Intelligent decision method for curing process of cement production

Abstract

The present application relates to the technical field of cement curing, in particular to an intelligent decision-making method for the curing process of cement production. The technical scheme comprises the following steps: collecting multi-dimensional parameter data in the cement curing process and preprocessing the data. The cement curing stages are divided, and a stage-based prediction model is constructed. Based on the real-time obtained multi-dimensional parameter data and the prediction model linkage analysis, the prediction result is obtained. It is judged whether the prediction result is beyond the curing target threshold range of the stage: if not, the current curing parameters are maintained; if yes, the prediction result is labeled. The labeled prediction result is used to obtain the optimal parameter combination through curing parameter optimization calculation. Through multi-source data acquisition and stage-based prediction model, the cement hydration process is accurately perceived, the curing parameters are dynamically optimized, and the problems of insufficient curing or excessive curing are effectively avoided. Water consumption and heat energy consumption can be reduced, which conforms to the concept of green production.

Description

Technical Field

[0001] This invention relates to the field of cement curing technology, and more specifically to an intelligent decision-making method for the cement production and curing process. Background Technology

[0002] As a fundamental building material, the quality of cement production directly determines the structural safety and service life of construction projects. The curing process is a crucial step in cement production. Cement doesn't simply "dry" and harden; rather, it undergoes a complex hydration reaction with water, generating hard hydration products that bind aggregates such as sand and stone into a cohesive whole (i.e., concrete / mortar). Essentially, it provides the necessary conditions for continuous hydration. The core objectives are to ensure strength development, prevent cracking, improve durability, and guarantee surface quality. Ensuring strength development: The hydration reaction requires water. Insufficient water will cause the reaction to stop, and the strength will not reach the design value. Preventing cracking: The hydration reaction is exothermic and produces shrinkage. If not controlled, excessive internal and external temperature differences or excessively rapid surface shrinkage will result in temperature cracks and shrinkage cracks. Improving durability: Adequate curing makes the structure denser, stronger in resisting the intrusion of harmful substances (chloride ions, carbon dioxide, water), thereby improving impermeability, frost resistance, and corrosion resistance. Guaranteeing surface quality: Preventing surface sanding and powdering. The core of curing is controlling temperature and humidity. Standard curing methods include: watering and covering for moisture retention. Watering: The most common method. After the concrete has set (generally 12-24 hours after pouring), begin watering regularly to keep the surface consistently moist. Suitable for horizontal surfaces (floor slabs, roads). Covering for moisture retention: Covering with damp burlap sacks, straw mats, or geotextiles; after covering, water regularly to keep the covering moist. Covering with plastic film: Cover tightly against the concrete surface, utilizing its own condensation for moisture retention. Effective and water-saving.

[0003] Post-concrete curing is the "last mile" that determines the final performance, durability, and structural lifespan of concrete, and it's also the easiest part to overlook and where shortcuts are taken. The saying "three parts pouring, seven parts curing" is no exaggeration. Scientific, timely, and sufficient curing is the most economical and effective way to ensure project quality, prevent cracking and leakage, and extend the building's lifespan.

[0004] In existing technologies, although some cement production and maintenance equipment have introduced simple sensing and monitoring functions, they can only collect data and lack the ability to conduct intelligent analysis and decision-making based on multi-source data, thus failing to fundamentally solve the drawbacks of traditional maintenance methods. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides an intelligent decision-making method for the cement production and curing process. Through multi-source data fusion analysis, staged model construction, and real-time dynamic optimization, it achieves accurate decision-making regarding curing parameters. The intelligent decision-making method for the cement production and curing process includes the following steps: S1. Collect multi-dimensional parameter data during the cement curing process and preprocess it.

[0006] S2. Divide the cement curing stages and construct a staged prediction model based on the pre-processed multi-dimensional parameter data and curing effect data.

[0007] S3. Based on the real-time acquisition of multi-dimensional parameter data and the linkage analysis of the prediction model, the prediction results are obtained.

[0008] S4. Then compare the prediction result with the maintenance target threshold for this stage to determine whether the prediction result exceeds the maintenance target threshold range for this stage: if not, maintain the current maintenance parameters; if yes, mark the prediction result.

[0009] S5. The optimal parameter combination is obtained by optimizing the maintenance parameters based on the predicted results of the markers.

[0010] S6. Adjust the equipment maintenance parameters according to the optimal parameter combination.

[0011] Preferred: Multi-dimensional parameter data includes basic parameters, environmental parameters, and state parameters.

[0012] Preferred basic parameters include: cement raw material ratio, initial cement temperature, and product molding process parameters.

[0013] Preferred environmental parameters: real-time temperature, relative humidity, wind speed, and atmospheric pressure of the maintenance area.

[0014] Preferred parameters include: surface temperature of cement products, internal core temperature, and surface humidity.

[0015] The preferred prediction model is divided into three core stages: the initial hydration stage (0-72h), the intensity growth stage (72h-14d), and the intensity stabilization stage (14d-28d).

[0016] Preferred method for obtaining the prediction model for the initial hydration stage includes: using basic parameters, ambient temperature, and surface / core temperature difference of the product from historical data as input features; using the initial setting time and final setting time of cement as output targets, and constructing the model using a BP neural network; during model training, using relevant data of the initial hydration stage from historical data as the training set, optimizing the model parameters through gradient descent, and controlling the prediction error to within 5%.

[0017] The preferred method for obtaining the strength growth stage prediction model includes: using the initial hydration stage prediction results, real-time environmental parameters, and product state parameters as input features, and the 7-day compressive strength and 14-day compressive strength of cement products as output targets, a random forest algorithm is used to construct the prediction model; and the generalization ability of the prediction model is improved by introducing a cross-validation mechanism.

[0018] Preferred method for obtaining the strength stabilization stage prediction model includes: using the prediction results of the strength growth stage and the trend of subsequent environmental parameter changes as input features, and using 28-day compressive strength, flexural strength, and crack occurrence rate as output targets, and constructing the model using support vector machine (SVM); optimizing the kernel function parameters of SVM through grid search method to ensure prediction accuracy.

[0019] Preferred: The optimization calculation of maintenance parameters may include: using a multi-objective optimization algorithm, with "quality compliance and minimum energy consumption" as the optimization objectives, and the operating parameter range of the maintenance equipment as the constraint condition, to calculate the optimal combination of maintenance parameters; the optimized parameters include spraying frequency, spraying duration, insulation layer temperature, ventilation intensity, etc.

[0020] Preferred: An intelligent decision-making method for cement production and curing processes further includes the following steps: S7. Construct a prediction result deviation curve based on the labeled prediction results, and calculate the curing deviation coefficient for each prediction result of cement curing.

[0021] S8. Determine whether the maintenance deviation coefficient of each prediction result is greater than a pre-set standard deviation coefficient. If it is, mark the prediction result; otherwise, do not mark it.

[0022] Preferably, the method for constructing the prediction result deviation curve includes: constructing a plane coordinate system, wherein the horizontal axis of the plane coordinate system is time and the vertical axis is the prediction result deviation value; then, the prediction result deviation value, which is the difference between the prediction result and the maintenance target threshold, is inserted into the coordinate system according to the time points; and then, the points are connected in time order to form the prediction result deviation curve.

[0023] Preferred: Maintenance deviation coefficients for each prediction result Where t is the horizontal axis value, T is the end time of the current maintenance stage or all maintenance stages; F is the prediction result deviation value, f is the unit deviation value, and W is the prediction result deviation value. t These are the weights of the prediction results at each time stage.

[0024] Preferred: Unit deviation value , where F m The maximum deviation value of the prediction result in the historical data is α, where α is the preset level order.

[0025] The technical effects and advantages of this invention are as follows: Improving curing precision and quality stability: Through multi-source data acquisition and staged prediction models, the cement hydration process can be accurately perceived, curing parameters can be dynamically optimized, and problems of insufficient or excessive curing can be effectively avoided. Experimental verification shows that after adopting this method, the 28-day compressive strength compliance rate of cement products is significantly improved, and the crack incidence rate is significantly reduced.

[0026] Real-time response and intelligent control: Sensor networks collect data in real time, and combined with predictive models and optimization algorithms, decision instructions can be generated quickly when environmental parameters change abruptly, reducing the response time to less than 10 minutes, thus solving the problem of lag in traditional manual control.

[0027] Reduce production energy consumption and resource waste: With minimizing energy consumption as one of the optimization goals, precise control of parameters such as spraying and heat preservation can reduce water consumption and heat energy consumption, which is in line with the concept of green production.

[0028] Improve data utilization and experience reusability: By integrating historical data and learning models online, scattered maintenance experience is transformed into reusable algorithm models, enabling standardized and regulated control of different batches of maintenance processes and improving the enterprise's production management level. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating Embodiment 1 of the intelligent decision-making method for cement production and maintenance processes proposed in this invention.

[0030] Figure 2 This is a flowchart illustrating the method for obtaining the prediction model for the initial hydration stage in Embodiment 1 of the intelligent decision-making method for the cement production and maintenance process proposed in this invention.

[0031] Figure 3 This is a flowchart illustrating the method for obtaining the strength growth stage prediction model in Embodiment 1 of the intelligent decision-making method for the cement production and maintenance process proposed in this invention.

[0032] Figure 4 This is a flowchart illustrating the method for obtaining the strength stabilization stage prediction model in Embodiment 1 of the intelligent decision-making method for the cement production and maintenance process proposed in this invention.

[0033] Figure 5 This is a flowchart illustrating Embodiment 2 of the intelligent decision-making method for cement production and maintenance processes proposed in this invention. Detailed Implementation

[0034] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the invention, and should not be construed as limiting the invention. Rather, embodiments of the invention include all variations, modifications, and equivalents falling within the spirit and scope of the appended claims.

[0035] Example 1 refer to Figure 1 This embodiment proposes an intelligent decision-making method for the cement production and curing process. Through multi-source data fusion analysis, staged model construction, and real-time dynamic optimization, it achieves accurate decision-making for curing parameters. The intelligent decision-making method for the cement production and curing process includes the following steps: S1. Collect multi-dimensional parameter data during the cement curing process and preprocess it. Multi-dimensional parameter data can be obtained through real-time monitoring via a sensor network deployed in the cement production and curing area. This data can include basic parameters, environmental parameters, state parameters, and historical data. Basic parameters can include: cement raw material ratio (clinker, gypsum, and admixture proportions), initial cement temperature, and product molding process parameters (vibration time, molding pressure). Environmental parameters include: real-time temperature, relative humidity, wind speed, and atmospheric pressure in the curing area, collected by deploying temperature sensors, humidity sensors, wind speed sensors, and pressure sensors, with a collection frequency of 5-10 minutes per instance. State parameters can include: surface temperature, internal core temperature, and surface humidity of the cement product, collected by embedding contact temperature sensors, non-contact infrared temperature sensors, and humidity sensors at different locations on the cement product, with a collection frequency of 3-5 minutes per instance. Historical data can include multi-dimensional parameter data (raw material ratio, environmental parameters, state parameters), curing effect data (28-day compressive strength, flexural strength, crack incidence rate), and anomaly handling records for historical curing batches. This data needs to be extracted and organized from the enterprise's production management system. For example, this embodiment uses the curing process of commercial concrete cement products from a cement plant as an application scenario. The cement products are C30 concrete components, and the curing period is 28 days. The curing area is a closed curing workshop equipped with a spray system (maximum flow rate 5m³ / h). 3 / h), intelligent insulation layer (temperature adjustment range 10-30℃), ventilation fan (wind speed adjustment range 0.5-2m / s). Three sets of temperature-humidity sensors (model: SHT30) are evenly deployed in the curing workshop to collect ambient temperature and relative humidity; five contact temperature sensors (model: PT100) are embedded on the surface of the concrete components, and one infrared temperature sensor (model: MLX90614) is embedded in the center of the component to collect surface and core temperatures. One wind speed sensor (model: FS3000) is deployed to collect wind speed in the workshop. Data on raw material proportions, molding processes, and curing effects of nearly 100 batches of C30 concrete are extracted from the enterprise's MES system as historical datasets. The preprocessing methods for the multi-dimensional parameter data may include: data cleaning (removing outliers and filling missing values), data standardization (converting parameters of different dimensions into standardized data in the [0,1] interval), and data fusion (using a weighted average method to fuse multi-sensor data of the same monitoring indicator to improve data reliability). For example, in actual production, the Python Pandas library is used to preprocess the collected data. Outliers from temperature sensors (such as data exceeding -10℃ or 50℃) can be removed using the 3σ criterion. Missing data can be filled using linear interpolation. Min-Max normalization can convert all parameters to the [0, 1] interval. A weighted average method can be used to fuse temperature data from three environmental sensors (each with a weight of 1 / 3) to improve data reliability. Of course, this is just a simple example and may not be universally applicable; other cases will not be elaborated here.

[0036] S2. Divide the cement curing process into stages and construct a staged prediction model based on pre-processed multi-dimensional parameter data and corresponding curing effect data. The multi-dimensional parameter data here refers to historical data, i.e., multi-dimensional parameter data for which curing effect data has already been obtained. According to the cement hydration reaction mechanism, the curing process can be divided into three core stages, which, in chronological order, are: initial hydration stage (0-72h), strength growth stage (72h-14d), and strength stabilization stage (14d-28d). The required curing parameters differ for each stage's hydration characteristics, therefore, it is necessary to construct corresponding curing effect prediction models.

[0037] refer to Figure 2The initial hydration stage prediction model uses basic parameters from historical data, ambient temperature, and surface / core temperature difference of the product as input features, and initial setting time and final setting time of cement as output targets. A backpropagation (BP) neural network is used to construct the model. During model training, relevant data from the initial hydration stage in historical data are used as the training set, and the model parameters are optimized using gradient descent to keep the prediction error within 5%. For example, the input features are clinker ratio, gypsum ratio, initial temperature, ambient temperature, and surface-core temperature difference; the output targets are initial setting time and final setting time. Sixty batches of historical data are selected as the training set, 20 batches as the validation set, and 20 batches as the test set. A BP neural network is constructed using the TensorFlow framework, with 6 neurons in the input layer, 2 hidden layers (12 and 8 neurons respectively), and 2 neurons in the output layer. The model is trained using the Adam optimizer with 500 iterations and a learning rate of 0.001. The final prediction error for the initial setting time on the test set is 3.2%, and the prediction error for the final setting time is 2.8%.

[0038] refer to Figure 3 The strength growth stage prediction model uses the initial hydration stage prediction results, real-time environmental parameters, and product state parameters as input features, and the 7-day and 14-day compressive strength of cement products as output targets. A random forest algorithm is used to construct the model; cross-validation is introduced to improve the model's generalization ability. For example, the input features are the predicted initial setting time, ambient temperature, relative humidity, surface temperature, and spraying frequency, and the output targets are the 7-day and 14-day compressive strength. 70 batches of historical data are selected as the training set, and 30 batches as the test set. A random forest model is constructed using the Scikit-learn library, with 100 decision trees and a maximum depth of 10. The model's 7-day strength prediction error on the test set is 4.1%, and the 14-day strength prediction error is 3.5%.

[0039] refer to Figure 4 The strength stabilization stage prediction model uses the predicted results of the strength growth stage and the trend of subsequent environmental parameter changes as input features, and the 28-day compressive strength, flexural strength, and crack incidence rate as output targets. A Support Vector Machine (SVM) is used to construct the model. The kernel function parameters of the SVM are optimized using a grid search method to ensure prediction accuracy. For example, the input features are the 14-day strength prediction value, the rate of change of ambient temperature, relative humidity, and insulation duration, and the output targets are the 28-day compressive strength, flexural strength, and crack incidence rate. The kernel function parameters of the SVM are optimized using a grid search method (RBF kernel, penalty coefficient C=10, gamma=0.1). The 28-day strength prediction error on the model test set is 2.5%, and the crack incidence rate prediction accuracy is 92%.

[0040] S3. Based on the real-time acquisition of multi-dimensional parameter data and the linked analysis of the prediction model, the prediction results are obtained. The pre-processed multi-dimensional parameter data obtained from real-time monitoring is input into the corresponding stage's curing effect prediction model to obtain the effect prediction results under the current curing state (such as initial setting time and 7-day strength prediction value). For example, at the 36th hour of curing (initial hydration stage), the real-time monitoring data are: ambient temperature 28℃, relative humidity 78%, surface-core temperature difference 6.2℃; inputting the data into the initial hydration stage BP model predicts an initial setting time of 12.5 hours.

[0041] S4. Then, compare the predicted results with the curing target thresholds for that stage to determine whether the predicted results exceed the curing target threshold range for that stage. If not, maintain the current curing parameters; if yes, mark the predicted results. For example, the predicted initial setting time is 12.5h (target threshold 10-14h, which meets the requirements), but the temperature difference and humidity both exceed the target thresholds; then we mark the multi-dimensional parameter data that cause the temperature difference and humidity. The determination of the curing target thresholds for each stage can be based on the quality standards of cement products and combined with historical best curing batch data to set the curing target thresholds for each stage, including the product temperature difference threshold (≤5℃) and humidity threshold (≥85%) for the initial hydration stage; the ambient temperature threshold (15-25℃) and spraying frequency threshold (1-2 times / hour) for the strength growth stage; and the humidity threshold (≥70%) and heat preservation time threshold (≥10d) for the strength stabilization stage. For example, in the initial hydration stage: surface-core temperature difference ≤ 5℃, ambient humidity ≥ 85%; in the strength growth stage: ambient temperature 15-25℃, spraying frequency 1-2 times / hour, 7-day strength ≥ 25MPa; in the strength stabilization stage: ambient humidity ≥ 70%, heat preservation time ≥ 10 days, 28-day strength ≥ 30MPa, crack incidence ≤ 5%. This is just a simple example and may not be universally applicable; other situations will not be elaborated upon here. The above method involves comparing multi-dimensional parameter data one by one. This comparison is targeted, facilitating the identification of corresponding multi-dimensional parameter data for maintenance, thereby enabling targeted adjustments to each multi-dimensional parameter data.

[0042] S5. The optimal parameter combination is obtained by optimizing the predicted results of the markers through maintenance parameter optimization calculations. The maintenance parameter optimization calculation can employ a multi-objective optimization algorithm (NSGA-III), with "quality meeting standards (predicted results meeting thresholds) and minimum energy consumption (minimum spray water volume and insulation energy consumption)" as the optimization objectives, and the operating parameter ranges of the maintenance equipment (such as the spray pump flow rate range and the insulation layer temperature adjustment range) as constraints, to calculate the optimal maintenance parameter combination. Optimized parameters include spray frequency, spray duration, insulation layer temperature, and ventilation intensity. For example, using the NSGA-III algorithm, with objectives of "temperature difference ≤ 5℃, humidity ≥ 85%, and minimum spray water volume," and constraints of a spray pump flow rate of 0-5 m³ / h...3 / h, insulation layer temperature 10-30℃; the optimal parameter combination was calculated: spraying frequency increased to 2 times / hour, spraying duration 10 minutes / time, and insulation layer temperature adjusted to 22℃.

[0043] S6. Adjust equipment maintenance parameters according to the optimal parameter combination. If maintenance is carried out through automatic control equipment, the maintenance parameter command with the optimal parameter combination can be sent to the maintenance execution equipment (spraying system, insulation system, ventilation system) to achieve automatic parameter adjustment. If maintenance is carried out through manual control equipment, a maintenance parameter list with the optimal parameter combination can be exported, and each piece of equipment can be adjusted manually according to the maintenance parameter list for subsequent maintenance. During maintenance, multi-dimensional parameter data after parameter adjustment is collected in real time and fed back to the maintenance effect prediction model to dynamically update the model parameters (using an online learning algorithm, updating the model training set after each batch of maintenance) to improve the accuracy of subsequent decisions. For example, the optimized parameter command is sent to the spraying system and insulation system, and the equipment automatically adjusts its operating status; after 1 hour, real-time data collection shows: ambient humidity 86%, surface-core temperature difference 4.8℃, both meeting the target threshold; the real-time data of this batch and the maintenance effect data (final 28-day strength 32.5MPa, no cracks) are added to the model training set to update the prediction model parameters at each stage. In addition, an intelligent decision-making and monitoring platform is built to display maintenance data, prediction results, decision parameters and equipment operating status in real time, and supports manual intervention. When sudden situations such as sensor failure or equipment abnormality occur, the platform will automatically issue an alarm and switch to preset emergency maintenance parameters to ensure the continuity of the maintenance process.

[0044] Example 2 refer to Figure 5S7. Construct a prediction result deviation curve based on the labeled prediction results and calculate the curing deviation coefficient for each prediction result of cement curing. We statistically analyze the prediction results mentioned above, including their occurrence time, end time, duration, and prediction result. For example, if we collect multi-dimensional parameter data every 10 minutes, we can obtain the data by the number of consecutive occurrences of the prediction result. If it occurs twice consecutively, the duration is 20 minutes. Of course, other methods of obtaining this data are not excluded, but will not be elaborated here. Then, a prediction result deviation curve is constructed for each prediction result. The method for constructing the prediction result deviation curve may include: constructing a plane coordinate system, where the horizontal axis of the plane coordinate system is time. The calculation can be performed by statistically analyzing time in stages or by calculating the entire curing process uniformly, but will not be elaborated here. The vertical axis can represent the prediction deviation value, which is the difference between the predicted result and the maintenance target threshold. When the maintenance target threshold is a range, this value represents the difference between the extreme values ​​of the predicted result that deviate from the range. For example, if the temperature range is 20-30℃ and the predicted result is 32℃, then the prediction deviation value is the difference between 32℃ and 20℃. When the predicted result is within the maintenance target threshold range, the prediction deviation value is 0. Of course, this is just a simple example and may not be universally applicable. Then, the maintenance deviation coefficient for each prediction result can be calculated. Where t is the horizontal axis value, T is the end time of the current maintenance stage or all maintenance stages; F is the prediction result deviation value, and f is the unit deviation value, which can be set manually or obtained through calculation. Where Fm is the maximum deviation value of the prediction result in historical data, α is the preset level order, the value of which can be set manually, generally between 5 and 10, which will not be elaborated here. W t This refers to the weights of the prediction results at each time stage. The weights of the prediction results vary at different times, and their values ​​can be obtained from the prediction model or experience. The specific values ​​can be between 1 and 10, which will not be elaborated here. The maintenance deviation coefficient calculated by this method can quickly separate out the stages with excessive deviations and critical stages based on the weights of the prediction results at each time stage and the deviation values ​​of the prediction results, avoiding data distortion and increasing the accuracy of the calculation, which is convenient for subsequent calculations.

[0045] S8. Determine if the curing deviation coefficient of each predicted result is greater than a pre-set standard deviation coefficient. If yes, mark the predicted result; otherwise, do not mark it. After marking the predicted result, a pre-set predicted result-inspection item information table can be found to obtain the inspection items. Then, during the curing stage or after all curing stages, the cement curing item is inspected to determine if the quality of the cement curing item is qualified. This ensures the quality of the cement curing item and avoids huge consequences in later use. The pre-set standard deviation coefficient can be set according to actual needs. An original standard factor can be given, and its value can be set according to actual needs, generally between 2 and 20, but other values ​​are not excluded. When no quality problems are found after marking, the standard factor can be reduced. If no quality problems are found in the corresponding predicted result but no marking is performed, the standard factor can be slightly increased, and the curing deviation coefficient calculated from the corresponding predicted result can be used as the standard deviation coefficient. Specific details are not elaborated here. The pre-set predicted result-inspection item information table can be obtained through experience. For example, temperature and humidity data deviations can cause cracks, which can be addressed by visually observing the cement curing items.

[0046] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this invention can be achieved, and this is not limited herein.

[0047] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. An intelligent decision-making method for cement production and curing processes, characterized in that, The intelligent decision-making method for the cement production and curing process includes the following steps: S1. Collect multi-dimensional parameter data during the cement curing process and preprocess it; S2. Divide the cement curing stages and construct a staged prediction model based on the pre-processed multi-dimensional parameter data and curing effect data. S3. Based on the real-time multi-dimensional parameter data and the prediction model, the prediction results are obtained through linkage analysis. S4. Determine whether the prediction result exceeds the maintenance target threshold range for this stage: If not, maintain the current maintenance parameters; If so, then label the prediction result; S5. The optimal parameter combination is obtained by optimizing the maintenance parameters based on the predicted results of the markers. S6. Adjust the equipment maintenance parameters according to the optimal parameter combination.

2. The intelligent decision-making method for cement production and curing processes according to claim 1, characterized in that, Multidimensional parameter data includes basic parameters, environmental parameters, and status parameters.

3. The intelligent decision-making method for cement production and curing processes according to claim 2, characterized in that, The basic parameters include: cement raw material ratio, initial cement temperature, and product molding process parameters.

4. The intelligent decision-making method for cement production and curing processes according to claim 2, characterized in that, Environmental parameters include: real-time temperature, relative humidity, wind speed, and atmospheric pressure of the maintenance area.

5. The intelligent decision-making method for cement production and curing processes according to claim 2, characterized in that, The state parameters include: surface temperature of cement products, internal core temperature, and surface humidity.

6. The intelligent decision-making method for cement production and curing processes according to claim 1, characterized in that, The prediction model is divided into three core stages: the initial hydration stage, the intensity growth stage, and the intensity stabilization stage.

7. The intelligent decision-making method for cement production and curing processes according to claim 6, characterized in that, The method for obtaining the prediction model for the initial hydration stage includes: using basic parameters, ambient temperature, and surface / core temperature difference of the product from historical data as input features, and the initial setting time and final setting time of cement as output targets, and constructing the model using a BP neural network; during the model training process, using relevant data of the initial hydration stage from historical data as the training set, optimizing the model parameters through gradient descent, and controlling the prediction error within 5%.

8. The intelligent decision-making method for cement production and curing processes according to claim 6, characterized in that, The method for obtaining the strength growth stage prediction model includes: using the initial hydration stage prediction results, real-time environmental parameters, and product state parameters as input features, and the 7-day compressive strength and 14-day compressive strength of cement products as output targets, a random forest algorithm is used to construct the prediction model; and the generalization ability of the prediction model is improved by introducing a cross-validation mechanism.

9. The intelligent decision-making method for cement production and curing processes according to claim 6, characterized in that, The method for obtaining the strength stabilization stage prediction model includes: using the strength growth stage prediction results and the trend of subsequent environmental parameter changes as input features, and using 28-day compressive strength, flexural strength, and crack occurrence rate as output targets, and constructing the model using support vector machine; optimizing the kernel function parameters of SVM through grid search method to ensure prediction accuracy.

10. The intelligent decision-making method for cement production and curing processes according to claim 1, characterized in that, The intelligent decision-making method for the cement production and curing process also includes the following steps: S7. Construct a prediction result deviation curve based on the labeled prediction results, and calculate the curing deviation coefficient for each prediction result of cement curing. S8. Determine whether the maintenance deviation coefficient of each prediction result is greater than the standard deviation coefficient. If it is, mark the prediction result; otherwise, do not mark it.

Images