the key factors selected in step S2 as input, the basic predicted value and relative error sequence of material utilization rate are obtained through fractional gray model, and the relative error is predicted and compensated through BP neural network, and finally the dynamic predicted value of the utilization rate of each material is output. Step S4: Global optimization of batching scheme based on multi-constraint and multi-objective dynamic programming: Based on the material effective utilization rate predicted in step S3, the theoretical input of materials is calculated back, and a multi-constraint and multi-objective dynamic programming model is constructed with the objectives of minimizing batching error, minimizing material loss, optimizing production cost, achieving concrete strength standards, and balancing equipment operating load. The global optimal batching scheme and the whole process control parameters are solved. Step S5: Real-time fusion of multi-sensor data and dynamic correction of batching parameters: The real-time data collected by multiple sensors is denoised and preprocessed by Kalman filtering. An adaptive weighted fusion algorithm is used to obtain the comprehensive measurement value. Combined with the optimal batching scheme and the predicted value of material effective utilization rate, the material input and equipment operating parameters in the batching process are dynamically corrected in real time. Step S6, Multi-system collaborative automated control and closed-loop iterative optimization: Based on the optimal control parameters and real-time correction results, fully automatic collaborative control of the mixing system, tilting feeding system and silica fume weighing system is realized. After each batching is completed, the historical database is updated, and the model is iteratively optimized to form a closed-loop control of the entire process.

[0027] In step S1, the multi-sensor acquisition unit includes material property sensors, operating environment sensors, and metering and equipment sensors. The material property sensors include sand and gravel moisture content sensors, cement fineness sensors, silica fume specific surface area sensors, and admixture solid content detectors. The operating environment sensors include mixing tank temperature and humidity sensors, ambient temperature and humidity sensors, online concrete slump detectors, and mixing host current sensors. The metering and equipment sensors include high-precision weighing sensors for each weighing hopper, infrared displacement sensors for the feeding system, pressure sensors for the silica fume storage tank, equipment vibration sensors, valve response time detectors, and belt conveyor speed encoders. All sensors have a sampling frequency of no less than 10Hz and communicate bidirectionally with the core control unit in real time.

[0028] The specific implementation process for concrete batching in the production of C30 strength grade utility poles is as follows: Please see Figure 2 As shown, a batching control system is built. The core control unit adopts a Siemens S7-1500 PLC + industrial computer, and the multi-sensor acquisition unit is configured as follows: Material property sensor: Sand and gravel moisture content sensor (measuring range) Precision ), cement fineness sensor, silica fume specific surface area sensor, admixture solid content detector; Environmental sensors: Mixing tank temperature and humidity sensor, ambient temperature and humidity sensor (measuring range) Precision ;humidity Precision ), online concrete slump tester, and current sensor for the mixing unit; Measurement and Equipment Sensors: 5 high-precision weighing hopper load cells (measuring range) Precision The system includes an infrared displacement sensor for the feeding system, a pressure sensor for the silica fume storage tank, a vibration sensor for the equipment, a valve response time detector, and a belt conveyor speed encoder; all sensors are set to a sampling frequency of 20Hz and communicate with the PLC in two-way real time.

[0029] Nearly 1200 sets of historical production data for C30 concrete from this batching plant were collected to construct a historical database. The criteria include removing outliers, filling in missing data using mean interpolation, and classifying data by production batch; zero-point calibration and range calibration are performed on all sensors, and the C30 concrete mix design specifications, equipment operating safety thresholds, and national standards for concrete performance (slump) are preset. 28-day compressive strength ).

[0030] In step S2, the specific implementation process of the improved principal component analysis is as follows: Step S21: Standardize the total impact indicators. Assume there are m total impact indicators for concrete mix design, and N sets of historical samples. The first sample Each indicator takes a value of The standardized formula is: ; In the formula, The standardized index value, For the first The sample mean of each indicator, For the first The sample standard deviation of each indicator; Step S22: Construct the correlation coefficient matrix of the standardized indicators The formula is: ; In the formula, Let be the correlation coefficient between the i-th and j-th concrete mix design indices. This represents the total number of historical production samples. Step S23: Solve for the correlation coefficient matrix eigenvalues and the corresponding feature vectors, and construct from the feature vectors For each principal component, calculate the information contribution rate of each principal component. With cumulative contribution rate The calculation formula is: ; When the cumulative contribution rate When, select the previous Each principal component is used as the core principal component, and the absolute value of the loading in each principal component is selected. The indicators are used as key influencing factors.

[0031] Specifically, based on the actual concrete production process, the indicators affecting the effective utilization rate of materials and the batching effect are divided into four categories (raw material attributes, working environment, equipment operation, and historical performance). A complete set of indicators is constructed to ensure coverage of the core influencing dimensions, while obviously irrelevant indicators (such as operator name, equipment number, etc.) are removed.

[0032] Raw material properties include sand and gravel moisture content, sand and gravel mud content, cement fineness (specific surface area), cement strength grade, silica fume specific surface area, admixture solid content, and fly ash loss on ignition; operating environment includes ambient temperature, ambient relative humidity, mixing tank temperature, and sand and gravel stockpile temperature; equipment operation includes mixer speed, mixing time, belt conveyor speed, valve response time, and weighing hopper discharge speed; historical performance includes the slump of the previous batch of concrete and the 28-day compressive strength of the previous batch of concrete; assuming the final determined total quantity impact indicators are... One, denoted as .

[0033] Collect at least N sets (N≥500 recommended to ensure statistical significance) of historical production full data. The first sample Each indicator takes a value of And perform the following preprocessing: Outlier removal uses the 3σ criterion to remove distorted data. If the sample value of a certain indicator... satisfy If so, it is considered an outlier and removed; Missing value imputation uses mean imputation to fill in removed or missing data, with the imputed value being the sample mean of the indicator. ; Because different indicators have significantly different dimensions (e.g., temperature is measured in °C, moisture content in %), the data needs to be Z-score standardized to eliminate the influence of dimensions. The formula is as follows: ; Calculate the Pearson correlation coefficient matrix of the standardized indicators. This quantifies the linear correlation between any two indicators, providing a basis for dimensionality reduction. Among them, : No. The first indicator and the first The correlation coefficients of the indicators range from [−1, 1]; when A strong correlation indicates a high degree of overlap between the two indicators, which can be merged through dimensionality reduction; when Indicates moderate relevance; This indicates a weak correlation and low information overlap.

[0034] In specific implementation, the concrete used for producing C30 strength grade utility poles at a certain batching plant was selected as the research object. A typical influencing indicator, collected. The following table shows the indicators for a set of historical samples (simulating a reasonable data range):

[0035] Assuming the sand and gravel moisture content of the 12th sample group is... The mean of the indicator was calculated. Standard deviation ,but These are identified as outliers and removed. Missing value imputation: The mean of the remaining 49 data sets using this indicator. Fill in Group 12 ; Data standardization: Z-score standardization was performed on all 50 groups of 8 indicators to obtain a dimensionless standardized data matrix. Some examples are shown in the table below:

[0036] The 8×8 correlation coefficient matrix R is calculated (partial example):

[0037] You can see (Moisture content of sand and gravel) and (Ambient temperature) (Ambient humidity) is strongly correlated. (Specific surface area of ​​cement) and The strong correlation between (specific surface area of ​​silica ash) indicates information overlap, necessitating dimensionality reduction.

[0038] Eight eigenvalues ​​were obtained, and the information contribution rate and cumulative contribution rate were calculated:

[0039] Screening results: Cumulative contribution rate of the first 3 principal components Therefore, select One core principal component , , The dimensions were reduced from 8 to 3 while retaining 90.7% of the original information.

[0040] Analyze the eigenvectors (factor loadings) corresponding to the first three core principal components and filter by absolute value. Indicators:

[0041] Six key factors were selected as inputs for the subsequent predictive model: principal component (Environmental and sand / gravel property factors): (Moisture content of sand and gravel) (Ambient temperature), x5 (Ambient humidity); principal component (Cementitious material property factors): (Specific surface area of ​​cement) (Specific surface area of ​​silica ash); principal component (Equipment operating factors): (mixer speed) (Valve response time).

[0042] In step S3, the effective utilization rate of materials is defined as: the ratio of the mass of material that actually participates in the hydration reaction and effectively contributes to the strength and workability of the hardened concrete to the total mass of the same material added during the batching process. The specific formula is as follows: ; In the formula, For the first Effective utilization rate of various materials For the first The effective quality of the material For the first The total mass of the input materials.

[0043] In step S3, the specific implementation process of the fractional-order gray-multi-sensor fusion BP neural network combined model is as follows: Step S31, Fractional Grey Model Preprocessing: Let the non-negative original sequence of the effective utilization rate of a certain material be... The accumulated sequence is obtained by using an r-order accumulation operator, as shown in the formula: ; In the formula, For fractional order, For combinations, construct The cumulative discrete grey model is used to estimate the model parameters using the least squares method, and the basic predicted value of the effective utilization rate is obtained. And calculate the relative error sequence. The formula is: ; Specifically, given the characteristics of small sample sizes and limited data in concrete production, the study aims to capture the overall trend of material utilization efficiency, filter out random noise, and provide basic predicted values ​​for subsequent error compensation.

[0044] Extract the actual effective utilization rate of a certain material from N consecutive production batches from the historical database, and construct a non-negative original time series: ;in, Let be the actual effective utilization rate of the material in the k-th batch. Perform smoothness and quasi-exponential regularity tests on the original sequence to ensure that the grey modeling conditions are met. Smoothness ratio: To meet the smoothness requirements; Grade ratio: This satisfies the requirements of quasi-exponential laws; The original sequence is generated by an r-order accumulation, which weakens the randomness of the sequence and strengthens its trend. The accumulation operator formula is as follows: ; In the formula, The combination number is calculated using the following formula: ; Define boundary conditions: ; when It degenerates into the accumulation operator of the traditional first-order grey model GM(1,1). The advantage of fractional order is that it can be adapted to data with different fluctuation levels by adjusting r. The Particle Swarm Optimization (PSO) algorithm is used to find the optimal solution. The optimization process, with the objective function being the minimum model mean relative error (MAPE), is as follows: Initialize the particle swarm: particle dimension is 1, search range is... Population size 20, number of iterations 100, learning factor Inertial weight ; Fitness calculation: for each particle We construct a fractional gray model and calculate the MAPE (Marginal Value of the Model Predicted Value) between the model's predicted value and the actual value as the fitness value. Iterative update: Update the individual optimal position and the global optimal position of the particle until the maximum number of iterations is reached, and output the global optimal value of r.

[0045] Based on the r-order cumulative sequence corresponding to the optimal r value Construct a fractional grey differential equation: ;in, The development coefficient reflects the changing trend of the sequence; The gray action quantity reflects the level of change in the data; it is solved using the least squares method. , The formula is: ; Among them, data matrix With observation vector Construction method: ; The solution obtained , Substituting into the time response function, we obtain the predicted value of the r-order cumulative sequence: ; Perform r-order cumulative subtraction restoration (r-IAGO) on the accumulated predicted values ​​to obtain the basic predicted values ​​of the original sequence. The formula for the cumulative reduction operator is: In the formula, It is a fractional combination number.

[0046] The final predicted sequence of material utilization rates is obtained as follows: Calculate the relative error sequence between the base predicted values ​​and the actual values, and use this sequence as the training target for the subsequent BP neural network. In the formula, For the first The relative error of each batch reflects the nonlinear fluctuation part that the fractional gray model cannot fit.

[0047] Step S32: Construct a multi-sensor fusion BP neural network: The network has a 3-layer structure, and the number of neurons in the input layer is the same as the number of core principal components selected in step S2. The number of neurons in the output layer is equal to the number of core material types. The formula for optimizing the number of neurons in the hidden layer is: In the formula, This represents the number of neurons in the hidden layer. The range of values ​​for the adjustment constant is as follows: Using real-time multi-sensor data of key factors as input and the relative error sequence of a fractional-order grey model as the training target, the relative error compensation value is predicted after network training is completed. ; Specifically, step S32 uses the key operating condition factors selected by PCA and collected in real time from multiple sensors as input, and the relative error of the fractional grey model as the training objective. It fits the nonlinear mapping relationship between operating condition fluctuations and prediction errors to achieve dynamic error compensation for trend predictions, thus addressing the problem of insufficient fitting ability of the grey model for nonlinear fluctuations. The input variables of the input layer are the real-time multi-sensor collected values ​​of the key influencing factors selected by PCA, and the number of neurons in the input layer... Equal to the number of key factors (as in the examples) (Corresponding to sand and gravel moisture content, ambient temperature, ambient humidity, silica fume specific surface area, stirring speed, and valve response delay); the output variable of the output layer is the relative error prediction value of the fractional gray model. Number of neurons in the output layer Equal to the number of material types to be predicted (e.g., when predicting a single material). Total Material Forecasting ).

[0048] A three-layer feedforward neural network structure (input layer - hidden layer - output layer) is adopted, and the specific design is as follows: Input layer: Number of neurons The number of key factors corresponding to PCA screening; Hidden layer: The number of neurons is determined by an optimization formula for the concrete scene to avoid overfitting and underfitting. ;in, The range of values ​​for the adjustment constant is as follows: The optimal value is determined through cross-validation; the activation function is adopted. Function, solve The vanishing gradient problem is explained by the following formula: ; Output layer: Number of neurons The relative error to be predicted corresponds to the number of errors to be predicted; the activation function used is the linear function purelin, which is suitable for regression prediction scenarios; Training algorithm: Adaptive gradient descent with momentum (traingdx) is used to speed up convergence and avoid getting stuck in local minima.

[0049] The historical dataset was divided into training, validation, and test sets in an 8:1:1 ratio. The training set was used for updating network weights, the validation set was used to prevent overfitting, and the test set was used to verify the model's generalization ability. The initial learning rate was 0.01, the momentum factor was 0.9, the maximum number of iterations was 1000, and the training objective was the mean squared error. Training is stopped early if the validation set error does not decrease after 6 consecutive iterations. Through error backpropagation, the connection weights and thresholds between the input layer and hidden layers, and between the hidden layer and output layer, are continuously adjusted until the training objective is achieved. After training, the normalized key factor data (collected in real-time from multiple sensors in the current batch) is input into the network, and the predicted relative error compensation value is output. After inverse normalization, it is used to calculate the final predicted value.

[0050] Step S33: Output of the final predicted value of the combined model: In the formula, For the first The final predicted value of the effective utilization rate of the material.

[0051] In specific implementation, the trend prediction of the phased grey model in experimental step S31 is as follows: Feasibility verification: Calculations show that the smoothness ratio of the original sequences is... The grade ratio is in Between these conditions, the gray modeling conditions are met; Optimal Optimization: The optimal fractional order is obtained by using the PSO algorithm. At this point, the model's MAPE is minimized; Generation by order accumulation: The accumulated sequence is calculated. The first five values ​​are: (0.912, 1.657, 2.432, 3.201, 3.938); Parameters were obtained using the least squares method. , ; Basic forecast calculation: After step-by-step reduction and restoration, the basic predicted values ​​of cement effective utilization rate for the first 20 batches are obtained. The values ​​for the first 5 batches are: (0.912, 0.910, 0.919, 0.921, 0.915). Relative error sequence calculation: The relative error sequences of the first 20 batches are obtained, and the values ​​of the first 5 are: ( ), which serves as the training target for the BP neural network.

[0052] The specific process for training a BP neural network with error compensation is as follows: Input and output determined: The input is sensor data of 6 key factors from 20 batches, and the output is the relative error value of the corresponding batch; Data normalization: Normalize the input and output data to [-1, 1]. Network structure settings: Number of neurons in the input layer Number of neurons in the output layer Cross-validation yielded Number of hidden layer neurons ; Network training: The dataset was divided into 8:1:1 ratios, training parameters were set, and after 286 iterations, the validation set MSE reached 0.00087, which met the training objective, and training was stopped. Error prediction for batch 21: Collect real-time key factor data from multiple sensors for batch 21, input it into the trained network, and obtain the relative error prediction value. .

[0053] Final predicted value output: Basic predicted value of cement effective utilization rate for the 21st batch using the fractional grey model. Substituting into the final prediction formula of the combined model: The final predicted effective utilization rate of the 21st batch of cement was 0.917.

[0054] In step S4, the specific implementation process of the multi-constraint, multi-objective dynamic programming model is as follows: Step S41: Calculate the theoretical material input based on the predicted material utilization rate. The specific formula is as follows: ; In the formula, For the first Theoretical input quality of the material, For the first The target effective quality of this material The standard density of concrete, For the total volume of concrete required, For the target strength grade of concrete, the first Standard mass percentage of each type of material; Step S42: Construct the overall objective function for multi-objective optimization: In the formula, To optimize the overall objective value for multiple objectives, The weight coefficients for each sub-objective satisfy the following conditions: ; To account for the overall batching error rate, To account for the overall material loss rate, For the overall cost of ingredients, This is the concrete strength deviation coefficient. The overall load rate of the equipment; Constraints also need to be set, as follows: ; In the formula, For the first Optimize the input quality of various materials. , For material usage limits, To optimize the concrete compressive strength corresponding to the mix proportions, The target compressive strength of concrete. The slump of the concrete. The linear speed of the equipment; Step S43: Solve the dynamic programming model using the reverse order solution method to obtain the global optimal batching scheme and the whole process control parameters, including the optimal input amount of each material, belt conveyor speed, mixing speed, distributor opening, feeding speed, and silica fume addition amount.

[0055] Specifically, the entire concrete batching process is divided into n stages based on the types of core materials (in this invention). (Corresponding to the five core materials: cement, sand, water, admixtures, and silica fume), denoted as a stage. Each stage corresponds to a decision on the input of a specific material.

[0056] Define each stage State variables This describes the state of the system up to the current stage and must include all the information needed for decisions in subsequent stages, specifically: In the formula, To reach the first Total mass of all materials put in so far (unit: kg). To reach the first The 28-day compressive strength of concrete (unit: MPa) based on the estimated value of materials already used up to this stage. To reach the first Concrete slump (unit: mm) up to this stage, based on the estimated value of materials already used.

[0057] Define each stage Decision variables Let be the input mass of materials at the current stage (unit: kg), and establish the state transition equation. Describe from the stage status Through decision , transition to stage status Relationship: .

[0058] The state equation can be decomposed into: 1. Total mass transfer: ; Strength prediction value transfer (based on simplified formulas for the contribution of cement and silica fume to strength): ;in, For the first The contribution coefficient of each material to the strength (positive values ​​for cement and silica fume, and 0 values ​​for other materials). For the first Predicted effective utilization rate of the material The standard density of concrete, This represents the total volume of concrete required.

[0059] Slump estimate transfer (based on a simplified formula for the contribution of water to slump): ;in, For the first The contribution coefficient of each material to the slump (water is taken as a positive value, and other materials are taken as 0).

[0060] The boundary conditions are for the last stage ( The optimal value function after that is: Meanwhile, the first phase ( The initial state of ) is: .

[0061] Using a reverse-order solution method, starting from the last stage (k=5), we recursively solve for the optimal value function of each stage. and optimal decision : (Silica Ash Stage): Given a state In the feasible decision set Select ,calculate Take the minimum value and the corresponding optimal decision ; (Admixture Stage): Given State In the feasible decision set Select The state transition equation is obtained as follows: ,calculate Take the minimum value and the corresponding optimal decision ; Recursively: Repeat the above process until... (Cement stage), initial state obtained. The optimal value function and optimal decision ; Output the optimal solution in sequence: from Optimal decision Starting from this point, we obtain the state transition equation. Check again The optimal decision table is obtained Output the optimal decision for each stage in sequence. That is, the globally optimal ingredient ratio.

[0062] In practical implementation, taking the concrete mix design for C30 strength grade utility poles as an example, the specific parameters are as follows: Total concrete volume required. Standard density of concrete Overall quality target Weighting coefficients: (Ingredient precision) (Material loss) (cost), (strength), (Equipment load); Predicted material utilization rate (from step S3): Cement Sand and gravel =0.972, water =0.985, admixture =0.941, silica ash .

[0063] The specific constraints are as follows: Total mass constraint is... Strength constraints are ; slump constraint is The equipment load constraint is that the amount of material input is within the upper and lower limits of usage.

[0064] The reverse solution process is as follows: (Silica fume stage) Given a state Feasible decision Must meet: ; ; Calculate the sub-objective function. (Simplified to local cost + local strength deviation), combined with To obtain the optimal decision and Relationship table.

[0065] (Admixture stage) Given a state Feasible decision Must meet The state transition equation is obtained as follows: ,check The relation table is obtained ,calculate Take the minimum value and Relationship table.

[0066] Proceeding sequentially to Repeat the above process to obtain the following results sequentially. (water), (Sand and gravel) The optimal decision table and optimal value function table for (cement).

[0067] Initial state solution ( Initial state In feasible decision Choose the optimal decision The state transition equation is obtained as follows: .

[0068] from Depart, check The optimal decision table is obtained By iteratively applying these steps, the globally optimal ingredient ratio is finally obtained as shown in the table below:

[0069] The final globally optimal ingredient ratio was verified: (1) Total mass: 28,800 kg, in compliance with constraints; (2) Intensity estimate: It complies with the constraints; (3) Estimated slump: It meets the constraints; (4) Overall objective value: After calculation, the overall objective function is... Compared with the traditional fixed mix ratio scheme, it reduces by 18.2%, achieving global optimization.

[0070] (5) Simultaneously output the optimal control parameters: belt conveyor speed 1.2m / s, mixer speed 35r / min, distributor opening 65%, and feed mass flow rate 28kg / s, to complete the output of the whole process batching scheme.

[0071] In step S5, the specific process for real-time fusion of multi-sensor data and dynamic correction of ingredient parameters is as follows: Step S51, Kalman filtering noise reduction preprocessing: Based on the optimal system state estimate and system control quantity at time k-1, the system state prediction, error covariance prediction, and Kalman gain calculation at time k are completed sequentially. Combined with the original measured values ​​of the sensors at time k, the optimal system state estimate that fits the actual working conditions on site is obtained. The error covariance matrix is ​​updated synchronously to provide a basis for the next acquisition cycle. Filtering parameters that match the measurement accuracy and equipment characteristics are configured for different types of sensors such as weighing, temperature and humidity, and displacement. This process is executed cyclically with the sensor acquisition frequency, effectively eliminating noise interference caused by equipment vibration, temperature drift, and electromagnetic interference on the production site, and obtaining the effective measurement data corresponding to each sensor. Specifically, based on the optimal estimate of the system state at time k-1, and combined with the system control input at time k-1, the predicted system state at time k is calculated using the state transition matrix and the control input matrix. Based on the error covariance matrix at time k-1, combined with the state transition matrix and the process noise covariance matrix, the error covariance prediction value corresponding to the state prediction value at time k is calculated, and the uncertainty of the state prediction is quantified. Based on the predicted error covariance at time k, combined with the observation matrix and the observation noise covariance matrix, the Kalman gain at time k is calculated. This Kalman gain is used to balance the acceptance weights of the system state prediction and the sensor measured values. The original measured values ​​of the sensor at time k are collected, and the predicted system state at time k is corrected by Kalman gain to obtain the optimal estimated value of the system state at time k that fits the actual working conditions on site, which is the effective measurement data of the sensor after noise reduction. Based on the Kalman gain, observation matrix, and error covariance prediction value at time k, the error covariance matrix at time k is updated to provide a basis for filtering calculation in the next acquisition cycle. The above steps are executed cyclically with the real-time acquisition frequency of the sensor. For different types of sensors such as weighing, temperature and humidity, and displacement, process noise covariance matrix, observation noise covariance matrix, and state transition matrix are configured to match their measurement accuracy and equipment characteristics.

[0072] The formulas involved are as follows: ; In the formula, for The predicted state value at time 10:00. for The optimal state estimate at time t. Here is the state transition matrix. To control the input matrix, for System control variables at any given time. Let covariance matrix be the variance matrix. The process noise covariance matrix is... For Kalman gain, For the observation matrix, To observe the noise covariance matrix, for Actual measured value from the time sensor. It is an identity matrix.

[0073] Step S52, Adaptive Weighted Data Fusion: For the effective sensor measurement data of the same measurement target in the concrete batching process after Kalman filtering, adaptive weighted data fusion processing is performed. First, the measurement fluctuation of each sensor in the current and consecutive previous acquisition cycles is statistically analyzed to calculate the real-time measurement variance. Then, dynamic weight coefficients are assigned to the effective measurement data of each sensor based on the measurement variance. The smaller the measurement variance, the larger the weight coefficient, and the sum of all weight coefficients is 1. Finally, all effective measurement data are weighted and fused according to the dynamic weight coefficient to obtain the unique, accurate and stable comprehensive measurement value of the measurement target at the current moment. The dynamic weight coefficient is updated synchronously with the real-time measurement variance of the sensor to realize the adaptive adjustment of the fusion process. Specifically, the effective measurement data of all sensors targeting the same measurement target are collected after processing by S51. The measurement fluctuation of each sensor in the current and consecutive previous acquisition cycles is statistically analyzed, and the measurement variance of each sensor at the current moment is calculated. This measurement variance is used to quantify the real-time measurement stability of each sensor. Based on the real-time measurement variance of each sensor, a dynamic weighting coefficient is assigned to the effective measurement data of each sensor. The smaller the measurement variance, the larger the corresponding dynamic weighting coefficient, and the larger the measurement variance, the smaller the corresponding dynamic weighting coefficient. The sum of the dynamic weighting coefficients of all sensors is 1. According to the dynamic weight coefficients corresponding to each sensor, all filtered effective measurement data are weighted and fused to obtain the comprehensive measurement value of the target at the current moment. The aforementioned dynamic weight coefficients are dynamically updated with the real-time measurement variance of each sensor to achieve adaptive adjustment of the fusion process and ensure the accuracy and stability of the comprehensive measurement value. Adaptive weighted data fusion, the formula is: ; In the formula, The value is the combined measurement after fusion at time t. Let be the dynamic weight of the i-th sensor at time t. The effective measurement value after filtering for the i-th sensor. Let be the measurement variance of the i-th sensor at time t.

[0074] Step S53, Real-time Dynamic Correction of Batching Parameters: Based on the comprehensive material measurement value obtained by adaptive weighted fusion, combined with the predicted material effective utilization rate obtained in step S3 and the real-time environmental temperature and humidity data, the input amount of each material in the batching process is dynamically corrected in real time. First, the comprehensive measurement value is compared with the target material weighing value to obtain the real-time deviation. Then, the basic correction is completed through the material effective utilization rate, and the secondary correction is completed through the temperature and humidity comprehensive correction coefficient. The corrected material input mass at the current moment is calculated by combining the results of multiple corrections. Then, it is converted into the execution unit control parameters and sent to the PLC core control unit. The PLC adjusts the opening of the weighing hopper valve, the speed of the material conveying equipment, and other execution actions in real time. This correction process is continuously executed throughout the entire process of material weighing and conveying. When the weighing is close to the target value, the feeding speed is adjusted to avoid overfeeding, so as to realize the closed-loop dynamic correction of batching parameters and the precise control of batching accuracy.

[0075] Specifically, based on the comprehensive measurement values ​​obtained from S52, combined with the predicted material utilization rate and the influence of ambient temperature and humidity, the input quantities of each material in the batching process are dynamically corrected in real time. The corrected material input quality is then obtained and sent to the execution unit. The specific correction process is as follows: The current comprehensive material measurement value is compared with the target weighing value of the material at the current moment set in the global optimal batching scheme to obtain the real-time deviation value between the two. Based on the predicted effective utilization rate of the material obtained in step S3, the initial material input amount is adjusted to ensure that the actual effective material usage in the concrete hydration reaction matches the requirements of the optimal batching scheme. Collect the ambient temperature and relative humidity data at the current moment, and use the preset temperature and humidity comprehensive correction coefficient to make a second correction to the material input amount after the basic correction, so as to eliminate the impact of changes in ambient temperature and humidity on material characteristics and batching accuracy. Based on the above real-time deviation value, the basic correction of material effective utilization rate, and the secondary correction of temperature and humidity, the corrected input quality of the i-th material at the current moment is calculated. The corrected material input mass is converted into corresponding control parameters for the execution unit and sent to the PLC core control unit in real time. The PLC then adjusts the valve opening of the weighing hopper, the rotation speed of the material conveying equipment, the feeding speed, and other execution actions to achieve closed-loop dynamic correction of the batching parameters. The above correction process is continuously executed throughout the entire process of material weighing and conveying. When the material weighing approaches the target value, the operating speed of the feeding equipment is adjusted synchronously to avoid overfeeding and ensure that the batching error is controlled within the preset threshold range.

[0076] The ingredient parameters are dynamically adjusted in real time, using the following formula: ; In the formula, for The time correction The quality of input materials for The target weighing value of the material at any given time. This is a comprehensive correction factor for temperature and humidity. For ambient temperature, This refers to the relative humidity of the environment.

[0077] In step S6, the specific implementation process of multi-system collaborative automation control is as follows: Step S61: The mixing system and the inclined feeding system are controlled in a coordinated manner, and the coordination formula is as follows: ; In the formula, for Constantly monitor the mixer speed. For the synergy coefficient, for The feed quality flow rate after constant mixing for The total mass of materials in the mixing tank is monitored at all times; simultaneously, the material distribution rate is corrected using an aerial material compensation formula, which is: ; In the formula, Let t be the actual material distribution mass. To achieve the target material quality, To compensate for the mass of materials falling from the air. The feeding delay time is used to monitor the feeding system's operating status in real time. When the deviation continuously exceeds the threshold, it is judged as a fault and an alarm and emergency shutdown are triggered. In step S62, the silica fume weighing and stirring system is under fully automatic coordinated control. The silica fume inventory is calculated in real time using a pressure sensor, and the formula is as follows: ; In the formula, for The total mass of silica fume in the silica fume storage tank at any given time. for The pressure measurement value at the bottom of the storage tank at all times. The cross-sectional area of ​​the bottom of the storage tank. The acceleration due to gravity is used; based on the optimal batching scheme, the silica fume conveying pump is controlled to achieve automatic extraction and one-time weighing. After weighing, it is directly discharged into the mixing tank, and the silica fume dosage is adjusted based on the ambient humidity.

[0078] In step S6, the specific process of closed-loop iterative optimization is as follows: After each batching is completed, the system automatically collects the actual material usage, measured concrete strength, slump, material loss, and equipment operation data of this production, calculates the actual effective utilization rate of each material, and updates it to the historical database; after every 50 production batches are completed, the PCA model, combined prediction model, and dynamic programming model are retrained and their parameters are optimized to achieve adaptive iteration of the model.

[0079] Specifically, the core of collaborative control is "data interoperability + command linkage + status feedback." With the core control unit (PLC + industrial computer) as the central hub, it integrates real-time data and control commands from the mixing system, tilting feeding system, and silica fume weighing system to achieve "one-button start, fully unmanned operation, and self-diagnosis of faults." The specific process is divided into three sub-processes: Sub-process 1: System initialization and command issuance Parameter synchronization: The core control unit reads the optimal batching scheme (amount of each material input and equipment operating parameters) output in step S4 and synchronizes it to the local controllers of the mixing system, feeding system and silica fume weighing system. Status self-check: Each subsystem performs an initialization self-check (sensor zero point, valve opening, motor speed, material inventory), and sends a "ready" signal back to the core control unit after the self-check passes. Command issuance: The core control unit issues start commands according to a preset sequence, with the priority being: silica fume weighing system → material weighing → belt conveyor → mixing system preheating → feeding system positioning.

[0080] Sub-process 2: Coordinated control of the mixing system and the inclined feeding system The core of this sub-process is the dynamic matching of "feed flow rate - stirring speed - dispensing accuracy", with specific steps as follows: Material weighing and conveying: The core control unit sends weighing instructions to each weighing hopper according to the optimal input amount. The weighing value is calibrated in real time by the weighing sensor data after Kalman filtering. When the error is ≤0.1%, the "weighing completed" signal is triggered. The belt conveyor starts at the optimal speed (output of step S4). The infrared displacement sensor monitors the material position on the belt in real time. The core control unit adjusts the belt speed according to the material position to avoid material accumulation / material breakage. Mixing system preheating and parameter setting: The mixing host preheats at the optimal speed (output of step S4), and the current sensor monitors the motor load in real time. If the load exceeds 80% of the rated value, the speed is automatically reduced and an alarm is triggered. The core control unit dynamically adjusts the mixing speed according to the feed mass flow rate (after fusion) through a collaborative formula. The synergy coefficient k is dynamically adjusted according to the concrete slump target (150 for C30 and 180 for C50). Feeding and dispensing coordination: The feeding system starts after receiving the signal that "the material in the mixing tank has reached 50% of its volume". The infrared displacement sensor monitors the position of the feeding hopper to ensure precise docking with the discharge port of the mixing tank. The core control unit calculates the compensation amount for material falling in mid-air (based on the material falling delay time). The material distribution rate is adjusted in real time by fitting the valve response time plus the material fall time. ; The actual value of the feed sample is compared with the target value in real time. If the deviation is greater than 0.2%, it is determined to be a feeding system malfunction. The system will be stopped immediately and an audible and visual alarm will be triggered. At the same time, the fault data will be recorded. Status feedback: Real-time data from each stage (speed, flow rate, mass, position) is uploaded to the core control unit every 100ms to form a visual monitoring interface.

[0081] Sub-process 3: Fully automated coordinated control of silica fume weighing and stirring system This sub-process completely eliminates manual operation, achieving full automation of the silica fume process: "inventory monitoring → automatic extraction → accurate weighing → direct feeding." Specific steps include: Real-time monitoring of silica fume inventory: The core control unit calculates the silica fume inventory in real time using data from the pressure sensor at the bottom of the silica fume storage tank. The cross-sectional area S at the bottom of the storage tank is preset to 0.8 m², and the gravitational acceleration g is taken as 9.8 m / s². If the storage volume is less than 1.2 times the optimal addition amount, a "replenishment warning" will be automatically triggered to notify the operation and maintenance personnel. Automatic extraction and precise weighing: The core control unit sends an extraction command to the silica fume delivery pump, which starts at a preset flow rate (0.5 kg / s). The weighing sensor monitors the silica fume mass in the weighing hopper in real time. When the weighing value reaches 98% of the optimal dosage, the delivery pump speed is reduced to 0.1 kg / s. When the optimal value is reached, extraction stops immediately, and the weighing accuracy is controlled within ±0.08%. Temperature and humidity correction and feeding: The core control unit reads data from the ambient humidity sensor. If the humidity is greater than 60% RH, it fine-tunes the silica fume dosage according to the correction formula.

[0082] After the mixing system sends a signal that "the material in the mixing tank has reached 80% of its volume", the valve of the silica fume weighing hopper will automatically open, and the silica fume will be directly discharged into the mixing tank without the need for secondary transfer. Post-feeding verification: After the weighing hopper is emptied, the core control unit verifies the remaining mass. If it is greater than 0.1 kg, it is determined that the valve is stuck and a cleaning warning is triggered.

[0083] In practice, taking the concrete batching scenario for C30 utility poles as an example, the complete implementation steps are as follows: 1. Basic parameters are as follows: Optimal batching scheme (output of step S4): cement 4352kg, sand and gravel 21687kg, water 1462kg, admixture 306kg, silica fume 2251kg; Optimal equipment parameters: belt conveyor speed 1.2m / s, mixer speed 35r / min, feed flow rate 28kg / s; Environmental parameters: temperature 22℃, humidity 58%RH.

[0084] 2. Implementation process of multi-system collaborative control Step 1: System Initialization The core control unit synchronizes the optimal parameters to each subsystem. Self-test results: zero-point error of the weighing sensor is 0.02%, silica fume inventory is 2500kg (meeting the requirement of 2251kg), and the infrared displacement sensor of the feeding system has been calibrated. The core control unit issues a "start" command with the following sequence: silica fume weighing system starts (t=0s) → cement / sand weighing starts (t=5s) → belt conveyor starts (t=30s) → mixing system preheats (t=60s).

[0085] Step 2: Coordinated control of mixing and feeding The cement weighing hopper completed weighing (4352 kg, error 0.03%), and the belt conveyor transported the material at a speed of 1.2 m / s. The infrared displacement sensor detected the material in the middle section of the belt, and the speed remained stable. The mixing host preheated to 35 r / min, and the current sensor showed a load of 75% (<80% of the rated value), with no warning. When the material in the mixing tank reached 50% of its volume (14400 kg), the feeding system started, and the infrared displacement sensor calibrated the position of the feeding hopper, with a deviation of 0.5 mm, achieving precise docking. The material discharge delay time... Feed mass flow rate Compensation amount The target material distribution was 1200 kg / rod, and the actual material distribution was 1207.8 kg, with a deviation of 0.65% (≤0.2%). There were no fault alarms throughout the process, and the feeding system operated stably.

[0086] Step 3: Fully Automatic Control of Silica Fume Actual measurement of pressure sensor Calculate stock The requirements are met; the transfer pump extracts silica fume up to 2251 kg with a weighing error of 0.05%; ambient humidity... No need to adjust the dosage; when the material in the mixing tank reaches 80% of the volume (23040kg), the silica fume is directly discharged into the mixing tank. After emptying, the remaining mass in the weighing hopper is 0.05kg, with no material jamming. 3. Closed-loop iterative optimization implementation process Step 1: Production Data Collection Actual production data: 4350kg cement, 21690kg sand and gravel, 1460kg water, 305kg admixtures, and 2250kg silica fume; Actual concrete measurement data: slump 125mm, 28-day compressive strength 34.2MPa (target 30MPa); Calculated actual effective utilization rate: cement (Predicted value 0.926, deviation 0.54%), silica ash (Predicted value 0.893, deviation 0.56%) Step 2: Data import The data showed no anomalies and was updated to the historical database. The cumulative production batches reached 50 (this is the 50th batch), triggering model iteration. Step 3: Iterative Model Optimization PCA model: After adding 50 sets of data, the core principal components remain at 6, and the cumulative contribution rate increases to 92.8%; Combinatorial prediction model: The BP neural network was retrained, and the mean squared error of the validation set decreased from 0.001 to 0.0008, with the optimal fractional order r fine-tuned from 0.85 to 0.86; Dynamic programming model: The material loss rate decreased from 12% to 11.5%, and the weights were fine-tuned. From 0.1 to 0.12; Step 4: Verify the iterative effect After iteration, the model was applied to the next batch of C30 feedstock. The predicted effective utilization rate of cement was 0.930, the actual value was 0.932, the deviation was 0.21%, the prediction accuracy was improved by 0.33%, and the new parameters were solidified.

[0087] Example 1

[0088] This embodiment provides a fully automated concrete batching optimization method based on multi-sensor fusion, applied to the concrete batching scenario for the production of C30 strength grade utility poles. The specific implementation process is as follows: 1. System setup, data acquisition and initialization A batching control system was built, with the core control unit using a Siemens S7-1500 PLC and an industrial computer. The multi-sensor acquisition unit was configured as follows: Material property sensor: Sand and gravel moisture content sensor (measuring range) (Accuracy ±0.2%), cement fineness sensor, silica fume specific surface area sensor, and admixture solid content detector; Environmental sensors: Mixing tank temperature and humidity sensor, ambient temperature and humidity sensor (measuring range) , accuracy ±0.2℃; humidity (Accuracy ±2%RH), online concrete slump tester, and current sensor for mixing host; Measurement and Equipment Sensors: 5 high-precision weighing hopper load cells (measuring range) The system includes an infrared displacement sensor for the feeding system (accuracy 0.05%FS), a pressure sensor for the silica fume storage tank, a vibration sensor for the equipment, a valve response time detector, and a belt conveyor speed encoder; all sensors are set to a sampling frequency of 20Hz and communicate with the PLC in real time.

[0089] Nearly 1200 sets of historical production data for C30 concrete from this batching plant were collected to construct a historical database. Outliers were removed according to the criteria, missing data were filled in using the mean interpolation method, and data were classified according to production batches. Zero-point calibration and range calibration were performed on all sensors, and the C30 concrete mix proportion specification range, equipment operation safety threshold, and national standard requirements for concrete performance (slump 120±20mm, 28d compressive strength ≥30MPa) were preset.

[0090] 2. Key factor screening based on improved principal component analysis Twenty-two total influencing indicators across four categories—raw material properties, working environment, equipment operation, and performance parameters—were selected from 1200 historical samples, and data standardization was completed. A 22×22 correlation coefficient matrix was constructed, and eigenvalues ​​and eigenvectors were obtained. The principal component contribution rate was calculated, and the top six core principal components were finally selected, with a cumulative contribution rate of 92.3%, meeting the requirement of ≥90%. Further screening revealed six key influencing factors with an absolute load value ≥0.7: sand and gravel moisture content, ambient temperature, ambient humidity, silica fume specific surface area, stirring speed, and valve response delay. These six key factors served as input variables for subsequent prediction models.

[0091] 3. Dynamic prediction of material utilization rate Using six selected key factors as input and the effective utilization rates of five core materials (cement, sand, water, admixtures, and silica fume) as output, a fractional-order gray-multi-sensor fusion BP neural network combined model is constructed: Fractional gray model: The optimal fractional order r=0.85 was obtained by particle swarm optimization. The historical effective utilization rate sequences of 5 materials were preprocessed to obtain the basic predicted values ​​and relative error sequences. BP Neural Network: A 3-layer network structure is constructed, with 6 neurons in the input layer and 5 neurons in the output layer. The number of neurons in the hidden layer is calculated to be 10 through optimization formula. The tanh activation function is used, and the network is trained using gradient descent with momentum. The historical dataset is divided into training and validation sets in an 8:2 ratio. Training is stopped when the mean square error of the validation set is ≤0.001. Final prediction: Input key factor data collected in real time by multiple sensors, and output the predicted effective utilization rates of 5 materials through a combined model, which are: cement 0.926, sand and gravel 0.972, water 0.985, admixture 0.941, and silica fume 0.893.

[0092] 4. Overall optimization of ingredient formulation Concrete demand for this production Standard density of concrete The standard mass proportions of each material in C30 concrete are: cement 14%, sand and gravel 73%, water 5%, admixtures 1%, and silica fume 7%. Based on the predicted effective utilization rate, the theoretical input of each material is calculated, and a multi-objective dynamic programming model is constructed, with weight coefficients set as follows: By setting constraints and using the reverse order solution method, the globally optimal batching scheme is obtained. The optimal input of core materials is: cement 4352kg, sand and gravel 21687kg, water 1462kg, admixture 306kg, and silica fume 2251kg. At the same time, the optimal control parameters are output: belt conveyor speed 1.2m / s, mixer speed 35r / min, distributor opening 65%, feed mass flow rate 28kg / s, and silica fume addition amount 2251kg.

[0093] 5. Multi-sensor data fusion and real-time parameter correction Kalman filtering is used to preprocess the real-time data from each weighing sensor, infrared displacement sensor, and temperature and humidity sensor to remove noise. An adaptive weighted fusion algorithm is then used to obtain the comprehensive measurement value. The fused data is compared with the optimal batching scheme parameters in real time. When the deviation of the sand and gravel weighing value exceeds 0.1%, the amount of sand and gravel input is adjusted in real time using a correction formula. At the same time, based on the real-time ambient temperature of 22℃ and relative humidity of 58%RH, the amount of cement and silica fume is finely adjusted using temperature and humidity correction coefficients to ensure that the batching accuracy is controlled within 0.1%.

[0094] 6. Multi-system collaborative control and closed-loop optimization Mixing-feeding coordinated control: Based on the fused feed mass flow rate and the total mass of material in the mixing tank, the mixer speed is adjusted in real time through a coordinated formula to achieve synchronous matching of feeding and mixing; the material distribution amount is corrected through an aerial material drop compensation formula, the material drop delay time is set to 0.3s, and the material distribution accuracy is controlled within 0.2%; the weighing data of the feeding system is monitored in real time, and when the deviation exceeds 1% continuously, an alarm is triggered and the machine is stopped to avoid insufficient feeding weight of the pole; The silica fume system is fully automated: the pressure sensor monitors the silica fume level in real time, and the silica fume conveying pump is automatically controlled to extract the fume based on the optimal batching scheme. After weighing to the target value of 2251kg in one go, it is directly discharged into the mixing tank without manual operation or secondary weighing. At the same time, the silica fume dosage is corrected based on the ambient humidity, and the weighing accuracy reaches 0.07%. Closed-loop iterative optimization: After the batching is completed, the measured 28-day compressive strength of concrete is 34.2 MPa and the slump is 125 mm, which meets the national standard requirements. The actual effective utilization rate of each material is calculated and updated to the historical database. Every 50 production batches are completed, the model is retrained and the parameters are optimized to achieve adaptive iteration.

[0095] Example 2

[0096] The difference between this embodiment and Embodiment 2 is that it is applied to the C50 strength grade high-performance concrete batching scenario. The preset concrete performance requirements are a slump of 180±20mm and a 28-day compressive strength ≥50MPa. An improved PCA model was used to screen out eight core principal components, with a cumulative contribution rate of 91.7%. Key influencing factors include sand and gravel moisture content, cement fineness, ambient temperature, silica fume specific surface area, admixture solid content, stirring speed, stirring current, and valve response delay. The combined model predicted the following material utilization rates: cement 0.908, sand and gravel 0.965, water 0.981, and admixture 0.9. 32. Silica fume 0.887; The optimal batching scheme was obtained by solving the dynamic programming model. The final produced concrete had a measured compressive strength of 56.8 MPa after 28 days, a slump of 186 mm, a batching error of 0.08%, and a material loss reduction of 13.2%, fully meeting the production requirements of high-performance concrete. It is worth noting that in the above system embodiment, the various units included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the scope of protection of this invention.

[0097] Furthermore, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware, and the corresponding program can be stored in a computer-readable storage medium.

[0098] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A fully automated concrete batching optimization method based on multi-sensor fusion, characterized in that, Includes the following steps: Step S1, System Setup, Data Acquisition and Initialization: Build a batching control system including a core control unit, a multi-sensor acquisition unit, an execution unit, and a data storage unit; collect historical production data to build a database and complete preprocessing; and initialize and calibrate the system and sensors. Step S2: Dimensionality reduction and key factor screening of batching influencing factors based on improved principal component analysis: Standardize the total influencing indicators of concrete batching, construct the correlation coefficient matrix and solve the eigenvalues ​​and eigenvectors, screen the core principal components through the cumulative contribution rate, and determine the key factors affecting the effective utilization rate of materials and batching effect; Step S3: Dynamic prediction of material utilization rate based on fractional gray-multi-sensor fusion BP neural network combined model: Using the real-time data of multi-sensor corresponding to the key factors selected in step S2 as input, the basic predicted value and relative error sequence of material utilization rate are obtained through fractional gray model, and the relative error is predicted and compensated through BP neural network, and finally the dynamic predicted value of the utilization rate of each material is output. Step S4: Global optimization of batching scheme based on multi-constraint and multi-objective dynamic programming: Based on the material effective utilization rate predicted in step S3, the theoretical input of materials is calculated back, and a multi-constraint and multi-objective dynamic programming model is constructed with the objectives of minimizing batching error, minimizing material loss, optimizing production cost, achieving concrete strength standards, and balancing equipment operating load. The global optimal batching scheme and the whole process control parameters are solved. Step S5: Real-time fusion of multi-sensor data and dynamic correction of batching parameters: The real-time data collected by multiple sensors is denoised and preprocessed by Kalman filtering. An adaptive weighted fusion algorithm is used to obtain the comprehensive measurement value. Combined with the optimal batching scheme and the predicted value of material effective utilization rate, the material input and equipment operating parameters in the batching process are dynamically corrected in real time. Step S6, Multi-system collaborative automated control and closed-loop iterative optimization: Based on the optimal control parameters and real-time correction results, fully automatic collaborative control of the mixing system, tilting feeding system and silica fume weighing system is realized. After each batching is completed, the historical database is updated, and the model is iteratively optimized to form a closed-loop control of the entire process.

2. The fully automated concrete batching optimization method based on multi-sensor fusion according to claim 1, characterized in that, In step S1, the multi-sensor acquisition unit includes material property sensors, environmental sensors, and metering and equipment sensors. The material property sensors include sand and gravel moisture content sensors, cement fineness sensors, silica fume specific surface area sensors, and admixture solid content detectors. The environmental sensors include mixing tank temperature and humidity sensors, ambient temperature and humidity sensors, online concrete slump detectors, and mixing host current sensors. The metering and equipment sensors include high-precision weighing sensors for each weighing hopper, infrared displacement sensors for the feeding system, pressure sensors for the silica fume storage tank, equipment vibration sensors, valve response time detectors, and belt conveyor speed encoders. All sensors have a sampling frequency of no less than 10Hz and communicate bidirectionally with the core control unit in real time.

3. The fully automated concrete batching optimization method based on multi-sensor fusion according to claim 1, characterized in that, In step S2, the specific implementation process of the improved principal component analysis is as follows: Step S21: Standardize the total impact indicators. Assume there are m total impact indicators for concrete mix design, and N sets of historical samples. The first sample Each indicator takes a value of The standardized formula is: ； In the formula, The standardized index value, For the first The sample mean of each indicator, For the first The sample standard deviation of each indicator; Step S22: Construct the correlation coefficient matrix of the standardized indicators The formula is: ； In the formula, For the first The and the first The correlation coefficients between the various concrete mix design influencing factors This represents the total number of historical production samples. Step S23: Solve for the correlation coefficient matrix eigenvalues and the corresponding feature vectors, and construct from the feature vectors For each principal component, calculate the information contribution rate of each principal component. With cumulative contribution rate The calculation formula is: 。 4. The fully automated concrete batching optimization method based on multi-sensor fusion according to claim 1, characterized in that, In step S3, the effective utilization rate of materials is defined as: the ratio of the mass of material that actually participates in the hydration reaction and effectively contributes to the strength and workability of the hardened concrete to the total mass of the same material added during the batching process. The specific formula is as follows: ； In the formula, For the first Effective utilization rate of various materials For the first The effective quality of the material For the first The total mass of the input materials.

5. The fully automated concrete batching optimization method based on multi-sensor fusion according to claim 1, characterized in that, In step S3, the specific implementation process of the fractional gray-multi-sensor fusion BP neural network combination model is as follows: Step S31, Fractional Grey Model Preprocessing: Let the non-negative original sequence of the effective utilization rate of a certain material be... The accumulated sequence is obtained by using an r-order accumulation operator, as shown in the formula: ; In the formula, For fractional order, For combinations, construct The cumulative discrete grey model is used to estimate the model parameters using the least squares method, and the basic predicted value of the effective utilization rate is obtained. And calculate the relative error sequence. The formula is: ； Step S32: Construct a multi-sensor fusion BP neural network: The network has a 3-layer structure, and the number of neurons in the input layer is the same as the number of core principal components selected in step S2. The number of neurons in the output layer is equal to the number of core material types. The formula for optimizing the number of neurons in the hidden layer is: In the formula, This represents the number of neurons in the hidden layer. The range of values ​​for the adjustment constant is as follows: ; Using real-time multi-sensor data of key factors as input and the relative error sequence of a fractional-order grey model as the training objective, the relative error compensation value is predicted after network training is completed. ; Step S33: Output of the final predicted value of the combined model: In the formula, For the first The final predicted value of the effective utilization rate of the material.

6. The fully automated concrete batching optimization method based on multi-sensor fusion according to claim 1, characterized in that, In step S4, the specific implementation process of the multi-constraint, multi-objective dynamic programming model is as follows: Step S41: Calculate the theoretical material input based on the predicted material utilization rate. The specific formula is as follows: ； In the formula, For the first Theoretical input quality of the material, For the first The target effective quality of this material The standard density of concrete, For the total volume of concrete required, For the target strength grade of concrete, the first Standard mass percentage of each type of material; Step S42: Construct the overall objective function for multi-objective optimization: In the formula, To optimize the overall objective value for multiple objectives, The weight coefficients for each sub-objective satisfy the following conditions: ; To account for the overall batching error rate, To account for the overall material loss rate, For the overall cost of ingredients, This is the concrete strength deviation coefficient. The overall load rate of the equipment; Step S43: Solve the dynamic programming model using the reverse order solution method to obtain the global optimal batching scheme and the whole process control parameters, including the optimal input amount of each material, belt conveyor speed, mixing speed, distributor opening, feeding speed, and silica fume addition amount.

7. The fully automated concrete batching optimization method based on multi-sensor fusion according to claim 6, characterized in that, In step S43, the specific process of solving the dynamic programming model using the reverse order method is as follows: Step S431: Starting from the last stage, recursively solve for the optimal value function of each stage. and optimal decision : Step S432: In Stage given state In the feasible decision set Select ,calculate Take the minimum value and the corresponding optimal decision ; Step S433: In Stage given state In the feasible decision set Select The state transition equation is obtained as follows: ,calculate Take the minimum value and the corresponding optimal decision ; Step S434: Repeat the above process sequentially until... To obtain the initial state The optimal value function and optimal decision ; Step S435: Output the optimal solution sequentially: from Optimal decision Starting from this point, we obtain the state transition equation. Check again The optimal decision table is obtained Output the optimal decision for each stage in sequence. That is, the globally optimal ingredient ratio.

8. The fully automated concrete batching optimization method based on multi-sensor fusion according to claim 1, characterized in that, In step S5, the specific process for real-time fusion of multi-sensor data and dynamic correction of ingredient parameters is as follows: Step S51, Kalman filtering noise reduction preprocessing: Based on the optimal system state estimate and system control quantity at time k-1, the system state prediction, error covariance prediction, and Kalman gain calculation at time k are completed in sequence. Combined with the original measured value of the sensor at time k, the optimal system state estimate that fits the actual working conditions on site is obtained. The error covariance matrix is ​​updated synchronously to provide a basis for the next acquisition cycle. Step S52, Adaptive Weighted Data Fusion: For the effective sensor measurement data of the same measurement target in the concrete batching process after multiple Kalman filtering, adaptive weighted data fusion processing is performed. First, the measurement fluctuation of each sensor in the current and consecutive previous acquisition cycles is statistically analyzed to calculate the real-time measurement variance. Then, dynamic weight coefficients are assigned to the effective measurement data of each sensor based on the measurement variance. Finally, all effective measurement data are weighted and fused according to the dynamic weight coefficients to obtain the comprehensive measurement value at the current moment. Step S53, Real-time Dynamic Correction of Batching Parameters: Based on the comprehensive material measurement value obtained by adaptive weighted fusion, combined with the predicted material effective utilization rate obtained in step S3 and the real-time ambient temperature and humidity data, the input amount of each material in the batching process is dynamically corrected in real time. First, the comprehensive measurement value is compared with the target material weighing value to obtain the real-time deviation. Then, the basic correction is completed through the material effective utilization rate, and the secondary correction is completed through the temperature and humidity comprehensive correction coefficient. The corrected material input quality at the current moment is calculated by combining the multiple correction results, and then converted into the execution unit control parameters and sent to the PLC core control unit.

9. The fully automated concrete batching optimization method based on multi-sensor fusion according to claim 1, characterized in that, In step S6, the specific implementation process of multi-system collaborative automation control is as follows: Step S61: The mixing system and the inclined feeding system are controlled in a coordinated manner, and the coordination formula is as follows: ； In the formula, for Constantly monitor the mixer speed. For the synergy coefficient, for The feed quality flow rate after constant mixing for The total mass of materials in the mixing tank is monitored at all times; simultaneously, the material distribution rate is corrected using an aerial material compensation formula, which is: ； In the formula, Let t be the actual material distribution mass. To achieve the target material quality, To compensate for the mass of materials falling from the air. The feeding delay time is used to monitor the feeding system's operating status in real time. When the deviation continuously exceeds the threshold, it is judged as a fault and an alarm and emergency shutdown are triggered. In step S62, the silica fume weighing and stirring system is under fully automatic coordinated control. The silica fume inventory is calculated in real time using a pressure sensor, and the formula is as follows: ； In the formula, for The total mass of silica fume in the silica fume storage tank at any given time. for The pressure measurement value at the bottom of the storage tank at all times. The cross-sectional area of ​​the bottom of the storage tank. The acceleration due to gravity is used; based on the optimal batching scheme, the silica fume conveying pump is controlled to achieve automatic extraction and one-time weighing. After weighing, it is directly discharged into the mixing tank, and the silica fume dosage is adjusted based on the ambient humidity.

10. The fully automated concrete batching optimization method based on multi-sensor fusion according to claim 9, characterized in that, In step S6, the specific process of closed-loop iterative optimization is as follows: after each batching is completed, the system automatically collects the actual material usage, measured concrete strength, slump, material loss, and equipment operation data of this production, calculates the actual effective utilization rate of each material, and updates it to the historical database; after every 50 production batches are completed, the PCA model, combined prediction model, and dynamic programming model are retrained and their parameters are optimized to achieve adaptive iteration of the model.

Images