Intelligent manufacturing method and system for foaming cement composite board

Abstract

The invention relates to the technical field of intelligent manufacturing, and discloses an intelligent manufacturing method and system for a foamed cement composite board. The method comprises the steps that a foam quality analysis model is adopted to analyze a foam solution, and a foam analysis result is obtained; according to the foam analysis result and the slurry viscosity value, an optimal mixing parameter set of the foam solution and the cement slurry base material is obtained, a mixed slurry quality analysis model is adopted to analyze the foamed cement mixed slurry, a mixed slurry quality analysis result is obtained, and therefore pouring parameters of the foamed cement mixed slurry are controlled. And according to the pouring amount data, the slab state data and the slab space-time reference data, a virtual maintenance model is constructed in the intelligent manufacturing digital twin platform to generate an optimal temperature and humidity maintenance curve, and according to the optimal temperature and humidity maintenance curve, the temperature and humidity set value of the maintenance kiln is dynamically adjusted. According to the method provided by the invention, the self-adaptability and the intelligent level of the production process of the foamed cement composite board are improved, and the product quality is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology, and in particular to an intelligent manufacturing method and system for foamed cement composite panels. Background Technology

[0002] Foamed cement composite panels, as a lightweight building material with excellent thermal insulation and fire resistance, are widely used in prefabricated buildings and energy-saving projects. The core of its production process lies in introducing uniform and stable air bubbles into cement-based slurry through physical or chemical methods, forming a lightweight material with a closed-cell or porous structure.

[0003] In the traditional process of producing foamed cement composite boards, the quality assessment of the foam solution relies heavily on visual observation and experience-based judgment by workers. There is a lack of quantitative monitoring methods for foam stability, pore size distribution, and uniformity, leading to unstable foaming quality and directly affecting the strength and thermal insulation performance of the final board. During the mixing stage, the mixing parameters of the foam solution and cement slurry are often based on fixed formulas, making adaptive adjustments impossible. This easily results in problems such as foam breakage, slurry segregation, or uneven mixing, causing internal defects in the board.

[0004] The curing process is a crucial step in determining the degree of hydration and final performance of foamed cement. Currently, batch curing is commonly carried out using curing kilns with fixed temperature and humidity curves, ignoring the dynamic influence of different formulations, environments, and green body conditions on the hydration process. This "one-size-fits-all" curing model makes it difficult to achieve optimal coupling of hydration heat and humidity diffusion, which can easily lead to shrinkage cracking of the boards, insufficient strength development, or energy waste.

[0005] Therefore, improving the adaptability and intelligence of the foamed cement composite board production process, as well as improving product quality, have become technical problems that urgently need to be solved by those skilled in the art. Summary of the Invention

[0006] This invention provides an intelligent manufacturing method and system for foamed cement composite boards, which addresses the technical issues of how to improve the adaptability and intelligence level of the foamed cement composite board production process and how to improve product quality.

[0007] In a first aspect, the present invention provides an intelligent manufacturing method for foamed cement composite panels, the method comprising: collecting foam flow image data and foam density data corresponding to foam solution in the manufacturing of foamed cement composite panels, and using a pre-constructed foam quality analysis model to analyze the foam flow image data and the foam density data to obtain foam analysis results;

[0008] Based on the foam analysis results and the viscosity value of the cement slurry base, the optimal mixing parameter set of the foam solution and the cement slurry base is obtained. Then, using a pre-constructed mixed slurry quality analysis model, the image data and wet density data of the foamed cement mixed slurry after the mixing operation based on the optimal mixing parameter set are analyzed to obtain the mixed slurry quality analysis results.

[0009] Based on the quality analysis results of the mixed slurry, the pouring parameters of the foamed cement mixed slurry are controlled, and a virtual curing model is constructed in the pre-constructed intelligent manufacturing digital twin platform based on the synchronously collected pouring volume data, the state data of the poured slab, and the spatiotemporal reference data of the slab. The intelligent manufacturing digital twin platform is set to simulate the hydration heat field coupling relationship during the curing process of the slab.

[0010] Based on the virtual curing model, an optimal temperature and humidity curing curve is generated in the intelligent manufacturing digital twin platform, and the temperature and humidity setpoints of the curing kiln are dynamically adjusted according to the optimal temperature and humidity curing curve.

[0011] The multi-axis cutting machine is controlled to cut the cured slab and assign a unique identification code to the cut finished slab.

[0012] Preferably, the construction of the intelligent manufacturing digital twin platform includes:

[0013] Based on the multi-physics coupling and reaction path theory of foamed cement composite board manufacturing, a physical theoretical model is established to map the performance to the production formula. The physical theoretical model includes: strength model, thermal conductivity model and density model.

[0014] Obtain historical production formulas and corresponding historical board performance indicators, and input the historical production formulas into the physical theory model to obtain theoretical board performance indicators. Based on the historical board performance indicators and the theoretical board performance indicators, obtain the board performance indicator residuals.

[0015] Based on the historical production formula and the residual performance index of the board, a training dataset is constructed, and the selected deep neural network is trained using the training dataset to obtain the index residual prediction model. The physical theory model and the index residual prediction model are combined to obtain the board performance index prediction model.

[0016] Based on the coupling relationship between temperature, humidity and hydration of the foamed cement composite board, a hydration thermal field simulation engine for the board is constructed. The hydration thermal field simulation engine includes: temperature field control equation, humidity field equation, hydration process equation, coupling relationship equation and kiln environment model.

[0017] The performance index prediction model of the board material and the hydration thermal field simulation engine of the board material are imported into the digital twin platform to obtain the intelligent manufacturing digital twin platform.

[0018] Preferably, the method further includes:

[0019] Based on the production order, an initial production formula is generated, and the predicted board performance index is obtained using the board performance index prediction model.

[0020] The loss value between the predicted board performance index and the target board performance index in the production order is calculated, and based on the loss value, the initial production formula is iteratively optimized using an optimization algorithm until the optimal production formula is found. The optimal production formula is used to control the proportioning of the foam solution and the cement slurry base material by the IoT weighing system.

[0021] Preferably, the step of using a pre-constructed foam quality analysis model to analyze the foam flow image data and the foam density data to obtain foam analysis results includes:

[0022] A first convolutional neural network is used to extract texture features, bubble morphology features, and bubble stability features from the foam flow image data to obtain a first high-dimensional image feature vector.

[0023] A first dual-head attention mechanism is used to fuse the first high-dimensional image feature vector and the foam density data to obtain a first fused feature. A first fully connected neural network is then used to analyze the first fused feature to obtain foam quality analysis results. The foam quality analysis results include: foam quality grade, foam performance quantitative index, root cause prediction results, and foaming process adjustment suggestions.

[0024] Preferably, the step of obtaining the optimal set of mixing parameters for the foam solution and the cement slurry base based on the foam analysis results and the slurry viscosity value includes:

[0025] Based on the viscosity value of the cement slurry base material and the preset viscosity threshold range, a threshold judgment is made on the viscosity value of the slurry to obtain the quality judgment result of the cement slurry base material;

[0026] Based on the foam analysis results, the slurry viscosity value, and the quality judgment results, a dynamic optimization model pre-built based on a machine learning algorithm is used to obtain the optimal set of mixing parameters for the foam solution and the cement slurry base. The dynamic optimization model includes logic gates, an attention mechanism layer, a physical information layer, and a multi-task output layer. The logic gates perform conditional judgments on the foam quality level and the quality judgment results, and output logical judgment results. The attention mechanism layer performs feature fusion on the slurry viscosity value and the foam performance quantification index based on the logical judgment results. The physical information layer includes a differentiable physical calculation unit and an expert rule residual connection. The differentiable physical calculation unit embeds physical constraints, and the expert rule residual connection embeds expert rules.

[0027] Preferably, the pre-constructed mixed slurry quality analysis model is used to analyze the mixed slurry image data and wet density data corresponding to the foamed cement mixed slurry after mixing based on the optimal mixing parameter set, to obtain the mixed slurry quality analysis results, including:

[0028] A second convolutional neural network is used to extract macroscopic uniformity features, surface bubble morphology features, and rheological visual features from the mixed slurry image data to obtain a second high-dimensional image feature vector;

[0029] A second dual-head attention mechanism is used to fuse the second high-dimensional image feature vector and the wet density data to obtain a second fused feature. A second fully connected neural network is then used to analyze the second fused feature to obtain the mixed slurry quality analysis results. The mixed slurry quality analysis results include: mixed slurry quality grade, mixed slurry performance quantification index, and pouring recommendations.

[0030] Preferably, the step of constructing a virtual curing model in a pre-built intelligent manufacturing digital twin platform based on synchronously collected pouring volume data, poured slab state data, and slab spatiotemporal reference data includes:

[0031] Based on the size data of the production order, a three-dimensional geometric model that is completely consistent with the physical slab is generated through the intelligent manufacturing digital twin platform. The synchronously collected casting volume data, the slab status data after casting, and the slab spatiotemporal reference data are used as the initial physical attribute values ​​of the three-dimensional geometric model to obtain a virtual slab.

[0032] The virtual slab curing process is simulated using the pre-built hydration thermal field simulation engine of the digital twin platform to obtain a virtual curing model.

[0033] Preferably, the step of dynamically adjusting the temperature and humidity setpoints of the curing kiln according to the optimal temperature and humidity curing curve includes:

[0034] The current measured temperature and humidity values ​​of the curing kiln are obtained by the distributed temperature and humidity sensors of the curing kiln. Based on the current measured temperature and humidity values ​​and the optimal temperature and humidity curing curve, the temperature and humidity setpoints of the curing kiln in the predicted time domain are optimized. The objective function of the optimization solution includes: temperature setpoint error term, humidity setpoint error term, and setpoint change smoothing term.

[0035] Preferably, the control of the multi-axis cutting machine to cut the cured slab includes:

[0036] Based on the obtained 3D dimensional data of the slab and the specifications of the finished slab, an optimal layout diagram is generated using a pre-constructed layout optimization model. The layout optimization model aims to maximize material utilization.

[0037] Based on the optimal layout diagram, the optimal cutting task sequence is obtained by using a pre-built path planning model, wherein the path planning model aims to minimize the total idle time.

[0038] Based on the optimal cutting task sequence, the optimal cutting command is obtained by using a pre-built motion trajectory optimization model. The optimal cutting command includes cutting position, cutting speed and cutting acceleration.

[0039] The multi-axis cutting machine is controlled to cut the cured slab according to the optimal cutting command to obtain the finished slab.

[0040] Secondly, the present invention also provides an intelligent manufacturing system for foamed cement composite boards, used to implement the intelligent manufacturing method for foamed cement composite boards described above. The system includes: a foaming monitoring and analysis module, a mixing monitoring and analysis module, a virtual curing model construction module, a curing control module, and a cutting control module.

[0041] The foaming monitoring and analysis module is used to collect foam flow image data and foam density data corresponding to the foam solution in the manufacturing of foamed cement composite boards, and to analyze the foam flow image data and foam density data using a pre-built foam quality analysis model to obtain foam analysis results.

[0042] The mixing monitoring and analysis module is used to obtain the optimal mixing parameter set of the foam solution and the cement slurry base material based on the foam analysis results and the slurry viscosity value of the cement slurry base material, and to analyze the image data and wet density data of the foamed cement slurry after mixing based on the optimal mixing parameter set using a pre-constructed mixing slurry quality analysis model, so as to obtain the mixing slurry quality analysis results.

[0043] The virtual curing model construction module is used to control the pouring parameters of the foamed cement slurry based on the quality analysis results of the mixed slurry, and to construct a virtual curing model in a pre-constructed intelligent manufacturing digital twin platform based on the synchronously collected pouring volume data, the state data of the poured slab, and the spatiotemporal reference data of the slab. The intelligent manufacturing digital twin platform is set to simulate the hydration heat field coupling relationship during the curing process of the slab.

[0044] The curing control module is used to generate an optimal temperature and humidity curing curve in the intelligent manufacturing digital twin platform based on the virtual curing model, and dynamically adjust the temperature and humidity setpoints of the curing kiln based on the optimal temperature and humidity curing curve.

[0045] The cutting control module is used to control the multi-axis cutting machine to cut the cured slab and assign a unique identification code to the cut finished slab.

[0046] This invention provides an intelligent manufacturing method and system for foamed cement composite boards. Compared with the prior art, the embodiments of this invention have the following beneficial effects:

[0047] The intelligent manufacturing method for foamed cement composite boards provided in this application employs a foam quality analysis model, breaking through the limitations of traditional methods that only focus on density or simple morphology. Through deep learning, it automatically extracts deep features such as texture, morphology, and stability, and combines this with density data to achieve multi-dimensional comprehensive evaluation. The provided foam quality analysis results not only indicate the foam quality level but also clearly point out the root causes and provide process adjustment suggestions, directly elevating quality inspection to decision support and process control recommendations, effectively supporting closed-loop production optimization and predictive maintenance. Based on the foam analysis results and the viscosity value of the cement slurry base, a dynamic optimization model is used to achieve adaptive control of the mixing process between the foam solution and the cement slurry base. A hybrid intelligent optimization model integrating physical laws, expert experience, and data-driven approaches is constructed, overcoming the limitations of traditional empirical formulas and the risk that pure data models may output results that violate physical laws. For the board curing process, an intelligent manufacturing digital twin model is constructed, changing the current situation where curing processes are estimated based on experience, achieving precise, differentiated, and dynamic curing control, and greatly improving the initiative and efficiency of quality management. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the steps of an intelligent manufacturing method for foamed cement composite panels provided in a preferred embodiment of the present invention;

[0049] Figure 2 This is a schematic diagram of the structure of an intelligent manufacturing system for foamed cement composite panels provided in a preferred embodiment of the present invention;

[0050] Figure Labels

[0051] 1-Fogging monitoring and analysis module, 2-Hybrid monitoring and analysis module, 3-Virtual curing model construction module, 4-Cure control module, 5-Cutting control module. Detailed Implementation

[0052] The embodiments of the present invention are described in detail below with reference to the accompanying drawings. The embodiments are provided for illustrative purposes only and should not be construed as limiting the scope of the invention. The accompanying drawings are for reference and illustration only and do not constitute a limitation on the scope of protection of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of this invention.

[0053] In the description of this invention, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0054] Please see Figure 1 In an embodiment of the present invention, a smart manufacturing method for foamed cement composite panels is provided, the method comprising:

[0055] S1. Collect foam flow image data and foam density data corresponding to the foam solution during the manufacturing of foamed cement composite boards, and analyze the foam flow image data and foam density data using a pre-constructed foam quality analysis model to obtain foam analysis results. Foamed cement composite boards are a new type of green building material, adopting a "sandwich" structure. The core layer is foamed cement core material, which is pre-made into a foam solution by a mechanical foaming machine, mixed with cement-based slurry, and cured to form a structure filled with uniformly closed fine pores. The upper and lower layers are high-strength panels, typically made of calcium silicate board, fiber cement board, magnesium oxide board, aluminum alloy board, etc., which are firmly bonded to the foamed cement core material with adhesives. Foamed cement composite boards combine the high strength of the panels with the lightweight and thermal insulation properties of the foamed cement core material, featuring lightweight and high strength, excellent thermal insulation performance, fire resistance and non-combustibility, good sound insulation, convenient construction, and environmental friendliness. Before the production of foamed cement composite boards, production orders are received. Production orders include detailed descriptions of the products, such as product model, specifications, color, and size, as well as the specific quantity ordered. Production plans can be determined based on the detailed product descriptions in production orders. For orders of foamed cement composite boards of standard specifications, historical data can be directly used to generate the corresponding production plan. For customized non-standard models, specifications, and sizes, a targeted production plan needs to be generated. In the preferred embodiment of this application, for non-standard models, specifications, and sizes of foamed cement composite boards, a combination of physical theory models and data-driven deep learning models is used to plan the production plan. Specifically, a physical theory model is constructed on the digital twin platform. The physical theory model includes a strength model, a thermal conductivity model, and a density model. The strength model is based on composite material mechanics, hydration kinetics, and pore structure theory. The mathematical formula for the strength model is expressed as:

[0056]

[0057] in, The theoretical strength of foamed cement composite boards is one of their main performance indicators. This represents the theoretical strength of a benchmark dense cement mortar matrix under standard curing conditions, determined by the cement grade, water-cement ratio, and mineral admixtures. The hydration degree correction factor represents the contribution of the cement hydration process to the matrix strength and is a function of curing time and curing conditions. This represents the structural constant, which is influenced by the characteristics of the pore structure. Indicates effective porosity. This represents the fiber efficiency factor, which is related to fiber type and orientation. Indicates fiber volume content (%). Indicates the fiber aspect ratio. Indicates fiber length. Indicates fiber diameter. Indicates the fiber orientation factor. This represents the interfacial bonding efficiency factor.

[0058] in,

[0059]

[0060] in, Indicates the water-cement ratio. and It is an empirical constant related to cement grade and mineral admixtures.

[0061] Different pore sizes have varying effects on weakening material strength; small pores and closed pores are less harmful than large pores and interconnected pores. Therefore:

[0062]

[0063]

[0064] in, This represents the volume fraction weighting for apertures smaller than the critical aperture. Indicates the average aperture. Indicates the critical aperture. Indicates the pore size sensitivity index. Indicates the gas phase volume fraction. The theoretical dry density of foamed cement composite boards is one of their main performance indicators. This represents the theoretical dry density of cement mortar when there are no air bubbles.

[0065] The interfacial bond efficiency factor reflects the bonding quality between the fiber and the cement matrix, and is affected by the fiber surface treatment and the matrix consistency, as shown below:

[0066]

[0067] in, Indicates the base bonding coefficient. This represents the adhesion sensitivity coefficient.

[0068] The thermal conductivity model is based on the multiphase heat transfer theory. Foamed cement composite boards are a three-phase composite consisting of solid, liquid, and gas phases. The mathematical formula for the thermal conductivity model is as follows:

[0069]

[0070]

[0071]

[0072] in, This indicates the overall thermal resistance of the foamed cement composite board. Indicates panel thickness. Indicates the thermal conductivity of the panel. Indicates the thickness of the foamed cement core material. Indicates the thermal conductivity of foamed cement core material. The thermal conductivity coefficient represents the overall theoretical thermal conductivity of foamed cement composite boards and is one of the main performance indicators of foamed cement composite boards. Indicates the thermal conductivity of a solid matrix. This indicates the structural factor of foamed cement core material. This represents a dimensionless correction factor between 0 and 1. This indicates the volumetric moisture content of the foamed cement core material. The thermal conductivity of water is 0.6 W / (m·K). The thermal conductivity of air is 0.026 W / (m·K).

[0073] The mathematical formula for the density model is expressed as follows:

[0074]

[0075] in, Indicates the dry-to-wet density conversion factor. Indicates cement quality. Indicates the quality of the mixing water. This indicates the mass of the foaming agent solution, including the foaming agent and the solvent water. This indicates the total mass of other solid additives. This indicates the apparent density of cement. This indicates the density of water. This indicates the density of the foaming agent solution. Indicates the density of other solid additives. This indicates the foaming ratio of the foaming agent solution. This indicates the foam volume conversion rate.

[0076] In a preferred embodiment of this application, to achieve accurate predictions from the model, the internal parameters (such as hydration correction factors, structural constants, fiber efficiency factors, pore size sensitivity indices, etc.) in the strength model, thermal conductivity model, and density model are calibrated based on historical data using a dynamic calibration strategy. This dynamic calibration strategy is implemented in a smart manufacturing digital twin platform. Its core is the construction of a parameter calibration module. This module mines and analyzes historical high-quality production data to reverse-calibrate the optimal values ​​of each model parameter under corresponding process conditions, forming a queryable and updatable parameter knowledge base. When simulating a new task, the corresponding parameter set is automatically matched or interpolated for use by the physical theory model.

[0077] The performance indicators of sheet metal obtained from physical theoretical models are theoretical values. In actual production, deviations exist that cannot be explained by the physical theoretical model. This application employs a data-driven prediction model to predict these deviations. Specifically, historical production formulas and corresponding historical sheet metal performance indicators are obtained. These historical production formulas are input into the physical theoretical model to obtain theoretical sheet metal performance indicators. The residuals of sheet metal performance indicators are obtained based on the difference between the historical and theoretical performance indicators. A training dataset is constructed based on the historical production formulas and the residuals. This training dataset is used to train a selected deep neural network to obtain an indicator residual prediction model. The physical theoretical model and the indicator residual prediction model are combined to construct a sheet metal performance prediction model. This model provides a robust prediction baseline using the physical theoretical model, while the indicator residual prediction model learns complex nonlinear residuals that the physical model cannot explain or cannot explain accurately. This integrates the reliability of the mechanistic model with the flexibility of the data-driven model. In areas with scarce data, the physical theoretical model can provide reasonable benchmark predictions, avoiding the absurd prediction results that pure data models may produce. Physical theory, as a strong prior, greatly reduces the hypothesis space that needs to be searched, making the training process converge faster, reducing the risk of overfitting, and showing better adaptability to new formulations and processes. The index residual prediction model has a simpler and more focused learning task, thus achieving higher accuracy with less data. Fluctuations in materials and processes often affect residuals in a relatively consistent way, making it easier to generalize to scenarios with similar fluctuation patterns than learning a complete input-output mapping.

[0078] The board performance prediction model is integrated into a digital twin platform. When a production order requires the generation of a production formula, the model is used as a performance simulator. An optimization algorithm then automatically adjusts the production formula parameters, ensuring that the board performance indicators output by the simulator closely approximate the target board performance indicators of the production order. Specifically, based on the production order, an initial production formula is generated based on experience. The board performance prediction model is then used to predict the corresponding board performance indicators, and the loss value between the predicted and target indicators is calculated. An optimization algorithm is then used to generate a new formula based on the loss value, and this process is iterative until the optimal production formula is found.

[0079] The production formula solution process in this application is decoupled from the forward model, making it flexible in application and capable of directly handling complex constraints. The optimization algorithm can simultaneously handle multiple objectives and constraints, finding the optimal balance point on the Pareto frontier of performance. Each formula iteration eliminates the need for actual mixing, pouring, curing, and testing in the laboratory; it only requires calling the board performance index prediction model for millisecond-level virtual calculations. The optimization algorithm can simultaneously explore tens of thousands of candidate formulas in a vast formula space and intelligently converge towards the target board performance index, avoiding the blindness and limitations of manual trial and error. Because the optimization is based on a high-precision prediction model, the generated optimal production formula can approach the target board performance index upon initial production, greatly reducing batch scrap, rework, and raw material waste caused by unqualified formulas.

[0080] In a preferred embodiment of this application, the proportions of foam solution and cement slurry base are controlled by an IoT weighing system based on the optimal production formula. A transparent observation channel, such as an acrylic tube or a flat flow trough, is bypassed from the outlet pipe of the foaming machine to ensure that the foam flows through the transparent observation channel at a uniform speed, stably, and without turbulence. The transparent observation channel has uniform backlighting, which facilitates imaging and is easy to clean, preventing stains from affecting the imaging. An industrial camera and lighting system are arranged in the transparent observation channel area to collect foam flow image data. The industrial camera is a high-frame-rate CMOS camera, and the lighting system uses backlighting to produce high-contrast foam outline images, with bubbles appearing dark and the liquid film appearing bright, greatly simplifying subsequent image processing. Foam density data is collected using an online foam density sensor. A microwave, ultrasonic, or nuclear online density meter is selected and directly installed on the main output pipe of the foaming machine to measure the apparent density of the foam solution in real time and continuously. During the data acquisition process, a precise timestamp is added to each frame of the foam image via a PLC or industrial control computer, and the foam density value and key process parameters at that moment are recorded simultaneously.

[0081] In a preferred embodiment of this application, the acquired foam flow image data is further subjected to denoising, contrast enhancement, and binarization operations. To extract information reflecting the essential attributes of foam from the foam flow image data, a convolutional neural network, named the first convolutional neural network, is pre-trained. The terminal convolutional layer of the first convolutional neural network outputs a feature map. By performing global average pooling or extracting activation values ​​from specific layers on the feature map, a first high-dimensional image feature vector representing the state of the foam flow is automatically obtained. The first high-dimensional image feature vector contains texture features, bubble morphology features, and bubble stability features. Texture features correspond to the arrangement pattern, gloss, and overall texture roughness of the foam liquid film, and are used to indirectly evaluate the uniformity and fineness of the foam. Bubble morphology features are the neural network's abstract understanding of the shape and contour of bubbles, reflecting the average size, size distribution breadth, and sphericity of bubbles, without the need for traditional precise image segmentation. Bubble stability features capture the trend of bubble merging and bursting by analyzing short-term changes in the feature vectors of consecutive frames or encoding dynamic textures within the network, thereby evaluating the time-varying stability of the foam.

[0082] To overcome the limitations of single-modal data, this invention introduces a dual-head attention mechanism module to fuse a first high-dimensional image feature vector and foam density data. The first high-dimensional image feature vector and the real-time acquired foam density data serve as two independent input heads. The dual-head attention mechanism first calculates its own query vector, key vector, and value vector for each input head, forming an image attention head and a foam density attention head. Then, the query vector of the first high-dimensional image feature vector is used to perform a dot product similarity calculation with the key vector of the foam density feature vector, thereby obtaining a first attention weight distribution for the foam density feature. The value vector of the foam density feature is then weighted and summed using this first attention weight distribution to retrieve the information most relevant to the visual context of the first high-dimensional image feature vector from the foam density data. Simultaneously, the query vector of the foam density data is also used to perform a dot product similarity calculation with the key vector of the first high-dimensional image feature vector, thereby obtaining a second attention weight distribution for the first high-dimensional image feature vector. The value vector of the first high-dimensional image feature vector is then weighted and summed using this second attention weight distribution to retrieve the information most relevant to the context of the foam density feature from the first high-dimensional image feature vector. Through the aforementioned bidirectional cross-attention calculation, a first fusion feature is generated that integrates the core information of the first high-dimensional image feature vector and the foam density data. This first fusion feature not only retains the original important information from both the first high-dimensional image feature vector and the foam density data, but also highlights the correlated, corroborating, or contradictory parts between the two sets of data. This provides a more comprehensive characterization of the complex quality state of the foam and significantly improves the model's ability to understand the complex quality state of the foam.

[0083] In this application, a first fully connected neural network is used to analyze the first fusion features to obtain foam quality analysis results. These results include: foam quality grade, quantitative indicators of foam performance, root cause prediction results, and suggestions for adjusting the foaming process. The foam quality grade is a direct assessment of the current foam solvent quality grade, including qualified, warning, and unqualified. Quantitative indicators of foam performance include: estimated average bubble diameter, bubble size uniformity coefficient, predicted foam half-life, and foam density deviation. During training data acquisition, several experienced process experts were invited to conduct comprehensive quality scoring and grade classification for each foam image sequence. The corresponding real physical parameters were measured using offline precision instruments as training targets. Specifically, a bubble segmentation algorithm (such as the watershed algorithm) is used to identify individual bubbles, calculate the average diameter (D50), diameter distribution span (e.g., D90 / D10), and calculate the degree to which the bubbles are nearly perfectly round; foam with high roundness is more stable. Uniform and dense small bubbles are a hallmark of high-quality foam. The standard deviation of the bubble area is calculated; the smaller the standard deviation, the better the uniformity. Therefore, the standard deviation is used to characterize the bubble size uniformity coefficient. For a continuous frame image sequence, the high similarity between adjacent frames and the continuity of its evolution over time are used as natural supervision signals.

[0084] The root cause prediction result, based on the learned first fusion feature, infers possible process causes leading to the current foam quality level. For example, the output might be: "Presumed cause: High gas-liquid ratio, probability 75%" or "Presumed cause: Insufficient foaming agent solution concentration, probability 65%." This function greatly shortens the fault diagnosis time. During the training of the foam quality analysis model, the training dataset not only records the foam quality level but also has expert-labeled or process log-labeled root causes for each quality anomaly. During training, one output head of the multi-head fully connected neural network learns to predict the probability distribution of root cause labels from the current fusion features, and the output is "Presumed cause: X (probability Y%)." In a preferred embodiment of this application, a solution database is established. After the root cause is predicted, historically validated and effective adjustment solutions for similar problems are matched from the solution database, and the output is a foaming process adjustment suggestion. The foaming process adjustment recommendations are based on the root cause analysis results, providing specific and actionable suggestions for adjusting process parameters, such as "it is recommended to reduce the air intake flow rate by 5%" or "it is recommended to check and replenish the foaming agent storage tank". These recommendations can be directly fed into the production control system or provided to operators for reference and implementation, thus achieving a closed loop from analysis to control.

[0085] In a preferred embodiment of this application, the foam quality analysis results are displayed on the on-site human-machine interface and transmitted to the intelligent manufacturing control platform via an industrial network for production recording, trend analysis, and advanced process optimization. This approach overcomes the limitations of traditional methods that only focus on density or simple morphology. Through deep learning, it automatically extracts deep features such as texture, morphology, and stability, and combines them with density data to achieve a multi-dimensional comprehensive evaluation. The dual-head attention mechanism employed achieves semantic-level deep fusion of foam flow image data and foam density data, enabling the foam quality analysis model to more accurately understand the relationship between the two. The provided foam quality analysis results not only indicate the foam quality level but also clearly point out the root causes and provide process adjustment suggestions, directly elevating quality inspection to decision support and process control recommendations, effectively supporting closed-loop production optimization and predictive maintenance.

[0086] S2. Based on the foam analysis results and the viscosity value of the cement slurry base, the optimal mixing parameter set for the foam solution and the cement slurry base is obtained. A pre-constructed mixed slurry quality analysis model is used to analyze the mixed slurry image data and wet density data corresponding to the foamed cement mixed slurry after mixing based on the optimal mixing parameter set, obtaining the mixed slurry quality analysis results. If the foam quality level is unqualified, the foam solution is directly discharged through a bypass, and the machine is shut down for maintenance. If the foam quality level is qualified or under warning, further mixing with the cement slurry base can be carried out. However, for foam solutions under warning levels, the mixing parameters need to be adjusted in subsequent mixing with the cement slurry base to meet the production requirements of foamed cement composite boards. Specifically, the slurry viscosity value of the cement slurry base is acquired in real time using an online rotary viscometer or vibratory viscosity sensor installed on the cement slurry conveying pipeline. The slurry viscosity value is a key indicator characterizing the workability and mixability of the slurry. The real-time viscosity value of the slurry is compared with a preset viscosity threshold range. This threshold range is determined by production process requirements and is typically a closed interval with upper and lower limits. Based on this viscosity threshold range, the threshold judgment logic is executed:

[0087] If the viscosity value of the slurry falls within the preset viscosity threshold range, the slurry state is determined to be ideal, and the judgment result is "qualified".

[0088] If the viscosity value of the slurry is higher than the upper limit, the slurry is judged to be too thick, and the judgment result is "overly thick warning".

[0089] If the viscosity value of the slurry is lower than the lower limit, the slurry is judged to be too thin, and the judgment result is "too thin warning".

[0090] When the slurry is too thick, the cement slurry base has poor fluidity. Although it has a strong ability to encapsulate and bear the foam, it has high mixing resistance. Under mechanical stirring, the foam is more likely to be sheared and broken, leading to bubble merging, larger pore size, and decreased porosity. Ultimately, the density of the board is too high, and the thermal insulation performance is reduced. During the casting process, it is difficult to fill the mold, which can easily form holes or defects inside the board, affecting the uniformity of the structure. After molding, the foamed cement composite board may not be fully hydrated due to insufficient moisture, resulting in more internal defects and reduced strength.

[0091] When the slurry is too thin, the density difference between the slurry bubbles increases, the foam rises faster, the foam stability decreases sharply, and the bubbles are prone to escape or merge, resulting in uneven density between the upper and lower layers of the board, and even collapse of the mold. During the casting process, the foamed cement slurry has a weak ability to suspend aggregates and fibers, and is prone to stratification. After molding, the foamed cement composite board has a high water-cement ratio and a loose cement stone structure, resulting in a significant decrease in strength.

[0092] The viscosity threshold range of this application includes cases where the foamed cement slurry can meet the requirements through subsequent fine-tuning of mixing parameters, while the "overly thick warning" and "overly thin warning" cases are cases where the foamed cement slurry cannot meet the requirements through subsequent fine-tuning of mixing parameters.

[0093] Furthermore, a dynamic optimization model is constructed based on machine learning algorithms. This dynamic optimization model is not a typical black-box neural network, but a hybrid intelligent system that deeply integrates logical judgment, attention mechanisms, physical laws, and expert experience. It includes logic gates, an attention mechanism layer, a physical information layer, and a multi-task output layer. The logic gates act as pre-condition filters, receiving the foam quality level and cement slurry base material quality judgment results output by the foam quality analysis model. The logic gates embed business logic, activating and outputting a condition met only when the foam quality level is qualified or a warning is issued, and the cement slurry base material quality judgment result is qualified. This allows subsequent optimization processes to begin; otherwise, a warning is output, prompting priority to address raw material quality issues. The attention mechanism layer, after both the cement slurry base material and the foam solution meet the conditions, receives the slurry viscosity value and the foam performance quantification index output by the foam quality analysis model. By calculating the correlation weight between the slurry viscosity value and the foam performance quantification index, the attention mechanism layer dynamically focuses on the key features that have the greatest impact on the current mixing process. For example, when the slurry is too thick, it pays more attention to the foam stability index; when the foam is too light, it pays more attention to the slurry's encapsulation ability. The physical information layer is the core that ensures the model output conforms to the laws of the physical world and has engineering feasibility. It consists of two parts: differentiable physical computation units and expert rule residual connections. The differentiable physical computation units embed the physical constraints that the mixing process must follow in a differentiable programming manner. During model forward propagation, the physical constraints participate in the calculation as strong constraints. During back propagation, their gradients are used to guide the optimization of the dynamic optimization model parameters, ensuring that the learning direction of the dynamic optimization model never violates the basic physical laws. In a preferred embodiment of this application, the physical constraints include: shear force foam stability constraints, slurry encapsulation capacity foam content constraints, mixing power slurry viscosity constraints, and feeding rate mixing process coordination constraints. The expert rule residual connections encode the experience rules of process experts into the backbone network in the form of residual connections. For example, "In low-viscosity slurries, the initial mixing speed should not be too high to prevent foam rupture." The rule is transformed into a correction term for the output of the intermediate layer of the network, which is equivalent to providing a reliable experience guide for the data-driven dynamic optimization model, making up for the shortcomings of pure data-driven methods that may fail under extreme or rare conditions.The multi-task output layer receives high-level features processed by the attention-weighted and physical information layers. Through multiple parallel fully connected sub-networks, it simultaneously predicts multiple optimal mixing parameters, forming a complete set of optimal mixing parameters. The set of optimal mixing parameters includes: optimal mixing speed, optimal mixing time, and optimal foam feeding rate. The unit of optimal mixing speed is revolutions per minute, which refers to the optimal rotational speed that the agitator should maintain. The unit of optimal mixing time is seconds, which refers to the optimal duration required from the addition of foam to uniform mixing. The optimal foam feeding rate is a time-series parameter that defines the rate at which the foam solution should be added to the cementitious slurry base within the mixing time to achieve the smoothest fusion and the most uniform pore structure.

[0094] The optimal set of mixing parameters will be directly sent to the mixer control system and foam metering and feeding system of the production line, driving the actuators to carry out automated mixing operations according to the optimal mixing speed, optimal mixing time and optimal foam feeding rate. This ensures that the foamed cement slurry always achieves the most ideal mixing uniformity and bubble structure retention rate, even when the raw material status fluctuates in real time.

[0095] In the preferred embodiment of this application, the dynamic optimization model uses the qualification judgment of foam quality and slurry viscosity as an optimization premise through logic gate units. This ensures that the optimization action is initiated only when the raw material condition meets the standards, avoiding the problem of producing unqualified foamed cement composite boards under abnormal raw material conditions, and improving the robustness and safety of foamed cement composite board production. A hybrid intelligent optimization model based on physical laws, expert experience, and data-driven approaches is constructed, overcoming the limitations of traditional empirical formulas and the risk that pure data models may output results that violate physical laws. Differentiable physical calculation units ensure the engineering rationality of output parameters, expert rule residual connections introduce the technological wisdom accumulated by humans over a long period of time, and attention mechanisms and multi-task learning endow the model with the ability to mine complex nonlinear relationships from data, making the dynamic optimization model accurate, reliable, and interpretable.

[0096] In a preferred embodiment of this application, after the mixing process is completed and before the slurry is transported to the casting process, an online quality monitoring station is set up to collect key data characterizing the state of the foamed cement slurry, including image data and wet density data. A high-definition industrial camera and a matching lighting system are deployed in the foamed cement slurry conveying channel or at a specially designed observation window. The lighting system preferably uses a side-facing or coaxial lighting scheme to highlight the macroscopic texture of the foamed cement slurry surface, the distribution of air bubbles, and any possible clumps or layering. The camera collects image data of the slurry at a fixed frequency. In process sections near the image acquisition points, online densitometers, such as non-contact microwave densitometers or X-ray densitometers, are installed. The densitometers continuously and in real time measure the wet density of the foamed cement slurry flowing through their detection area. Wet density is the most direct comprehensive physical indicator of the slurry, reflecting the overall compactness of the mixture of slurry, foam, and air, and is a key basis for subsequent calculations of dry density and porosity.

[0097] To perform intelligent analysis of the collected mixed slurry image data and wet density data, a mixed slurry quality analysis model is pre-constructed. This model includes the following processing layers: an image feature extraction network, a multi-source data fusion analysis layer, and a quality assessment attribution output layer. The image feature extraction network uses a second convolutional neural network to process the mixed slurry image data, obtaining a second high-dimensional image feature vector. The features extracted by the second convolutional neural network focus on: macroscopic uniformity features, identifying the presence of obvious color or brightness blocks or stripes in the image to determine whether the slurry is uniformly mixed; surface bubble morphology features, analyzing the approximate size distribution and aggregation of visible bubbles on the surface to indirectly assess the quality of the internal pore structure; and rheological visual features, inferring the apparent viscosity and workability based on the flow profile and adhesion characteristics of the foamed cement mixed slurry within the observation window. The multi-source data fusion analysis layer also introduces a dual-head attention mechanism to fuse the second high-dimensional image feature vector and wet density data. The second high-dimensional image feature vector and the real-time acquired wet density data serve as two independent input heads. The dual-head attention mechanism first calculates its own query vector, key vector, and value vector for each input head. Subsequently, the second high-dimensional image feature vector queries the context of the wet density data, while the wet density data also queries the context of the second high-dimensional image feature vector. Through cross-attention calculation, a second fused feature is generated. The second fused feature not only retains the original important information of the second high-dimensional image feature vector and the wet density data, but also highlights the correlated, corroborating, or contradictory parts between the two data, thus more comprehensively characterizing the composite quality state of the mixed slurry. The quality assessment attribution output layer is a fully connected neural network composed of multiple hidden layers. Through nonlinear transformation, it maps high-dimensional fused features to the final analysis result space. The output of the fully connected neural network is structured into a mixed slurry quality analysis result containing three core dimensions, mainly including: mixed slurry quality grade, mixed slurry performance quantitative index, and pouring recommendations. The mixed slurry quality grade is a classification output based on comprehensive evaluation, and outputs the probability distribution of each quality grade. The final mixed slurry quality grade is determined according to the highest probability of the probability distribution. The mixed slurry quality grades include: qualified, quality warning, and unqualified. Qualified means that pouring can be carried out according to normal process. Quality warning means that fine-tuning is required to meet the pouring requirements. Unqualified means that pouring cannot be carried out, and it is recommended to discard this batch of mixed slurry and check the foaming and mixing process parameters.The performance quantification index of the mixed slurry is a multi-task regression output, providing a series of accurate and continuous physical and technological indicators, including but not limited to: apparent bubble size index, which is the relative size index of bubbles on the slurry surface calculated based on the feature vector of the second high-dimensional image; mixed slurry uniformity coefficient, which quantifies the degree of uniformity of distribution of each component in the slurry; deviation between measured wet density and target value, which calculates the absolute deviation and relative percentage between the current wet density and the process-set target value; and rheological score, which predicts the fluidity and castability of the mixed slurry. The performance quantification index of the mixed slurry provides an objective and precise measure of the quality status of foamed cement mixed slurry. The pouring suggestion is a condition-generated output, which generates specific and actionable pouring suggestions based on the quality grade and performance quantification index of the mixed slurry, combined with a pre-built process knowledge base. For example, if the mixed slurry quality grade is qualified, the recommendation is: "The slurry condition is ideal; it is recommended to pour immediately according to the standard process." If the mixed slurry quality grade is a quality warning and the uniformity coefficient is low, the recommendation is: "The mixed slurry uniformity is average, with a slight risk of segregation. It is recommended to stir at low speed (≤30 RPM) for 30 seconds in the storage tank before pouring, and to prioritize its use on non-load-bearing components." If the mixed slurry quality grade is unqualified and the wet density is severely low, the recommendation is: "The slurry density is substandard, and the bubble structure may be unstable. It is recommended to discard this batch of slurry and check the parameters of the foaming and mixing processes." This application directly transforms mixed slurry quality analysis into production guidelines, lowering the operational threshold and improving response speed and decision consistency.

[0098] In a preferred embodiment of this application, the generated mixed slurry quality analysis results are pushed to the intelligent manufacturing control platform and human-machine interface in real time, and the pouring suggestions are directly displayed on the operation terminal of the pouring station to guide the workers' operations.

[0099] The mixed slurry quality analysis model in this application utilizes a dual-head attention mechanism to enable in-depth interaction between mixed slurry image data and wet density data, capturing more subtle and complex data correlations and significantly improving the model's feature utilization efficiency and discrimination accuracy. The mixed slurry quality analysis results are set at three levels: qualitative, quantitative, and suggestive, fully covering the cognitive chain from state perception to decision execution. This provides strong interpretability, making human-computer interaction more natural and efficient, and is particularly beneficial for inexperienced operators to quickly make correct responses.

[0100] S3. Based on the virtual curing model, an optimal temperature and humidity curing curve is generated in the intelligent manufacturing digital twin platform. The temperature and humidity setpoints of the curing kiln are dynamically adjusted according to the optimal temperature and humidity curing curve. In a preferred embodiment of this application, a preset pouring control model is used to convert the quantitative performance indicators of the mixed slurry from the mixed slurry quality analysis results and pouring suggestions into a set of pouring parameters to control the pouring operation. The control dimensions include: pouring speed and vibration parameters. In a preferred embodiment of this application, the pouring control model uses a gradient boosting tree as its basic architecture. Input features include: quantitative performance indicators of the mixed slurry from the mixed slurry quality analysis results; environmental and equipment conditions, such as mold temperature, ambient humidity, and pouring gate height; product specifications, such as target plate thickness and mold size. Output targets include: an optimal pouring speed correction coefficient and optimal vibration parameter suggestion values. The optimal pouring speed correction coefficient is an adjustment ratio relative to the standard speed, and the optimal vibration parameter suggestion values ​​include: amplitude (mm), frequency (Hz), and duration (s). The training data for the casting control model comes from the historical production database. Production records with excellent final slab quality are selected, and their corresponding casting parameters are matched with the corresponding quantitative indicators of mixed slurry performance, environmental and equipment status, and product specifications as training samples for the casting control model.

[0101] Furthermore, during the pouring process, pouring volume data, slab condition data after pouring, and slab spatiotemporal reference data are collected simultaneously. Pouring volume data includes the total pouring weight of a single slab and the pouring flow rate time series. Slab condition data includes the initial temperature of the slurry, the initial surface morphology of the slurry, the apparent color and texture of the slurry, and the surface moisture content of the slurry. Slab spatiotemporal reference data includes the slab's unique identifier, timestamp, and position coordinates within the curing kiln. Pouring volume data is collected by installing high-precision weighing sensors at the outlet of the pouring pipe or on the mold support unit. If the high-precision weighing sensor is installed on the mold support unit, the weight of the empty mold is measured and recorded before pouring, and the total weight of the mold and slurry is measured immediately after pouring to calculate the slab's net weight. For the pouring flow rate curve, the high-frequency readings of the weighing sensor are used to record the weight change curve over time, i.e., the pouring flow rate curve. For initial temperature acquisition of the slurry, a non-contact infrared thermal imager is fixedly installed above the pouring completion point to quickly scan the entire slab surface, generating a temperature distribution heat map, and then outputting the average temperature, maximum temperature, minimum temperature, and temperature standard deviation. For initial surface morphology, as well as the apparent color and texture of the slurry, an industrial-grade binocular or multi-lens color structured light 3D scanner is used, installed on a straight track above the pouring line. While the slab is stationary, the scanner moves at a uniform speed to complete the scan, generating high-precision surface 3D point cloud data. Each 3D point is mapped to a corresponding RGB color value, and the initial surface morphology of the slurry is finally output, along with annotations of the apparent color and texture. For surface moisture acquisition, a near-infrared spectrometer is used. By analyzing the characteristic absorption peaks of water molecules, the surface moisture content is inverted, and then the surface moisture content of the slurry is output.

[0102] Furthermore, based on the casting volume data, the completed slab state data, and the slab spatiotemporal reference data, a corresponding independent virtual entity is created for each physical slab in a pre-built intelligent manufacturing digital twin platform. Specifically, based on the production order's dimensional data, such as length, width, and thickness, as well as possible structural information such as grooving and chamfering, the parametric modeling engine within the intelligent manufacturing digital twin platform is automatically driven to generate a three-dimensional geometric model that is completely consistent with the physical slab to be produced. This three-dimensional geometric model accurately reflects the macroscopic shape and volume of the slab and serves as the computational domain foundation for subsequent physical field simulation. The synchronously collected casting volume data, the completed slab state data, and the slab spatiotemporal reference data are used as the initial physical attributes of this three-dimensional geometric model to obtain the virtual slab. Further, the pre-built slab hydration thermal field simulation engine of the intelligent manufacturing digital twin platform is used to simulate the curing process of the virtual slab, resulting in a virtual curing model. The intelligent manufacturing digital twin platform of this application includes a board hydration thermal field simulation engine comprising temperature field equations, humidity field equations, and hydration process equations. These equations are used to simulate the curing process of the boards, controlling the heater power, humidifier spray volume, and damper opening within the curing kiln to optimize the curing process of foamed cement composite boards. The temperature field equation is expressed as follows:

[0103]

[0104]

[0105]

[0106] in, This indicates the density of the foamed cement slurry as a function of time. This indicates the specific heat capacity of foamed cement paste as a function of hydration degree. , representing the temperature field distribution, Represents spatial location coordinates, Indicates a time index. Thermal conductivity, representing the change in hydration degree of foamed cement paste. Represents the heat conduction term. This represents the heat released during the complete hydration of a unit volume of foamed cement paste. This indicates the degree of hydration, which changes over time. Indicates the hydration reaction rate, This represents the specific heat capacity at initial hydration. This indicates the final specific heat capacity after hydration is complete. The thermal conductivity represents the initial hydration. Indicates the final thermal conductivity after hydration is complete. The power function representing the degree of hydration. It represents an empirical index that reflects the nonlinear growth of thermal conductivity as the microstructure develops.

[0107] The humidity field equation adopts the improved Bažant humidity diffusion model, as follows:

[0108]

[0109] in, This represents the relative humidity field distribution, specifically expressed as... , The humidity diffusion coefficient is a function of humidity, temperature, and degree of hydration. This represents the water diffusion term. This represents the decrease in relative humidity caused by the change in the degree of hydration of a unit volume of foamed cement paste. The left side of the humidity field equation represents the rate of change of relative humidity per unit volume over time, the first term on the right side is the net humidity change rate caused by water diffusion, and the second term on the right side is the rate of decrease in humidity caused by the consumption of pore water by the hydration reaction.

[0110] This represents the hydration sink, which occurs during the hydration process when cement minerals react chemically with pore water, converting free water into chemically bound water, thereby reducing the relative humidity in the pores. This term quantifies the rate of decrease in relative humidity of the foamed cement slurry per unit time due to water loss during hydration.

[0111] The humidity diffusion coefficient model is expressed as follows:

[0112]

[0113]

[0114]

[0115] in, Indicates the reference humidity diffusion coefficient. This represents the humidity-dependent function, describing the nonlinear relationship between the humidity diffusion coefficient and relative humidity. The diffusion mechanisms differ in low-humidity and high-humidity regions. The activation energy represents the rate of water diffusion and characterizes the sensitivity of the water migration process to temperature. Represents the universal gas constant. Represents absolute temperature. This represents the baseline diffusion coefficient in the fully hydrated state. This represents the dimensionless attenuation coefficient, a shape parameter controlling the rate of decay of the baseline humidity diffusivity coefficient with respect to hydration degree. It is a positive value, obtained through inversion fitting; the larger the value, the faster the attenuation. This represents the reference hydration degree, a reference value used to adjust the shape of the function, and is usually set to 1 or a value close to 1. Represents a minimal constant. This represents the low humidity plateau value, indicating the relative diffusion capacity under extremely dry conditions. The critical relative humidity, representing the point at which a significant change occurs in the relative humidity curve, is typically associated with the humidity at which a continuous water film begins to form in the major capillaries or capillary condensation occurs, and is approximately 0.7–0.85. Indicates the shape index.

[0116] The hydration process equation, using the extended Knutsen-Byskov model, is expressed as:

[0117]

[0118] in, This represents the effect function of temperature on hydration degree. The function representing the effect of humidity. The equation represents the reaction order and characterizes the mechanism of reaction kinetics. The left side of the hydration process equation represents the hydration reaction rate, while the right side is the intrinsic rate determined by temperature and reaction stage, multiplied by the availability factor determined by humidity and the driving force determined by the remaining unhydrated amount.

[0119]

[0120] in, Indicates the pre-exponential factor, with the dimension of time. -1 It was set to vary with the degree of hydration to reflect the changes in the dominant mechanism at different reaction stages. It represents the apparent activation energy, which is the energy barrier that the hydration reaction needs to overcome.

[0121]

[0122] in, This represents the initial activation energy, corresponding to the early stages of hydration, when the reaction is controlled by chemical reactions and the activation energy is relatively low. This represents the final activation energy.

[0123]

[0124] in, This indicates the critical relative humidity for hydration. The humidity effect index controls the rate at which the response rate recovers as humidity increases. It is an empirical value, usually obtained by fitting experimental data of maintenance under different humidity conditions, and ranges from 1 to 4.

[0125]

[0126]

[0127] in, and Indicates the stage reaction order. For the early stages (where chemical reaction control is dominant), the reaction order is typically close to 1. The corresponding reaction order in the later stages is usually greater than 1. , This represents the critical hydration degree of the mechanism transition, an approximate critical point where the mechanism shifts from early reaction control to later diffusion control. Indicates early pre-exponential factors. Indicates the pre-exponential factor in the later stage. This represents the steepness coefficient of the transition zone.

[0128] The coupling equation is expressed as follows:

[0129]

[0130]

[0131]

[0132]

[0133] in, This represents the heat release power of complete hydration of a unit volume of foamed cement paste. Indicates the attenuation coefficient. Indicates the hydration completion attenuation factor. Indicates the effective chemically bound water coefficient. This represents the maximum chemically bound water content, and is a material constant. Represents the binding rate coefficient, controlling Follow Increase and approach The speed.

[0134] The boundary conditions of the plate hydration thermal field simulation engine include thermal boundary conditions and wet boundary conditions, where the thermal boundary conditions are:

[0135]

[0136] in, Indicates the thermal conductivity at the surface of the material. This represents the temperature gradient along the direction normal to the boundary. Indicates the surface convective heat transfer coefficient. Indicates the surface temperature of the board. This indicates the dry-bulb temperature of the air inside the curing kiln. Represents the Stefan-Boltzmann constant. Emissivity represents the surface emissivity of a sheet material, characterizing its ability to emit or absorb energy in the form of thermal radiation. This represents the radiative heat transfer temperature difference driving term. The left side of the thermal boundary condition represents the heat flow from the interior of the plate to the surface through thermal conduction. The first term on the right side of the thermal boundary condition represents the heat flow lost from the plate surface to the air through convection. The second term on the right side of the thermal boundary condition represents the heat flow lost from the plate surface to the environment through radiation.

[0137] The wet boundary conditions are:

[0138]

[0139] in, This represents the humidity diffusion coefficient at the material surface. This represents the gradient of relative humidity along the direction normal to the boundary. The surface moisture exchange coefficient, expressed in m / s, is a parameter characterizing the ability of the plate surface to exchange moisture with the air inside the curing kiln. This indicates the relative humidity of the pores on the surface of the board. This indicates the relative humidity of the air inside the curing kiln. The left side of the wet boundary condition represents the volumetric flux of moisture from the interior of the plate to the surface via diffusion, while the right side represents the flux of moisture exchanged from the surface to the air via convection.

[0140] The surface moisture exchange coefficient and the surface convective heat transfer coefficient are related through the Lewis relation, as follows:

[0141]

[0142] in, Represents the Lewis number, Indicates the air density inside the kiln. This indicates the specific heat capacity of the air at constant pressure inside the kiln.

[0143] The kiln environment model includes the energy conservation equation for curing air inside the kiln and the moisture mass conservation equation for curing air inside the kiln. The energy conservation equation for curing air inside the kiln is expressed as follows:

[0144]

[0145] in, Indicates the air volume inside the curing kiln. This indicates the rate of change of air temperature inside the curing kiln with respect to time. Indicates the first The area of ​​each heat exchange surface Indicates the first The convective heat coefficient of each heat exchange surface Indicates the first The temperature of each heat exchange surface Indicates the heating power of the heating system. This represents the heat loss power transferred from the kiln body to the external environment, and is usually calculated as follows:

[0146]

[0147] in, This represents the overall heat transfer coefficient of the kiln body. Indicates the area of ​​the kiln body. This indicates the ambient temperature outside the kiln.

[0148] The equation for the conservation of moisture in the air inside the curing kiln is expressed as follows:

[0149]

[0150] in, Indicates the absolute humidity of the air inside the kiln. This indicates the rate of change of absolute humidity of the air inside the kiln. Indicates the first The wet exchange coefficient of a heat exchange surface Indicates the first The relative humidity of each heat exchange surface This indicates the steam injection rate of the steam humidification system. The rate of moisture loss due to ventilation / exhaust is usually expressed as:

[0151]

[0152] in, Indicates the volumetric flow rate of ventilation. This indicates the absolute humidity of the air outside the kiln.

[0153] In a preferred embodiment of this application, in order to achieve accurate prediction of the model, the internal parameters in the temperature field control equation, humidity field equation, hydration process equation, coupling relationship equation, energy conservation equation of air in curing kiln, and moisture mass conservation equation of air in curing kiln are also calibrated based on historical data using a dynamic calibration strategy through the parameter calibration module.

[0154] S4. Based on the virtual curing model, an optimal temperature and humidity curing curve is generated in the intelligent manufacturing digital twin platform. The temperature and humidity setpoints of the curing kiln are dynamically adjusted according to this optimal temperature and humidity curing curve. In a preferred embodiment of this application, boundary conditions connect the field inside the board to the environmental state inside the kiln. The kiln environment model further links changes in the kiln environment state with the heat and moisture exchange on all board surfaces and the actions of the control system, forming a closed-loop, two-way, strongly coupled system. Based on the virtual curing model, the optimal temperature and humidity curing curve is solved in the intelligent manufacturing digital twin platform. This optimal temperature and humidity curing curve enables the foamed cement composite board to theoretically achieve optimal performance.

[0155] In the specific curing process of the board, the current measured temperature and humidity values ​​of the curing kiln are obtained through distributed temperature and humidity sensors in the curing kiln. Based on the current measured temperature and humidity values ​​and the optimal temperature and humidity curing curve, the temperature and humidity setpoints in the prediction time domain are optimized. The objective function of the optimization solution includes: temperature setpoint error term, humidity setpoint error term, and setpoint change smoothing term. While tracking the optimal temperature and humidity curing curve, the stability of temperature and humidity changes is ensured.

[0156] In a preferred embodiment of this application, the kiln environment model is the basis for dynamically adjusting the curing temperature and humidity, by adjusting the heating power of the heating system. and the steam injection rate of the steam humidification system It can make and Track the optimal temperature and humidity curing curves, such as the stages of heating, constant temperature and humidity, and cooling.

[0157] In the preferred embodiment of this application, through strong coupling modeling of hydration reaction kinetics, heat and moisture transfer, and mechanical properties, based on the scientific principle that "chemical reactions drive heat and moisture migration, and heat and moisture conditions react with chemical reactions and final properties" during the curing process of foamed cement, the simulation of the intelligent manufacturing digital twin platform is no longer a surface phenomenon simulation, but a digital reproduction of the internal mechanism. The prediction results have a solid scientific basis and extremely high reliability. In the physical world, it is difficult to deploy large-scale points to measure the internal temperature and humidity fields of slabs. However, in the virtual model, these can be completely and continuously simulated, changing the current situation where curing processes are estimated based on experience. This enables precise, differentiated, and dynamic curing control, such as providing a direct decision map for adjusting the air supply temperature and humidity in different areas, greatly improving the initiative and efficiency of quality control.

[0158] S5. Control the multi-axis cutting machine to cut the cured slab and assign a unique identification code to the cut finished slab. In the preferred embodiment of this application, the actual length, width, and height of the cured large slab, the order specifications and quantity, and process constraint parameters are input into a pre-constructed layout optimization model to generate an optimal layout diagram. The layout optimization model places all the required finished slab sizes on the two-dimensional plane of the slab to maximize material utilization as the optimization objective. It considers saw kerf loss, equipment cutting boundaries, and possible unusable edge areas of the slab. A heuristic algorithm is used for optimization, and the output optimal layout diagram includes the precise position and orientation of each finished slab on the parent slab. The heuristic algorithm can be a particle swarm optimization algorithm. Furthermore, on the determined optimal layout diagram, a pre-constructed path planning model is used to plan the order and idle movement path of the tool accessing and cutting each sub-slab, with minimizing the total idle movement time as the optimization objective. The path planning model integrates expert experience rules, such as prioritizing cutting internal holes before cutting the outer contour; cutting adjacent edges continuously to reduce tool lifting; and following the inside-out principle to prevent plate displacement after cutting from affecting accuracy. The path planning model employs a graph search algorithm, using the cutting start point, line segment endpoints, and tool lifting points as nodes to construct a path graph. The A* algorithm (also known as the A* algorithm) is then used to quickly calculate the optimal connection path, ultimately outputting an ordered sequence of cutting tasks and the idle path points connecting these cutting segments. Finally, a pre-built motion trajectory optimization model transforms the optimal cutting task sequence into time-optimal and smooth cutting instructions that can be executed by each axis motor of the multi-axis cutting machine. The optimization variables of the motion trajectory optimization model are the maximum allowable speed, acceleration, and jerk of each motion segment. Constraints include: the speed, acceleration, and jerk of each independent axis must not exceed the maximum values ​​of the motor and mechanical structure; during multi-axis linkage, the combined speed of the tool tip must not exceed the set value; and at corners, speed look-ahead and corner smoothing are required to prevent overshoot and ensure cutting shape accuracy. The motion trajectory optimization model models the motion trajectory optimization problem as a convex optimization problem, uses the interior point method to quickly solve the global optimal velocity curve, and finally outputs a series of cutting position, cutting speed and cutting acceleration instruction sets with timestamps, which are directly sent to the motion controller of the multi-axis cutting machine to control the multi-axis cutting machine to cut the cured slab according to the optimal cutting instructions to obtain the finished slab.

[0159] In the preferred embodiment of this application, the cutting process is upgraded from automation that relies on worker experience and fixed programs to an intelligent production unit capable of autonomous optimization and flexible adaptation. Through scientific calculation using a nesting optimization model, material utilization is maximized, directly reducing raw material costs. The path planning model and motion trajectory optimization model work together to ensure that the cutting machine's movement path is minimized, idle time is reduced, and it operates at the most efficient and smooth speed, significantly improving production efficiency and reducing time and energy costs.

[0160] In a preferred embodiment of this application, a globally unique identification code is applied to a designated location on each finished board after cutting, such as the side or end face of the board. This identification code is linked to the entire production process data of the board in the database, including at least: raw material batch, mixing parameters, pouring time, complete historical temperature and humidity curves in the curing kiln, cutting dimensions, and quality inspection results. The identification code can be one of the following forms or a combination thereof: one-dimensional barcode or two-dimensional barcode, laser-engraved character code, or RFID (Radio Frequency Identification) electronic tag.

[0161] Accordingly, such as Figure 2 As shown, based on a smart manufacturing method for foamed cement composite boards, this embodiment of the invention also provides a smart manufacturing system for foamed cement composite boards, used to implement the smart manufacturing method for foamed cement composite boards disclosed in this embodiment of the invention, including: a foaming monitoring and analysis module 1, a mixing monitoring and analysis module 2, a virtual curing model construction module 3, a curing control module 4, and a cutting control module 5;

[0162] The foaming monitoring and analysis module 1 is used to collect foam flow image data and foam density data corresponding to the foam solution in the manufacturing of foamed cement composite boards, and to analyze the foam flow image data and foam density data using a pre-built foam quality analysis model to obtain foam analysis results.

[0163] The mixing monitoring and analysis module 2 is used to obtain the optimal mixing parameter set of the foam solution and the cement slurry base material based on the foam analysis results and the slurry viscosity value of the cement slurry base material, and to use a pre-constructed mixing slurry quality analysis model to analyze the mixing slurry image data and wet density data of the foamed cement mixing slurry after the mixing operation based on the optimal mixing parameter set, so as to obtain the mixing slurry quality analysis results.

[0164] The virtual curing model construction module 3 is used to control the pouring parameters of the foamed cement slurry according to the quality analysis results of the mixed slurry, and to construct a virtual curing model in the pre-constructed intelligent manufacturing digital twin platform according to the synchronously collected pouring volume data, the state data of the poured slab and the spatiotemporal reference data of the slab. The intelligent manufacturing digital twin platform is set to simulate the hydration heat field coupling relationship during the curing process of the slab.

[0165] The curing control module 4 is used to generate an optimal temperature and humidity curing curve in the intelligent manufacturing digital twin platform according to the virtual curing model, and dynamically adjust the temperature and humidity setpoints of the curing kiln according to the optimal temperature and humidity curing curve.

[0166] The cutting control module 5 is used to control the multi-axis cutting machine to cut the cured slab and assign a unique identification code to the cut finished slab.

[0167] Specific limitations regarding the intelligent manufacturing system for foamed cement composite panels can be found in the above-described intelligent manufacturing method for foamed cement composite panels, and will not be repeated here. Those skilled in the art will recognize that the various modules and steps described in conjunction with the embodiments disclosed in this invention can be implemented in hardware, software, or a combination of both. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0168] This embodiment provides an intelligent manufacturing method and system for foamed cement composite boards, which addresses the technical issues of improving the adaptability and intelligence level of the foamed cement composite board production process and improving product quality. The method includes: collecting foam flow image data and foam density data corresponding to the foam solution during the manufacturing of foamed cement composite boards; analyzing the foam flow image data and foam density data using a pre-constructed foam quality analysis model to obtain foam analysis results; obtaining the optimal mixing parameter set for the foam solution and cement slurry based on the foam analysis results and the slurry viscosity value of the cement slurry base material; and analyzing the mixed slurry image data and wet density data corresponding to the foamed cement mixed slurry after mixing based on the optimal mixing parameter set using a pre-constructed mixed slurry quality analysis model to obtain mixed slurry quality analysis results; and further analyzing the mixed slurry based on the mixed slurry... Based on the quality analysis results, the pouring parameters of the foamed cement slurry are controlled. A virtual curing model is constructed in a pre-built intelligent manufacturing digital twin platform using synchronously collected pouring volume data, finished slab state data, and slab spatiotemporal reference data. This platform is designed to simulate the coupling relationship of hydration and thermal fields during slab curing. An optimal temperature and humidity curing curve is generated within the platform based on the virtual curing model. The temperature and humidity settings of the curing kiln are dynamically adjusted according to this curve. A multi-axis cutting machine is controlled to cut the cured slab, and a unique identification code is assigned to each finished slab. The intelligent manufacturing method for foamed cement composite panels provided in this application employs a foam quality analysis model, overcoming the limitations of traditional methods that only focus on density or simple morphology. Through deep learning, it automatically extracts deep features such as texture, morphology, and stability, and combines this with density data to achieve a multi-dimensional comprehensive evaluation. The provided foam quality analysis results not only indicate the foam quality level but also clearly point out the root causes and process adjustment suggestions, directly elevating quality inspection to decision support and process control recommendations, effectively supporting closed-loop production optimization and predictive maintenance. Based on the foam analysis results and the viscosity value of the cement slurry base, an adaptive control of the mixing process of foam solution and cement slurry base is achieved through a dynamic optimization model. A hybrid intelligent optimization model integrating physical laws, expert experience, and data-driven approaches has been constructed, overcoming the limitations of traditional empirical formulas and the risk of pure data models outputting results that violate physical laws. For the board curing process, an intelligent manufacturing digital twin model has been constructed, changing the current situation of curing processes relying on experience-based estimations, achieving precise, differentiated, and dynamic curing control, and greatly improving the initiative and efficiency of quality management.

[0169] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0170] The above-described embodiments are merely preferred embodiments of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention. Therefore, the scope of protection of this invention should be determined by the scope of the claims.

Claims

1. An intelligent manufacturing method of foamed cement composite board, characterized in that, The method includes: Foam flow image data and foam density data corresponding to the foam solution in the manufacturing of foamed cement composite boards are collected, and the foam flow image data and foam density data are analyzed using a pre-constructed foam quality analysis model to obtain foam analysis results; Based on the foam analysis results and the viscosity value of the cement slurry base, the optimal mixing parameter set of the foam solution and the cement slurry base is obtained. Then, using a pre-constructed mixed slurry quality analysis model, the image data and wet density data of the foamed cement mixed slurry after the mixing operation based on the optimal mixing parameter set are analyzed to obtain the mixed slurry quality analysis results. Based on the quality analysis results of the mixed slurry, the pouring parameters of the foamed cement mixed slurry are controlled, and a virtual curing model is constructed in the pre-constructed intelligent manufacturing digital twin platform based on the synchronously collected pouring volume data, the state data of the poured slab, and the spatiotemporal reference data of the slab. The intelligent manufacturing digital twin platform is set to simulate the hydration heat field coupling relationship during the curing process of the slab. Based on the virtual curing model, an optimal temperature and humidity curing curve is generated in the intelligent manufacturing digital twin platform, and the temperature and humidity setpoints of the curing kiln are dynamically adjusted according to the optimal temperature and humidity curing curve. The multi-axis cutting machine is controlled to cut the cured slab and assign a unique identification code to the cut finished slab. 2.The intelligent manufacturing method of the foamed cement composite board according to claim 1, characterized in that, The construction of the intelligent manufacturing digital twin platform includes: Based on the multi-physics coupling and reaction path theory of foamed cement composite board manufacturing, a physical theoretical model is established to map the performance to the production formula. The physical theoretical model includes: strength model, thermal conductivity model and density model. Obtain historical production formulas and corresponding historical board performance indicators, and input the historical production formulas into the physical theory model to obtain theoretical board performance indicators. Based on the historical board performance indicators and the theoretical board performance indicators, obtain the board performance indicator residuals. Based on the historical production formula and the residual performance index of the board, a training dataset is constructed, and the selected deep neural network is trained using the training dataset to obtain the index residual prediction model. The physical theory model and the index residual prediction model are combined to obtain the board performance index prediction model. Based on the coupling relationship between temperature, humidity and hydration of the foamed cement composite board, a hydration thermal field simulation engine for the board is constructed. The hydration thermal field simulation engine includes: temperature field control equation, humidity field equation, hydration process equation, coupling relationship equation and kiln environment model. The performance index prediction model of the board material and the hydration thermal field simulation engine of the board material are imported into the digital twin platform to obtain the intelligent manufacturing digital twin platform.

3. The intelligent manufacturing method for foamed cement composite panels as described in claim 2, characterized in that, The method further includes: Based on the production order, an initial production formula is generated, and the predicted board performance index is obtained using the board performance index prediction model. The loss value between the predicted board performance index and the target board performance index in the production order is calculated, and based on the loss value, the initial production formula is iteratively optimized using an optimization algorithm until the optimal production formula is found. The optimal production formula is used to control the proportioning of the foam solution and the cement slurry base material by the IoT weighing system.

4. The intelligent manufacturing method for foamed cement composite panels as described in claim 1, characterized in that, The pre-constructed foam quality analysis model is used to analyze the foam flow image data and the foam density data to obtain foam analysis results, including: A first convolutional neural network is used to extract texture features, bubble morphology features, and bubble stability features from the foam flow image data to obtain a first high-dimensional image feature vector. A first dual-head attention mechanism is used to fuse the first high-dimensional image feature vector and the foam density data to obtain a first fused feature. A first fully connected neural network is then used to analyze the first fused feature to obtain foam quality analysis results. The foam quality analysis results include: foam quality grade, foam performance quantitative index, root cause prediction results, and foaming process adjustment suggestions.

5. The intelligent manufacturing method for foamed cement composite panels as described in claim 1, characterized in that, The optimal set of mixing parameters for the foam solution and the cement slurry base is obtained based on the foam analysis results and the slurry viscosity value, including: Based on the viscosity value of the cement slurry base material and the preset viscosity threshold range, a threshold judgment is made on the viscosity value of the slurry to obtain the quality judgment result of the cement slurry base material; Based on the foam analysis results, the slurry viscosity value, and the quality judgment results, a dynamic optimization model pre-built based on a machine learning algorithm is used to obtain the optimal set of mixing parameters for the foam solution and the cement slurry base. The dynamic optimization model includes logic gates, an attention mechanism layer, a physical information layer, and a multi-task output layer. The logic gates perform conditional judgments on the foam quality level and the quality judgment results, and output logical judgment results. The attention mechanism layer performs feature fusion on the slurry viscosity value and the foam performance quantification index based on the logical judgment results. The physical information layer includes a differentiable physical calculation unit and an expert rule residual connection. The differentiable physical calculation unit embeds physical constraints, and the expert rule residual connection embeds expert rules.

6. The intelligent manufacturing method for foamed cement composite panels as described in claim 1, characterized in that, The pre-constructed mixed slurry quality analysis model is used to analyze the mixed slurry image data and wet density data corresponding to the foamed cement mixed slurry after mixing based on the optimal mixing parameter set, and the mixed slurry quality analysis results are obtained, including: A second convolutional neural network is used to extract macroscopic uniformity features, surface bubble morphology features, and rheological visual features from the mixed slurry image data to obtain a second high-dimensional image feature vector; A second dual-head attention mechanism is used to fuse the second high-dimensional image feature vector and the wet density data to obtain a second fused feature. A second fully connected neural network is then used to analyze the second fused feature to obtain the mixed slurry quality analysis results. The mixed slurry quality analysis results include: mixed slurry quality grade, mixed slurry performance quantification index, and pouring recommendations.

7. The intelligent manufacturing method for foamed cement composite panels as described in claim 3, characterized in that, The process involves constructing a virtual curing model within a pre-built intelligent manufacturing digital twin platform based on synchronously collected pouring volume data, completed slab state data, and slab spatiotemporal reference data. This includes: Based on the size data of the production order, a three-dimensional geometric model that is completely consistent with the physical slab is generated through the intelligent manufacturing digital twin platform. The synchronously collected casting volume data, the slab status data after casting, and the slab spatiotemporal reference data are used as the initial physical attribute values ​​of the three-dimensional geometric model to obtain a virtual slab. The virtual slab curing process is simulated using the pre-built hydration thermal field simulation engine of the digital twin platform to obtain a virtual curing model.

8. The intelligent manufacturing method for foamed cement composite panels as described in claim 1, characterized in that, The step of dynamically adjusting the temperature and humidity setpoints of the curing kiln according to the optimal temperature and humidity curing curve includes: The current measured temperature and humidity values ​​of the curing kiln are obtained by the distributed temperature and humidity sensors of the curing kiln. Based on the current measured temperature and humidity values ​​and the optimal temperature and humidity curing curve, the temperature and humidity setpoints of the curing kiln in the predicted time domain are optimized. The objective function of the optimization solution includes: temperature setpoint error term, humidity setpoint error term, and setpoint change smoothing term.

9. The intelligent manufacturing method for foamed cement composite panels as described in claim 1, characterized in that, The control of the multi-axis cutting machine to cut the cured slab includes: Based on the obtained 3D dimensional data of the slab and the specifications of the finished slab, an optimal layout diagram is generated using a pre-constructed layout optimization model. The layout optimization model aims to maximize material utilization. Based on the optimal layout diagram, the optimal cutting task sequence is obtained by using a pre-built path planning model, wherein the path planning model aims to minimize the total idle time. Based on the optimal cutting task sequence, the optimal cutting command is obtained by using a pre-built motion trajectory optimization model. The optimal cutting command includes cutting position, cutting speed and cutting acceleration. The multi-axis cutting machine is controlled to cut the cured slab according to the optimal cutting command to obtain the finished slab.

10. A smart manufacturing system for foamed cement composite panels, used to implement the smart manufacturing method for foamed cement composite panels according to any one of claims 1-9, characterized in that, The system includes: a foaming monitoring and analysis module, a hybrid monitoring and analysis module, a virtual curing model construction module, a curing control module, and a cutting control module; The foaming monitoring and analysis module is used to collect foam flow image data and foam density data corresponding to the foam solution in the manufacturing of foamed cement composite boards, and to analyze the foam flow image data and foam density data using a pre-built foam quality analysis model to obtain foam analysis results. The mixing monitoring and analysis module is used to obtain the optimal mixing parameter set of the foam solution and the cement slurry base material based on the foam analysis results and the slurry viscosity value of the cement slurry base material, and to analyze the image data and wet density data of the foamed cement slurry after mixing based on the optimal mixing parameter set using a pre-constructed mixing slurry quality analysis model, so as to obtain the mixing slurry quality analysis results. The virtual curing model construction module is used to control the pouring parameters of the foamed cement slurry based on the quality analysis results of the mixed slurry, and to construct a virtual curing model in a pre-constructed intelligent manufacturing digital twin platform based on the synchronously collected pouring volume data, the state data of the poured slab, and the spatiotemporal reference data of the slab. The intelligent manufacturing digital twin platform is set to simulate the hydration heat field coupling relationship during the curing process of the slab. The curing control module is used to generate an optimal temperature and humidity curing curve in the intelligent manufacturing digital twin platform based on the virtual curing model, and dynamically adjust the temperature and humidity setpoints of the curing kiln based on the optimal temperature and humidity curing curve. The cutting control module is used to control the multi-axis cutting machine to cut the cured slab and assign a unique identification code to the cut finished slab.

Images