Intelligent preparation method and system of recycled aggregate based on material gene reverse modeling

Abstract

The application discloses a kind of based on material gene reverse modeling's recycled aggregate intelligent preparation method and system, and preparation method includes the following steps: obtaining the multimodal material gene feature of the building waste to be processed input material;Component gene feature, internal defect gene feature and geometric gene feature are fused, and material gene feature vector is generated;The performance index requirement of target recycled aggregate is obtained, and target performance vector is constructed;Material gene feature vector and target performance vector are input into physical constraint reverse model, and optimal crushing process parameters are obtained by reverse solution;According to optimal crushing process parameters, crushing equipment is controlled to crush the input material, and recycled aggregate is obtained.The preparation method of the application can identify the chemical composition, internal defect and original strength of the material, and according to the performance index requirement of target recycled aggregate, the optimal crushing process parameters are obtained by reverse solution, so as to improve the added value of product.

Description

Technical Field

[0001] This invention relates to the field of construction waste treatment technology, and in particular to a method and system for intelligent preparation of recycled aggregates based on reverse modeling of materials genes. Background Technology

[0002] Construction waste mainly refers to various solid wastes generated during construction activities such as demolition, new construction, decoration, and repair, and is a major component of urban solid waste. The large-scale disposal of construction waste through simple landfill or open-air dumping not only occupies valuable land resources but also causes serious pollution to soil, water, and air, and poses safety hazards such as geological disasters.

[0003] Among the many pathways for the resource utilization of construction waste, the most mature and widely used approach is to process waste concrete, bricks, and other inorganic hard materials, which constitute the largest proportion, into recycled aggregates through crushing, screening, and purification processes. Recycled aggregates can serve as partial or complete substitutes for natural sand and gravel in the production of recycled concrete, recycled mortar, and road base materials, thereby conserving natural mineral resources, reducing ecological damage from mining, and alleviating pressure on landfills, resulting in significant environmental and economic benefits.

[0004] However, compared with natural sand and gravel, recycled aggregates exhibit significant performance differences and inherent limitations. This is primarily due to the highly heterogeneous composition resulting from their complex origins. Recycled aggregates are composites composed of various materials, including virgin concrete particles, attached hydrated hardened mortar, brick and tile fragments, unhydrated cement particles, and microcracks. This complex internal structure leads to a series of drawbacks, such as high water absorption, high crushing value, high porosity, a large amount of thin mortar adhering to the surface, and unstable gradation and particle shape. When used in the preparation of new concrete, this often results in poor workability, reduced mechanical strength, decreased volumetric stability (shrinkage, creep), and deterioration of durability (freeze-thaw resistance, impermeability). This has long confined recycled aggregates, especially those made from demolition waste, to a "downgraded utilization" dilemma in practical applications. They are typically only used in low-value-added areas such as non-load-bearing roadbed layers and plaza paving stones, making it difficult to enter high-value applications such as structural concrete with stringent material performance requirements. This severely restricts the economic benefits and development potential of the construction waste recycling industry. Summary of the Invention

[0005] This invention aims to solve at least one of the technical problems existing in the prior art. To this end, this invention proposes a smart preparation method and system for recycled aggregates based on reverse modeling of materials genes. The preparation method can identify the chemical composition, internal defects, and original strength of materials, and, based on the performance requirements of the target recycled aggregates, reverse-engineer the optimal crushing process parameters, thereby solving the problems of unstable production quality, lagging control, and low added value of recycled aggregates.

[0006] This invention also proposes an intelligent preparation system for recycled aggregates.

[0007] The intelligent preparation method for recycled aggregate based on reverse modeling of materials genes according to the first aspect of the present invention includes the following steps: S1: Obtain the multimodal material gene characteristics of the construction waste to be processed, wherein the multimodal material gene characteristics include composition gene characteristics, internal defect gene characteristics and geometric gene characteristics; S2: The component gene features, internal defect gene features and geometric gene features are fused to generate a material gene feature vector; S3: Obtain the performance index requirements of the target recycled aggregate and construct the target performance vector; S4: Input the material gene feature vector and target performance vector into the physical constraint inverse model, and solve in reverse to obtain the optimal crushing process parameters; S5: Control the crushing equipment to crush the construction waste feed according to the optimal crushing process parameters to obtain recycled aggregate.

[0008] In some embodiments, obtaining multimodal material gene features in step S1 specifically includes: Near-infrared spectroscopy (NIRS) and / or laser-induced breakdown spectroscopy (LIBS) are used to identify the proportion of different materials in the feed in real time and obtain the genetic characteristics of the components. Acoustic emission (AE) sensors and / or microwave dielectric sensors are used to quantitatively analyze the microcrack density, porosity, and primary strength of the feed material, thereby obtaining the internal defect gene characteristics. A 3D vision sensor and / or laser profile scanner are used to obtain the particle size distribution and particle shape characteristics of the feed material, and to obtain the geometric gene characteristics.

[0009] In some embodiments, the specific method for feature fusion in step S2 is as follows: Instance-level normalization was performed on the component gene subvectors, defect gene subvectors, and geometric gene subvectors, respectively. A weighted average fusion algorithm is adopted, and each sub-vector is weighted and concatenated according to the confidence adaptive weight. The confidence adaptive weight is updated online in an exponential smoothing manner according to the prediction error of each modality on the target gradation. Output the material gene feature vector.

[0010] In some embodiments, the specific method for feature fusion in step S2 is as follows: A three-branch encoder is used to encode the component gene sequence, defect gene sequence, and geometric gene vector respectively to obtain the feature vector of each branch; After concatenating the feature vectors of each branch, the weights of each modal channel are learned through a channel attention network; The material gene feature vector is compressed and output through a fully connected layer.

[0011] In some embodiments, the method for constructing the physical constraint inverse model in step S4 includes: Based on Bond's theory of crushing power consumption and particle size distribution evolution theory, a positive physical driving model is established, which describes the mapping relationship between process parameters and product performance. An objective function is constructed based on the positive physics-driven model, wherein the objective function is the weighted error between the prediction performance and the target performance plus a physical constraint penalty term; The objective function is iteratively optimized using the Newton-Raphson iterative method or the Levenberg-Marquardt algorithm to solve for the optimal process parameters in reverse.

[0012] In some embodiments, step S5 specifically includes: Before the material enters the crusher, feedforward control is performed according to the optimal crushing process parameters to preset the discharge port opening and rotor speed of the crushing equipment; During the crushing process, the discharge port opening or rotor speed of the crushing equipment is dynamically fine-tuned based on real-time operating parameters using an adaptive adjustment algorithm. The real-time operating parameters include current load, vibration signal, and acoustic emission signal, and the adaptive adjustment algorithm includes a fuzzy PID control algorithm.

[0013] The intelligent preparation system for recycled aggregate based on reverse modeling of materials genes according to a second aspect of the present invention comprises: The multimodal material gene sensing unit is used to collect the component gene features, internal defect gene features and geometric gene features of the input material, and fuse the multimodal features to generate a unified material gene feature vector; The intelligent reverse decision-making unit is used to receive the performance index requirements of the target recycled aggregate and solve the optimal crushing process parameters in reverse through the physical constraint reverse model. An adaptive execution control unit is used to control the crushing equipment to perform crushing processing according to the optimal crushing process parameters.

[0014] In some embodiments, the multimodal material gene sensing unit includes: The component feature recognition module uses near-infrared spectroscopy (NIRS) and / or laser-induced breakdown spectroscopy (LIBS) technology to identify the proportion of different materials in the feed in real time. The internal defect sensing module uses acoustic emission (AE) sensors and / or microwave dielectric sensors to quantitatively analyze the microcrack density, porosity, and primary strength of the feed material. The geometric feature acquisition module uses a 3D vision sensor or a laser contour scanner to acquire the particle size distribution and particle shape characteristics of the feed material. The data fusion module integrates multi-source heterogeneous sensing information into a unified material gene feature vector.

[0015] In some embodiments, the intelligent reverse decision-making unit includes: The target performance input module is used to receive the target aggregate performance indicators set by the user. The physical constraint inverse model, based on the principles of material crushing mechanics and particle physics, solves in reverse the optimal combination of process parameters to satisfy the target performance. The dynamic optimization engine performs online self-learning and fine-tuning of model parameters based on real-time production feedback data; The process parameter library stores the historical optimal parameter combinations for different equipment models, which are used for model initialization and constraint boundary setting.

[0016] In some embodiments, the adaptive execution control unit includes: The feedforward control module presets the crusher parameters based on the results of the inverse model solution before the material enters the crusher. The adaptive adjustment module dynamically fine-tunes parameters based on real-time operating conditions during the crushing process; The status monitoring module monitors the equipment's operating status in real time. The actuators include jaw crushers, impact crushers, vertical shaft impact crushers, and their associated frequency conversion drives and hydraulic adjustment systems.

[0017] The present invention has at least the following beneficial effects: (1) Existing technologies mostly adopt the "trial and error method" or simple feedback control based on a single physical index (such as particle size), which cannot cope with the problem of complex composition and large differences in internal defects of construction waste. This invention precisely quantifies the composition, internal microcracks and geometric features of the feed material before crushing, and combines it with the "physical constraint inverse model" for feedforward control, thereby eliminating the impact of raw material fluctuations on product quality from the source, and ensuring that the key performance indicators such as crushing value and water absorption rate of recycled aggregates stably meet the requirements of high-standard engineering.

[0018] (2) This invention constructs a reverse model based on the principles of material crushing mechanics and particle science, which can solve for the theoretically optimal combination of process parameters for each batch of feed, considering the composition, internal microcracks, and geometric characteristics. This reduces and avoids ineffective energy consumption and equipment wear caused by over-crushing, and also prevents the increase in rework rate caused by under-crushing, thereby maximizing the output of qualified aggregates per unit energy consumption.

[0019] (3) The present invention adopts a dual control architecture of "feedforward preset + adaptive fine-tuning": the feedforward link sets the benchmark working condition in advance based on the material gene, which greatly shortens the system response lag time; the adaptive link uses algorithms such as fuzzy PID to dynamically compensate for load disturbances based on real-time signals such as current and vibration, so that the system can still operate smoothly when facing extreme inferior materials or sudden material blockage risks, which significantly reduces the downtime failure rate and the frequency of manual intervention.

[0020] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of an intelligent manufacturing system according to an embodiment of the present invention; Figure 2 This is a flowchart of an intelligent fabrication method according to an embodiment of the present invention. Detailed Implementation

[0023] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0024] The following disclosure provides numerous different embodiments or examples for implementing various structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples. Such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed. Additionally, examples of various specific processes and materials are provided in this invention; however, those skilled in the art will recognize the applicability of other processes and / or the use of other materials.

[0025] Below, refer to Figure 1 The present invention describes an intelligent preparation system for recycled aggregate based on reverse modeling of materials genes according to the second aspect of the present invention. The system consists of three core units: a perception layer, a decision layer, and an execution layer. Through the synergy of multimodal sensing, reverse modeling, and adaptive control technologies, the system realizes customized production of recycled aggregate.

[0026] The sensing layer includes a multimodal material gene sensing unit, which is used to collect the component gene features, internal defect gene features and geometric gene features of the input material, and fuse the multimodal features to generate a unified material gene feature vector. The decision-making layer includes an intelligent reverse decision-making unit, which receives the performance index requirements of the target recycled aggregate and solves the optimal crushing process parameters in reverse through a physical constraint reverse model. The execution layer includes an adaptive execution control unit, which is used to control the crushing equipment to perform crushing processing according to the optimal crushing process parameters.

[0027] In some embodiments, the multimodal material gene sensing unit includes: The component feature recognition module uses near-infrared spectroscopy (NIRS) and / or laser-induced breakdown spectroscopy (LIBS) technology to identify the proportion of different materials in the feed in real time. The internal defect sensing module uses acoustic emission (AE) sensors and / or microwave dielectric sensors to quantitatively analyze the microcrack density, porosity, and primary strength of the feed material. The geometric feature acquisition module uses a 3D vision sensor or a laser contour scanner to acquire the particle size distribution and particle shape characteristics of the feed material. The data fusion module integrates multi-source heterogeneous sensor information into a unified material gene feature vector. This data fusion module can employ either a weighted average or a deep learning feature fusion algorithm.

[0028] In some embodiments, the intelligent reverse decision-making unit includes: The target performance input module is used to receive the target aggregate performance indicators set by the user; for example, the gradation and crushing value of aggregate for C30 concrete ≤12%, the content of needle-shaped and flaky particles ≤10%, and the mud content ≤1%; it supports manual setting or direct issuance by the MES system.

[0029] The physical constraint inverse model, based on the principles of material crushing mechanics and particle physics, solves in reverse the optimal combination of process parameters to satisfy the target performance. The dynamic optimization engine performs online self-learning and fine-tuning of model parameters based on real-time production feedback data; the dynamic optimization engine can employ model predictive control or online learning algorithms.

[0030] The process parameter library stores the historical optimal parameter combinations for different equipment models, which are used for model initialization and constraint boundary setting.

[0031] In some embodiments, the adaptive execution control unit includes: The feedforward control module presets the crusher parameters based on the results of the inverse model solution before the material enters the crusher; the feedforward control module can send commands to the frequency converter or hydraulic adjustment system through the PLC.

[0032] The adaptive adjustment module dynamically fine-tunes parameters based on real-time operating conditions during the crushing process; the adaptive adjustment module is based on a fuzzy PID control algorithm to prevent material blockage or over-crushing.

[0033] The status monitoring module monitors the equipment's operating status in real time. For example, the status monitoring module integrates current sensors, temperature sensors, accelerometers, etc., to detect the crusher load, bearing temperature, and vibration intensity. When the parameters exceed the safety threshold, it automatically triggers protection or reduces the load.

[0034] The actuators include jaw crushers, impact crushers, vertical shaft impact crushers, and their associated frequency conversion drives and hydraulic adjustment systems.

[0035] The intelligent recycled aggregate preparation system based on materials genome reverse modeling of this invention can achieve fully closed-loop intelligent control from "raw material perception - target decision - process execution", significantly improving the product quality stability and production efficiency of recycled aggregates, and reducing reliance on human experience. This system achieves a paradigm shift from "experience-driven" to "data-driven" processes through multi-source sensor fusion and physical constraint reverse modeling, enabling the recycled aggregate preparation process to have adaptive and self-optimizing capabilities. This allows for the stable production of high-quality aggregates that meet target performance even under complex conditions such as raw material fluctuations and equipment wear.

[0036] The intelligent preparation method and system for recycled aggregate based on reverse modeling of materials genes of the present invention can be applied to: 1. Construction waste recycling center / plant: Suitable for urban construction waste recycling bases, fixed or mobile construction waste processing production lines, used to convert solid waste such as demolition waste and decoration waste into high-quality recycled aggregates.

[0037] 2. Customized production of recycled aggregates: Suitable for high-value-added scenarios that require reverse customization of recycled aggregates with specific gradation, particle shape, and crushing value based on the performance requirements of downstream products (such as C30 and C40 concrete, precast components, permeable bricks, etc.).

[0038] 3. Co-processing of solid waste: In addition to typical construction waste, it is also suitable for the crushing, shaping and resource utilization of bulk solid waste such as tailings, coal gangue and industrial slag.

[0039] 4. Intelligent upgrade of crushing equipment: It can be integrated into various crushing equipment such as jaw crushers, impact crushers, cone crushers, and vertical shaft impact crushers, serving as the core decision-making module of their intelligent control system.

[0040] It is particularly suitable for the following specific application scenarios: Construction waste recycling project in cold northern regions (requiring compliance with freeze-thaw durability requirements).

[0041] The production of aggregates for structural concrete requires a high rate of recycled aggregate substitution (≥30%).

[0042] Green buildings, prefabricated buildings, and other projects that have strict requirements on the carbon footprint of building materials.

[0043] Reference Figure 2 According to the first aspect of the present invention, the intelligent preparation method of recycled aggregate based on reverse modeling of materials genes includes the following steps: S1: Obtain the multimodal material gene characteristics of the construction waste to be processed, wherein the multimodal material gene characteristics include composition gene characteristics, internal defect gene characteristics and geometric gene characteristics; S2: The component gene features, internal defect gene features and geometric gene features are fused to generate a material gene feature vector; S3: Obtain the performance index requirements of the target recycled aggregate and construct the target performance vector; S4: Input the material gene feature vector and target performance vector into the physical constraint inverse model, and solve in reverse to obtain the optimal crushing process parameters; S5: Control the crushing equipment to crush the construction waste feed according to the optimal crushing process parameters to obtain recycled aggregate.

[0044] Among them, the component gene characteristics refer to the chemical composition and mineral composition of various materials in construction waste, including but not limited to cement stone content, aggregate type, brick and tile ratio, and impurity content; the internal defect gene characteristics include microcrack density, porosity, and degree of damage in the interface transition zone; and the geometric gene characteristics include particle size distribution, shape index, and surface texture characteristics. The target performance vector includes key quality parameters such as particle size distribution, crushing index, water absorption rate, and content of needle-like and flaky particles in the recycled aggregate.

[0045] In some embodiments, obtaining multimodal material gene features in step S1 specifically includes: Near-infrared spectroscopy (NIRS) and / or laser-induced breakdown spectroscopy (LIBS) are used to identify the proportion of different materials in the feed in real time and obtain the genetic characteristics of the components. Acoustic emission (AE) sensors and / or microwave dielectric sensors are used to quantitatively analyze the microcrack density, porosity, and primary strength of the feed material, thereby obtaining the internal defect gene characteristics. A 3D vision sensor and / or laser profile scanner are used to obtain the particle size distribution and particle shape characteristics of the feed material, and to obtain the geometric gene characteristics.

[0046] In some embodiments, the specific method for feature fusion in step S2 is as follows: Instance-level normalization was performed on the component gene subvectors, defect gene subvectors, and geometric gene subvectors, respectively. A weighted average fusion algorithm is adopted, and each sub-vector is weighted and concatenated according to the confidence adaptive weight. The confidence adaptive weight is updated online in an exponential smoothing manner according to the prediction error of each modality on the target gradation. Output the material gene feature vector.

[0047] A specific method for feature fusion can be a weighted average fusion algorithm. For example, for the first Block material instance, three sub-vectors: Component genes: This includes cement stone content, aggregate type, brick and tile ratio, impurity content, and other components; Defective genes: This includes intrinsic strength, microcrack density, porosity, and the degree of damage in the interface transition zone; Geometric Gene: This includes particle size distribution, shape index, and surface texture features.

[0048] Each subvector is first standardized at the instance level: in Take the mean and standard deviation of the sliding window for this instance in the corresponding mode (to prevent mode drift).

[0049] Set static initial weights. In other words, composition determines strength with the highest weight, defects are second, and geometry only affects the gradation shape.

[0050] Online confidence correction (adaptive weighting), error between the current batch predicted gradation and the target gradation for each modality. Update weights using exponential smoothing: Finally, the output is fused and spliced ​​(⊕) to output a 12-dimensional unified gene feature: .

[0051] In some embodiments, the specific method for feature fusion in step S2 is as follows: A three-branch encoder is used to encode the component gene sequence, defect gene sequence, and geometric gene vector respectively to obtain the feature vector of each branch; After concatenating the feature vectors of each branch, the weights of each modal channel are learned through a channel attention network; The material gene feature vector is compressed and output through a fully connected layer.

[0052] For example, the specific method of feature fusion can be a deep learning feature fusion algorithm. The fusion module includes a three-branch encoder and an SE channel attention layer. After the component and defect sequences are extracted by one-dimensional convolution, they are concatenated with the geometric vector MLP output. The channel weights of each modality are learned through channel attention, and then compressed into 16-dimensional features through a fully connected layer. The network is jointly trained with a physical constraint inverse model, and the loss function includes a gradation error term, a Bond energy consumption constraint term, and a feature norm regularization term.

[0053] First, perform branch coding: ingredient sequence : 1D-CNN (Conv1d, kernel=3, channels=32, ReLU) → GlobalAvgPool → ; Defect sequence : 1D-CNN (kernel=5, channels=32) → GAP → ; Geometric vectors :MLP(3→64→32, ReLU)→ ; Then channel attention fusion is performed: splicing =ReLU, =Sigmoid Then compressed using a fully connected connection: Final joint training loss: in, .

[0054] In some embodiments, the method for constructing the physical constraint inverse model in step S4 includes: Based on Bond's theory of crushing power consumption and particle size distribution evolution theory, a positive physical driving model is established, which describes the mapping relationship between process parameters and product performance. An objective function is constructed based on the positive physics-driven model, wherein the objective function is the weighted error between the prediction performance and the target performance plus a physical constraint penalty term; The objective function is iteratively optimized using the Newton-Raphson iterative method or the Levenberg-Marquardt algorithm to solve for the optimal process parameters in reverse.

[0055] For example, the specific method for constructing a physical constraint inverse model is as follows: 1. Establish a positive physics-driven model: Bond's Third Fracture Theory: in: Energy required to crush a unit mass of material (kWh / t) Bond work index (derived from materials genes) Component parameters (obtained through table lookup or regression) : Screen aperture size (mm) of 80% of the feed material passing through. : Screen aperture size (mm) of 80% of the output material. Gradation prediction model (Rosin-Rammler distribution): in: Particle size greater than Cumulative mass score Characteristic particle size (gap between discharge port and discharge port) (linear correlation) Uniformity coefficient (related to rotor speed) (Related) Construct the mathematical expression for the positive model: Let the process parameter vector be... Material gene vector Then the forward model The output is the prediction performance. : Specific calculation formulas for each component: (1) Characteristic particle size prediction in (Based on historical data) This represents the microcrack density index.

[0056] (2) Prediction of uniformity coefficient in , This is the angle coefficient.

[0057] (3) Crushing value prediction in .

[0058] (4) Energy consumption prediction This is the no-load loss constant.

[0059] 2. Construct an objective function based on the aforementioned positive physics-driven model, wherein the objective function is the weighted error between prediction performance and target performance plus a physics constraint penalty term; Material gene vector (Measured by the multimodal sensing unit) Target performance vector Find the optimal process parameters , so that: Where the objective function Defined as: Weighting coefficients (0.6, 0.3, 0.1 respectively) Physical constraint penalties (e.g., the rotational speed must not exceed the equipment's rated value). 3. The objective function is iteratively optimized using the Newton-Raphson iterative method or the Levenberg-Marquardt algorithm to solve for the optimal process parameters in reverse: Step 1: Initialize parameters (Match the closest historical case from the process parameter library) Step 2: Calculate the Jacobian matrix : The partial derivatives are obtained analytically from the forward model, for example: Step 3: Calculate the parameter update amount: in This is the damping factor (Levenberg-Marquardt algorithm), initially set to 0.01.

[0060] Step 4: Update parameters: The step size (0.5~1.0) is determined by line search.

[0061] Step 5: Check convergence conditions: If not satisfied, return to step 2; if satisfied, output... .

[0062] The complete example is shown below: (1) Input data: Material gene vector (Measured by multimodal sensing):

[0063] Target performance (Aggregate for C30 concrete):

[0064] (2) Reverse solution process: Initialization: Matching from the process parameter library First iteration: Forward calculation: error: Jacobian matrix, row 1: , Update quantity calculation: renew: , Second iteration: Forward calculation: error: renew: 3rd iteration: renew: 4th iteration: renew: 5th iteration: (Below target, slightly over-adjusted) renew: 6th iteration: (Error < 0.5mm, convergence) (3) Final output:

[0065] (4) Verification results: Actual production verification: Actual measurement (Prediction error 2.6%) Measured crushing value (Meets C30 requirements) Actual energy consumption per ton (Prediction error 2.8%) In some embodiments, step S5 specifically includes: Before the material enters the crusher, feedforward control is performed according to the optimal crushing process parameters to preset the discharge port opening and rotor speed of the crushing equipment; During the crushing process, the opening of the crusher discharge port or the speed of the crusher is dynamically fine-tuned based on real-time operating parameters through an adaptive adjustment algorithm. The real-time operating parameters include current load, vibration signal and acoustic emission signal, and the adaptive adjustment algorithm includes a fuzzy PID control algorithm.

[0066] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0067] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for intelligent preparation of recycled aggregate based on reverse modeling of materials genes, characterized in that, Includes the following steps: S1: Obtain the multimodal material gene characteristics of the construction waste to be processed, wherein the multimodal material gene characteristics include composition gene characteristics, internal defect gene characteristics and geometric gene characteristics; S2: The component gene features, internal defect gene features and geometric gene features are fused to generate a material gene feature vector; S3: Obtain the performance index requirements of the target recycled aggregate and construct the target performance vector; S4: Input the material gene feature vector and target performance vector into the physical constraint inverse model, and solve in reverse to obtain the optimal crushing process parameters; S5: Control the crushing equipment to crush the construction waste feed according to the optimal crushing process parameters to obtain recycled aggregate.

2. The method according to claim 1, characterized in that, Step S1, obtaining the multimodal material gene characteristics, specifically includes: Near-infrared spectroscopy (NIRS) and / or laser-induced breakdown spectroscopy (LIBS) are used to identify the proportion of different materials in the feed in real time and obtain the genetic characteristics of the components. Acoustic emission (AE) sensors and / or microwave dielectric sensors are used to quantitatively analyze the microcrack density, porosity, and primary strength of the feed material, thereby obtaining the internal defect gene characteristics. A 3D vision sensor and / or laser profile scanner are used to obtain the particle size distribution and particle shape characteristics of the feed material, and to obtain the geometric gene characteristics.

3. The method according to claim 1, characterized in that, The specific method for feature fusion in step S2 is as follows: Instance-level normalization was performed on the component gene subvectors, defect gene subvectors, and geometric gene subvectors, respectively. A weighted average fusion algorithm is adopted, and each sub-vector is weighted and concatenated according to the confidence adaptive weight. The confidence adaptive weight is updated online in an exponential smoothing manner according to the prediction error of each modality on the target gradation. Output the material gene feature vector.

4. The method according to claim 1, characterized in that, The specific method for feature fusion in step S2 is as follows: A three-branch encoder is used to encode the component gene sequence, defect gene sequence, and geometric gene vector respectively to obtain the feature vector of each branch; After concatenating the feature vectors of each branch, the weights of each modal channel are learned through a channel attention network; The material gene feature vector is compressed and output through a fully connected layer.

5. The method according to claim 1, characterized in that, The method for constructing the physical constraint inverse model in step S4 includes: Based on Bond's theory of crushing power consumption and particle size distribution evolution theory, a positive physical driving model is established, which describes the mapping relationship between process parameters and product performance. An objective function is constructed based on the positive physics-driven model, wherein the objective function is the weighted error between the prediction performance and the target performance plus a physical constraint penalty term; The objective function is iteratively optimized using the Newton-Raphson iterative method or the Levenberg-Marquardt algorithm to solve for the optimal process parameters in reverse.

6. The method according to claim 1, characterized in that, Step S5 specifically includes: Before the material enters the crusher, feedforward control is performed according to the optimal crushing process parameters to preset the discharge port opening and rotor speed of the crushing equipment; During the crushing process, the discharge port opening or rotor speed of the crushing equipment is dynamically fine-tuned based on real-time operating parameters using an adaptive adjustment algorithm. The real-time operating parameters include current load, vibration signal, and acoustic emission signal, and the adaptive adjustment algorithm includes a fuzzy PID control algorithm.

7. A smart preparation system for recycled aggregates based on reverse modeling of materials genes, characterized in that, include: The multimodal material gene sensing unit is used to collect the component gene features, internal defect gene features and geometric gene features of the input material, and fuse the multimodal features to generate a unified material gene feature vector; The intelligent reverse decision-making unit is used to receive the performance index requirements of the target recycled aggregate and solve the optimal crushing process parameters in reverse through the physical constraint reverse model. An adaptive execution control unit is used to control the crushing equipment to perform crushing processing according to the optimal crushing process parameters.

8. The system according to claim 7, characterized in that, The multimodal material gene sensing unit includes: The component feature recognition module uses near-infrared spectroscopy (NIRS) and / or laser-induced breakdown spectroscopy (LIBS) technology to identify the proportion of different materials in the feed in real time. The internal defect sensing module uses acoustic emission (AE) sensors and / or microwave dielectric sensors to quantitatively analyze the microcrack density, porosity, and primary strength of the feed material. The geometric feature acquisition module uses a 3D vision sensor or a laser contour scanner to acquire the particle size distribution and particle shape characteristics of the feed material. The data fusion module integrates multi-source heterogeneous sensing information into a unified material gene feature vector.

9. The system according to claim 7, characterized in that, The intelligent reverse decision-making unit includes: The target performance input module is used to receive the target aggregate performance indicators set by the user. The physical constraint inverse model, based on the principles of material crushing mechanics and particle physics, solves in reverse the optimal combination of process parameters to satisfy the target performance. The dynamic optimization engine performs online self-learning and fine-tuning of model parameters based on real-time production feedback data; The process parameter library stores the historical optimal parameter combinations for different equipment models, which are used for model initialization and constraint boundary setting.

10. The system according to claim 7, characterized in that, The adaptive execution control unit includes: The feedforward control module presets the crusher parameters based on the results of the inverse model solution before the material enters the crusher. The adaptive adjustment module dynamically fine-tunes parameters based on real-time operating conditions during the crushing process; The status monitoring module monitors the equipment's operating status in real time. The actuators include jaw crushers, impact crushers, vertical shaft impact crushers, and their associated frequency conversion drives and hydraulic adjustment systems.

Images