Real-time monitoring and optimizing method and system for resource utilization of construction waste

Abstract

The invention relates to a real-time monitoring and optimizing method and system for resource utilization of construction waste. The method comprises the following steps: carrying out data acquisition on vibration frequencies, material images and weight data of a crusher, a screening machine and a sorting device through a multi-source sensor network to obtain an original parameter data set; performing feature extraction to obtain a standardized feature data set; establishing a crushing kinetic model, a screening efficiency model and a sorting process model, carrying out coupling through a material balance equation and an energy balance equation, and carrying out model parameter identification by adopting a least square method and a genetic algorithm to obtain a technological process mathematical model; inputting the current process parameter combination and the historical quality data into an epsilon-constraint optimizer to obtain a Pareto optimal solution of the resource conversion rate and the energy consumption; and calculating the control quantity of the execution mechanism through the model prediction controller, and generating an optimal process parameter combination. According to the implementation of the method, real-time optimization and adjustment of the process parameters for resource utilization of the construction wastes are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of process optimization, and in particular to a real-time monitoring and optimization method and system for resource utilization of construction waste. Background Art

[0002] Construction waste is one of the main solid wastes generated in the process of urban development. It has complex composition and huge output. If it is not handled properly, it will cause environmental pollution and waste of resources. At present, the resource utilization of construction waste has become an important way to solve this problem. However, in the actual treatment process, due to the uncertainty of material composition and properties, it is difficult to accurately control the process parameters of the treatment equipment.

[0003] Traditional construction waste treatment methods mainly rely on manual experience to adjust process parameters, lack real-time monitoring and optimization methods, and cannot respond to changes in material properties in a timely manner, resulting in low resource conversion efficiency and energy waste. At the same time, there is a complex coupling relationship between the multiple processes in the treatment process, and the operating parameters of each process affect each other, making it difficult to find the optimal combination of process parameters. In addition, existing construction waste treatment equipment generally has problems such as single monitoring methods and insufficient data analysis capabilities, and is unable to conduct comprehensive status monitoring and quality control of the treatment process, resulting in large fluctuations in product quality and unstable resource utilization effects. Summary of the invention

[0004] The main purpose of the present invention is to provide a real-time monitoring and optimization method and system for resource utilization of construction waste. The present invention realizes real-time optimization and adjustment of process parameters for resource utilization of construction waste.

[0005] To achieve the above object, the present invention provides a real-time monitoring and optimization method for resource utilization of construction waste, comprising the following steps: Based on the current process parameter combination, construction waste is processed for resource use, and the vibration frequency, material image and weight data of the crusher, screener and sorting device are collected through a multi-source sensor network to obtain the original parameter data set; Inputting the original parameter data set into a multidimensional feature extraction model to perform feature extraction and feature standardization processing to obtain a standardized feature data set; A crushing dynamics model, a screening efficiency model and a sorting process model are established according to the standardized characteristic data set, and the material balance equation and the energy balance equation are coupled to identify the model parameters by using the least square method and the genetic algorithm to obtain a mathematical model of the process; Performing autoregressive analysis based on the mathematical model of the process, and inputting the current process parameter combination and historical quality data into an ε-constrained optimizer to obtain a Pareto optimal solution for resource conversion rate and energy consumption; A control strategy is generated based on the Pareto optimal solution, and the control amount of the actuator is calculated through a model predictive controller. The process parameters are adjusted in combination with feedforward compensation and feedback compensation to generate an optimal process parameter combination.

[0006] The present invention also provides a real-time monitoring and optimization system for resource utilization of construction waste, comprising: The acquisition module is used to process the construction waste resources based on the current process parameter combination, and collect the vibration frequency, material image and weight data of the crusher, screener and sorting device through a multi-source sensor network to obtain the original parameter data set; A feature extraction module is used to input the original parameter data set into a multidimensional feature extraction model to perform feature extraction and feature standardization processing to obtain a standardized feature data set; Establishing a module for establishing a crushing dynamics model, a screening efficiency model and a sorting process model according to the standardized characteristic data set, coupling the material balance equation and the energy balance equation, using the least square method and genetic algorithm to identify the model parameters, and obtaining a process mathematical model; An analysis module, for performing autoregressive analysis based on the mathematical model of the process, and inputting the current process parameter combination and historical quality data into an ε-constrained optimizer to obtain a Pareto optimal solution of resource conversion rate and energy consumption; A generation module is used to generate a control strategy based on the Pareto optimal solution, calculate the control amount of the actuator through a model predictive controller, adjust the process parameters in combination with feedforward compensation and feedback compensation, and generate an optimal process parameter combination.

[0007] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0008] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0009] In summary, the technical solution provided by the present invention realizes comprehensive monitoring of key processes such as crushing, screening, and sorting through a multi-source sensor network, obtains multi-dimensional data such as equipment vibration frequency, material image and weight, and solves the problem of incomplete information acquisition in the traditional single monitoring method; a multi-dimensional feature extraction model is used to process the original parameter data, and a multi-head attention mechanism and a graph convolutional network are combined to extract deep features, thereby overcoming the defect that the traditional feature extraction method has insufficient processing capacity for multi-source heterogeneous data; a process mathematical model including a crushing dynamics model, a screening efficiency model and a sorting process model is established, and multi-process coupling modeling is realized through material balance equations and energy balance equations; based on autoregressive analysis and ε-constrained optimization methods, multi-objective optimization of resource conversion rate and energy consumption is realized, and the Pareto optimal solution set is obtained, which solves the problem that traditional single-objective optimization cannot balance multiple performance indicators; a model predictive controller is used in combination with a feedforward compensation and feedback compensation control strategy to realize real-time optimization and adjustment of process parameters, thereby overcoming the shortcoming of the traditional control method in responding laggingly to changes in material properties. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a schematic diagram of the steps of a real-time monitoring and optimization method for resource utilization of construction waste in one embodiment of the present invention; Figure 2 It is a structural block diagram of a real-time monitoring and optimization system for resource utilization of construction waste in one embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0011] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0012] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0013] Reference Figure 1 This embodiment provides a real-time monitoring and optimization method for resource utilization of construction waste, comprising the following steps: S1, based on the current process parameter combination, construction waste is processed for resource use, and the vibration frequency, material image and weight data of the crusher, screener and sorting device are collected through a multi-source sensor network to obtain the original parameter data set; Among them, construction waste resource processing is carried out based on the current process parameter combination. During the processing process, vibration sensors are installed at the main bearing seat positions of the crusher, screener and sorting device to collect the vibration signals of the equipment in real time, obtain the vibration frequency data of the equipment, and directly reflect the operation status of the equipment and the load change. At the same time, in order to fully obtain the dynamic characteristics of the material, high-precision image sensors are installed at the key positions of the material conveying channel, and the image of the material is collected in a non-contact manner to generate material morphology data that can describe the particle size, shape and distribution characteristics of the material. This data can intuitively reflect the crushing and screening effects of the material and assist in identifying abnormal characteristics in the material. In order to improve the grasp of the material flow characteristics, a weight sensor is installed on the conveyor belt to measure the weight data of the material in real time, calculate the flow information of the material, and characterize the dynamic changes in the material conveying and processing process. Based on the equipment vibration frequency data, material morphology data and material flow data, a sensor calibration curve database is established. By experimentally calibrating the output characteristics of the sensor, the response curve and deviation parameters of the sensor are calculated and stored, and a calibration parameter matrix is ​​generated. This matrix can quantify the deviation relationship between the sensor output signal and the actual physical quantity, ensuring the accuracy and reliability of the data acquisition process. After the sensor calibration is completed, the calibration parameter matrix is ​​used to perform compensation calculation and data integration on the collected equipment vibration frequency data, material shape data, and material flow data. The compensation calculation process corrects the deviation of the sensor output based on the calibration parameters to make it closer to the actual value; data integration matches and synthesizes the compensation data of multi-source sensors according to the time axis and material flow path to form an original parameter data set containing rich physical information.

[0014] S2, inputting the original parameter data set into the multidimensional feature extraction model for feature extraction and feature standardization to obtain a standardized feature data set; Specifically, the equipment vibration frequency data in the original parameter data set is input into the first graph convolution layer of the multidimensional feature extraction model. The first graph convolution layer performs convolution operation on the data through 8 convolution kernels of size 5×5, thereby extracting the key features in the vibration data and generating a vibration feature mapping matrix. This matrix can capture the complex dynamic characteristics of the equipment during operation, including frequency changes and amplitude patterns. At the same time, the material morphology data in the original parameter data set is input into the second graph convolution layer of the multidimensional feature extraction model, which contains 16 convolution kernels of size 3×3. By extracting and enhancing the local features of the material image data, an image feature mapping matrix is ​​generated to reveal the fine features of the material particle size, shape and distribution, and help understand the morphological characteristics of the material and the potential processing effect. The material flow data in the original parameter data set is input into the third graph convolution layer, which is equipped with 32 convolution kernels of size 7×7 to capture the time series characteristics and trend changes of the material flow data. Through the convolution operation of this layer, a weight feature mapping matrix is ​​generated to reflect the dynamic changes and circulation status of the material flow. The vibration feature mapping matrix, image feature mapping matrix and weight feature mapping matrix are input into the feature fusion layer in the multi-dimensional feature extraction model. In this layer, the weights of different modal features are calculated by adopting the multi-head attention mechanism to form a multi-modal feature fusion matrix. The introduction of the multi-head attention mechanism can dynamically allocate the weights of different features to highlight the contribution of each modal feature in a specific task, while effectively reducing the risk of feature redundancy and information loss. The multi-modal feature fusion matrix is ​​input into a deep neural network composed of three layers of fully connected layers for processing. Among them, the first fully connected layer contains 512 neurons, the second layer contains 256 neurons, and the third layer contains 128 neurons. Each layer uses the LeakyReLU activation function to enhance the nonlinear representation ability. In this process, the model can extract higher-level abstract information from the fused features and generate the target feature vector. In order to improve the applicability and consistency of the data, the target feature vector is standardized to eliminate the differences in scale and unit, which is convenient for subsequent analysis. At the same time, in order to enhance the ability to capture different frequency domain features, multi-scale wavelet decomposition is used to decompose and reconstruct the target feature vector to generate the final standardized feature data set.

[0015] S3, based on the standardized characteristic data set, the crushing dynamics model, the screening efficiency model and the sorting process model are established, the material balance equation and the energy balance equation are coupled, and the least square method and genetic algorithm are used to identify the model parameters to obtain the mathematical model of the process; It should be noted that the crushing process data in the standardized feature data set is analyzed in time series, and the kinetic equation describing the crushing process is constructed based on the material particle size distribution function and the crushing power function. In this process, the Bond power calculation formula is used to evaluate the crushing energy consumption per unit time, accurately quantify the energy consumption characteristics of the crushing equipment, and generate a crushing kinetic model, which can reflect the relationship between the change in material particle size and the crushing energy consumption. The screening process data in the standardized feature data set is analyzed probabilistically, and the passing rate of the material through the screen is calculated by the Rosin-Rammler particle size distribution function, and the kinetic equation describing the screening process is established in combination with the material stacking thickness function. The screening efficiency model reveals the comprehensive influence of the screen opening rate, vibration frequency and material flow rate on the screening effect. At the same time, the sorting process data in the standardized feature data set is analyzed by trajectory, and the motion trajectory of the material in the sorting device is calculated based on the multi-body system motion equation. Combined with the material characteristics and device parameters, the sorting kinetic equation is constructed through the separation function to generate a sorting process model. This model can quantify the sorting accuracy and evaluate the separation effect of materials with different particle sizes or densities. The output parameters of the crushing dynamics model are combined with the input parameters of the screening efficiency model to establish the material balance equation. By calculating the material flow rate of each process, the iterative method is used to solve the material balance constraint relationship to form a material balance equation group. This process can ensure the consistency of material flow between different processes and ensure that the physical meaning of the process model is clear. According to the energy consumption parameters of the crushing dynamics model, the screening efficiency model and the sorting process model, the energy balance equation is constructed, the energy transfer efficiency of each process is calculated, and the energy balance constraint relationship is solved by the matrix method to obtain the energy balance equation group. The introduction of the energy balance equation can effectively describe the energy distribution and conversion efficiency in the process flow, providing a basis for optimizing energy consumption. The material balance equation group and the energy balance equation group are integrated into the objective function, and the optimization variables, including crushing ratio, screening efficiency and sorting accuracy, are set. The model parameters are optimized by genetic algorithm. In the optimization process, the genetic algorithm simulates the mechanism of natural selection and genetic variation to perform a global search on the objective function to obtain the optimal solution of the model parameters. On this basis, the least squares method is used to further fit the optimized solution of the model parameters to ensure the accuracy and stability of parameter identification. Based on the model parameter identification results, the parameters of the crushing dynamics model, screening efficiency model and sorting process model are updated. Combined with the updated model, the coupling relationship between the process parameters is calculated through the material balance equation group and the energy balance equation group to generate a process mathematical model. This model can fully describe the dynamic behavior and mutual influence of each process in the process of resource utilization of construction waste.

[0016] S4, based on the process mathematical model, autoregressive analysis is performed, and the current process parameter combination and historical quality data are input into the ε-constrained optimizer to obtain the Pareto optimal solution of resource conversion rate and energy consumption; Specifically, based on the mathematical model of the process, the resource conversion rate objective function and the energy consumption objective function are calculated, and the two objective functions are weighted and combined to construct a multi-objective optimization function. In this process, the resource conversion rate objective function evaluates the process effect by measuring the resource recovery efficiency in the process of resource utilization of construction waste, while the energy consumption objective function evaluates the system energy efficiency by calculating the energy usage per unit resource conversion in the process. The weighted combination of multi-objective optimization functions dynamically adjusts the weights of different optimization objectives to achieve a balance between resource efficiency and energy consumption. After constructing the optimization function, based on the current process parameter combination, a p-order autoregressive model is established to analyze the time series characteristics of the process parameters. The order of the autoregressive model is evaluated and optimized by using the Akaike information criterion to determine the optimal autoregressive model structure. After determining the model structure, the historical quality data is input into the autoregressive model, and a complete autoregressive prediction model is generated by calculating the autoregressive coefficient and the white noise variance. This prediction model can use the current process parameter combination and historical data to make rolling predictions for future process parameter changes and obtain a series of process parameter prediction sequences. The process parameter prediction sequence is input into the ε-constrained optimizer, and the constraint interval of resource conversion rate and the constraint interval of energy consumption are set. The setting of the constraint interval is based on the actual process requirements and performance standards to ensure that the optimization results meet both the quality requirements of resource utilization and the goal of energy consumption control. During the optimization process, the ε-constrained optimizer comprehensively scans the process parameter space through the grid search method to construct the corresponding constraint matrix. The matrix contains the optimization information of each parameter combination under different constraint conditions. The constraint matrix is ​​segmented and solved, and the segmented optimal solution set is sorted by the non-dominated sorting method. The non-dominated sorting effectively distinguishes the better solutions in the decomposition set, and selects elite individuals based on the areas with higher congestion to generate a non-dominated solution set. The solutions in the non-dominated solution set represent the optimization results between resource conversion rate and energy consumption under different trade-off conditions, ensuring that the optimization solution can achieve a balance between multiple objectives. After obtaining the non-dominated solution set, in order to improve the accuracy and stability of the optimization solution, the simulated annealing strategy is used to perform local search optimization on the neighborhood of the solution. The simulated annealing strategy simulates the temperature change in the physical annealing process and accepts poor solutions with a certain probability, thereby jumping out of the local optimal trap and improving the global search capability. In the process of searching for the best solution in the neighborhood of the solution, it gradually converges to the Pareto optimal solution of resource conversion rate and energy consumption.

[0017] The process parameter prediction sequence is input into the ε-constrained optimizer, and a normality test is performed to evaluate the distribution characteristics of the parameters. The normality test of the process parameters is used to determine whether the prediction sequence conforms to the normal distribution or other distribution forms. Based on the obtained parameter distribution characteristics, the process parameter prediction sequence is normalized, and different parameters are mapped to a unified standardized range to eliminate the differences in data scales and generate a standardized prediction sequence. On this basis, the standardized prediction sequence is used to calculate the constraint boundary. By analyzing the resource conversion rate target and combining the process flow characteristics, the constraint interval of the resource conversion rate is calculated and determined; at the same time, the energy consumption analysis of the standardized prediction sequence is performed, and the constraint interval of energy consumption is calculated by quantifying the energy usage characteristics of each process. The determination of these two constraint intervals provides clear upper and lower boundaries for the optimization process, ensuring that the physical meaning of the optimization solution is reasonable and meets the requirements of practical applications. After the constraint interval is determined, grid points are generated based on the resource conversion rate constraint interval and the energy consumption constraint interval, and the continuous constraint space is discretized into a finite number of discrete constraint point sets. The generation of discrete constraint point sets needs to evenly cover the entire constraint space, thereby providing a comprehensive search range for subsequent optimization. The discrete constraint point set is input into the ε vector generator, and the ε constraint vector is generated by equally spaced sampling. The generation of the ε constraint vector is based on the distribution between points in the discrete constraint point set, and the equally spaced method ensures that the optimizer can explore all possible solutions in the constraint space with a uniform step size. The generated ε constraint vector and the discrete constraint point set are combined, and a complete constraint matrix is ​​formed by matching each constraint vector with the corresponding constraint point. This constraint matrix not only contains the constraint boundary information of resource conversion rate and energy consumption, but also reflects the optimization potential of the process parameter prediction sequence under different constraint conditions. Through this matrix, the ε-constrained optimizer can efficiently perform optimization search in the constraint space, ensuring that the optimization results not only meet the boundary requirements of resource conversion rate and energy consumption, but also achieve global optimality in the objective function.

[0018] S5, generates a control strategy based on the Pareto optimal solution, calculates the control quantity of the actuator through the model predictive controller, adjusts the process parameters in combination with feedforward compensation and feedback compensation, and generates the optimal process parameter combination.

[0019] Among them, the Pareto optimal solution is input into the model predictive controller. In the model predictive controller, the quadratic objective function is used to construct the control performance index, and the objective function is used to balance the relationship between resource conversion rate and energy consumption, while minimizing the control error and control increment. By analyzing the dynamic characteristics of the step response, the control increment sequence of the system response is determined, which reflects the adjustment strategy of the actuator at different time steps. Based on the control increment sequence, a feedforward compensation model is constructed, and the key process parameters of the crusher, screener and sorting device are feedforward compensated by analyzing the impact of system input changes on output performance. These parameters include the speed of the crusher, the vibration frequency of the screener and the sorting ratio of the sorting device. The parameter matrix obtained by the feedforward compensation calculation can pre-correct the potential disturbance effect, improve the system's ability to suppress input disturbances, and generate a feedforward compensation parameter matrix. The feedforward compensation parameter matrix is ​​input into the feedback compensator to dynamically correct the deviation between the actual output and the expected output of the system. By collecting the equipment operation data in real time and comparing it with the output predicted by the model, the feedback compensator calculates the feedback compensation amount of each actuator to correct the deviation of the actual system, and finally obtains the feedback compensation parameter matrix. The introduction of feedback compensation significantly improves the stability and robustness of the system, ensuring that the process adjustment can adapt to complex dynamic environments. After the feedforward compensation parameter matrix and the feedback compensation parameter matrix are generated, the two are organically combined through compensation control calculation to generate a compensation control parameter matrix. This matrix integrates feedforward and feedback information, and can coordinate the predictability and real-time performance of the control strategy. On the basis of the compensation control parameter matrix, in order to optimize the adjustment range, reasonable constraints are set for the speed of the crusher, the vibration frequency of the screener, and the sorting ratio parameter of the sorting device. These constraints construct the parameter optimization interval to ensure that the final generated process parameter combination meets both the actual operation requirements and the equipment performance limitations. The compensation control parameter matrix is ​​combined with the parameter optimization interval, and the system parameters are further adjusted through the optimization solution method, and the constrained optimization algorithm is used to search for the optimal solution in the objective function. In this process, by balancing the priority of resource conversion rate and energy consumption, the key process parameters of the crusher, screener and sorting device are optimized, and finally the optimal process parameter combination is obtained. This parameter combination can minimize energy consumption while meeting the process efficiency requirements, ensuring that the resource utilization process of construction waste achieves the optimal performance.

[0020] In one example, construction waste is processed based on the current process parameter combination, and the vibration frequency, material image and weight data of the crusher, screener and sorting device are collected through a multi-source sensor network to obtain the original parameter data set, including: Based on the current process parameter combination, construction waste is processed for resource use. Vibration sensors are installed at the main bearing seats of crushers, screeners, and sorting devices to collect vibration signals and obtain equipment vibration frequency data. An image sensor is installed at the material conveying channel to collect material images and obtain material shape data, and a weight sensor is installed on the conveyor belt to collect material weight and obtain material flow data; Based on the equipment vibration frequency data, material shape data and material flow data, a sensor calibration curve database is established, and calibration parameter calculation is performed to obtain a calibration parameter matrix; Based on the calibration parameter matrix, compensation calculation and data integration are performed on the equipment vibration frequency data, material shape data and material flow data to obtain the original parameter data set.

[0021] In this example, vibration sensors are installed at the main bearing seats of the crusher, screener, and sorting device to collect vibration signals during the operation of the equipment. These signals can reflect the working status of the equipment and the impact characteristics of the material, and generate equipment vibration frequency data. Vibration signals are analyzed in the time domain and frequency domain to obtain key features. For example, through fast Fourier transform, time series signals are converted into Transform to frequency domain signal , and its calculation formula is: ; in, represents the frequency domain signal, For time, is the frequency, is an imaginary unit. Through fast Fourier transform, the main frequency component in the equipment vibration signal is extracted to analyze the stability and abnormal state of equipment operation. At the same time, a high-precision image sensor is installed at the material conveying channel to collect real-time images of the conveyed materials and generate material morphology data. The image data is extracted through the edge detection algorithm to extract the particle size and shape characteristics of the material. For example, based on the Canny edge detection algorithm, the edge intensity of the material morphology It is calculated by gradient, and its formula is: ; in, is the image gray value, and are the horizontal and vertical coordinates of the image, respectively. By counting the edge features, the particle size distribution and shape coefficient are calculated. In order to obtain material flow data, a weight sensor is installed on the conveyor belt. According to Hawke's law, the weight sensor measures the change in the force exerted by the material on the elastic element of the sensor to obtain the real-time weight of the material. If the sensor output voltage signal is , then the flow By sensor sensitivity and material speed Calculated as: ; in, is the sensitivity coefficient of the sensor. By measuring the conveyor belt speed and the sensor output signal in real time, the instantaneous flow rate of the material can be accurately calculated. After completing the collection of vibration frequency, image and weight data, a sensor calibration curve database is established based on these data. The calibration process constructs a calibration model by experimentally measuring the deviation relationship between the actual output of the sensor and the theoretical value. For example, assuming that the output of the vibration sensor is , the actual vibration acceleration is , then the calibration curve is expressed by a linear model as: ; in, is the gain coefficient, is the offset. The calibration parameters are obtained by fitting the experimental data using the least squares method. and . Similarly, the image and weight sensors are calibrated to obtain the complete calibration parameter matrix , each element of the matrix corresponds to a calibration parameter group of a sensor. Based on the calibration parameter matrix, the collected data is compensated and integrated. For example, the compensation of vibration frequency data is achieved by the following formula: ; in, is the vibration acceleration after compensation, is the uncompensated sensor output data. Similarly, for the particle size distribution in the image data Flow and weight data Compensation is performed to generate a more consistent data set. By integrating the compensated vibration frequency data, material shape data, and material flow data, a complete original parameter data set is generated. For example, a data fusion algorithm is used to combine these multimodal data to construct a high-dimensional feature vector. : ; in, is the vibration characteristic, For morphological characteristics, is the traffic feature. After this fusion, the original parameter data set is obtained.

[0022] In one example, the original parameter data set is input into the multidimensional feature extraction model for feature extraction and feature standardization to obtain a standardized feature data set, including: The device vibration frequency data in the original parameter data set is input into the first graph convolution layer in the multidimensional feature extraction model, where the first graph convolution layer contains 8 5×5 convolution kernels to obtain a vibration feature mapping matrix; The material morphology data in the original parameter data set is input into the second graph convolution layer in the multidimensional feature extraction model, where the second graph convolution layer contains 16 3×3 convolution kernels to obtain an image feature mapping matrix; The material flow data in the original parameter data set is input into the third graph convolution layer in the multidimensional feature extraction model, where the third graph convolution layer contains 32 7×7 convolution kernels to obtain a weight feature mapping matrix; The vibration feature mapping matrix, the image feature mapping matrix and the weight feature mapping matrix are input into the feature fusion layer in the multi-dimensional feature extraction model, and the feature weights are calculated through the multi-head attention mechanism to obtain the multi-modal feature fusion matrix; The multimodal feature fusion matrix is ​​input into the three-layer fully connected layer in the multidimensional feature extraction model. The number of neurons in the first layer is 512, the number of neurons in the second layer is 256, and the number of neurons in the third layer is 128. The LeakyReLU activation function is used in each layer to obtain the target feature vector. The target feature vector is normalized and decomposed by multi-scale wavelet, and then a standardized feature data set is obtained by reconstruction.

[0023] In this example, the device vibration frequency data in the original parameter dataset is input into the first graph convolution layer in the multidimensional feature extraction model. The graph convolution layer is designed to contain 8 The convolution kernel is used to extract the spatial features in the vibration data through the local receptive field. Assume that the vibration data is represented as a two-dimensional matrix ,in is the time step, is the frequency component, the output feature map matrix of the convolution operation It is expressed as: ; in, For the The weight of the convolution kernel, is the bias term, is an activation function (e.g., ReLU). Through this convolution operation, features related to the dynamic characteristics of equipment operation, such as frequency fluctuation patterns and vibration intensity, are extracted from the vibration data. The material morphology data in the original parameter data set is input into the second graph convolution layer in the multidimensional feature extraction model. This convolution layer contains 16 graphs of size The convolution kernel is used to extract important features such as edges and textures in the material image. Assume that the material morphology data is a two-dimensional image matrix , the image feature map matrix obtained after the convolution operation The calculation formula is: ; in, and Respectively The weights and bias terms of the convolution kernels, The same is the activation function. The feature mapping matrix can capture the material size distribution and shape information. At the same time, the material flow data in the original parameter data set is input into the third graph convolution layer, which is designed to be 32 in size. The convolution kernel is used to extract the spatiotemporal features in the traffic data. Assume that the traffic data is a matrix ,in is the time step, Weight feature map matrix generated by convolution operation for the number of sensors Expressed as: ; in, For the The weight of the convolution kernel, is the bias term, is the activation function. Through this processing method, the periodic fluctuation characteristics of the flow and the dynamic mode of delivery are extracted. The vibration feature mapping matrix , image feature mapping matrix and the weight feature map matrix Input to the feature fusion layer, calculate the weight of each modal feature through the multi-head attention mechanism, and generate a multimodal feature fusion matrix. Suppose the input of the feature fusion layer is , the output is the fusion matrix , the calculation formula of the multi-head attention mechanism is: ; ; in, , , are the weight matrices for query, key, and value, respectively. is the dimension of the key, is the number of attention heads. The importance of each modal feature is dynamically allocated through the attention mechanism. The multimodal feature fusion matrix is ​​input into the three-layer fully connected layer network. The number of neurons in the first fully connected layer is 512, the second layer is 256, and the third layer is 128. Each layer uses the LeakyReLU activation function, and its nonlinear transformation formula is: ; in, is the negative slope coefficient, usually set to 0.01. After the nonlinear mapping of the fully connected layer, the target feature vector is generated. , which contains the comprehensive features of all modes. The target feature vector is normalized to eliminate the numerical scale difference, and the features are decomposed in the frequency domain by multi-scale wavelet decomposition. Suppose the input of wavelet decomposition is , the output is the decomposition coefficient and approximation coefficient , the decomposition formula is: ; in, is the wavelet function, is the scale function, is the decomposition level. By reconstructing the decomposed features, a standardized feature data set is generated for subsequent model training and optimization.

[0024] In one example, a crushing dynamics model, a screening efficiency model, and a sorting process model are established based on a standardized feature data set. The material balance equation and the energy balance equation are coupled, and the least squares method and genetic algorithm are used to identify the model parameters to obtain a mathematical model of the process, including: The crushing process data in the standardized feature data set is analyzed in time series, and the crushing dynamics equation is established based on the material particle size distribution function and the crushing power function. The Bond power calculation formula is used to calculate the crushing energy consumption per unit time, and the crushing dynamics model is obtained. The screening process data in the standardized feature data set were subjected to probability analysis, the material passing rate was calculated based on the Rosin-Rammler particle size distribution function, and the screening kinetic equation was established in combination with the material stacking thickness function to obtain the screening efficiency model. Conduct trajectory analysis on the sorting process data in the standardized feature data set, calculate the material sorting trajectory based on the multi-body system motion equation, construct the sorting dynamics equation through the separation function, and obtain the sorting process model; The material balance equation is established according to the output parameters of the crushing dynamics model and the input parameters of the screening efficiency model, the material flow rate of each process is calculated, and the material balance constraint relationship is solved by the iteration method to obtain the material balance equation group; The energy balance equation is established according to the energy consumption parameters of the crushing dynamics model, the screening efficiency model and the sorting process model, the energy transfer efficiency of each process is calculated, and the energy balance constraint relationship is solved by the matrix method to obtain the energy balance equation group; The material balance equations and energy balance equations are used to construct the objective function, and the optimization variables are set to include crushing ratio, screening efficiency and sorting accuracy. The parameters are optimized through genetic algorithm to obtain the optimal solution of model parameters. The least squares method is used to fit the parameters of the model parameter optimization solution to obtain the model parameter identification results. Based on the model parameter identification results, the parameters of the crushing dynamics model, screening efficiency model and sorting process model are updated. The coupling relationship between the process parameters is calculated through the material balance equation group and the energy balance equation group to obtain the mathematical model of the process.

[0025] In this example, the crushing process data involved in the standardized feature data set is analyzed in time series to establish a crushing dynamics model. The crushing dynamics model describes the energy consumption and particle size change characteristics of the material during the crushing process through the material particle size distribution function and the crushing power function. The particle size distribution is expressed by the Rosen-Rammler function as follows: ; in, For particles smaller than The percentage of materials is the characteristic particle size, is the distribution index, which describes the steepness of the particle size distribution. The crushing power consumption is calculated based on the Bond power formula: ; in, is the crushing energy consumption per unit time, is the Bond work index, The particle sizes of 80% of the materials before and after crushing are respectively. By combining the particle size distribution function with the Bond power formula, the crushing dynamics equation is established: ; in, is the fragmentation rate constant, The model describes the dynamic relationship between particle size change and energy consumption. After completing the crushing dynamics modeling, the screening process data is subjected to a probability analysis. The screening efficiency depends on the probability of the material passing through the screen, which is related to the Rosin-Rammler particle size distribution function. The material passing rate is expressed as: ; in, The particle size is The probability of the material passing through the screen, is the characteristic size of the sieve hole, is the screening index. Combined with the material stacking thickness function , the screening kinetic model is expressed as: ; in, is the distance during the screening process, is the screening efficiency coefficient. Through this model, the separation efficiency of the material particle size in the screening process is predicted. For the sorting process, the material sorting trajectory is calculated based on the motion equation of the multi-body system. Assume that during the sorting process, the material is under the action of gravity and fluid force, and the trajectory is described by the following equation: ; in, For material quality, is the material location, is gravity, is the fluid resistance, is the drag coefficient, is the fluid density, is the windward area of ​​the material, is the relative speed. The sorting efficiency is calculated by the separation function: ; in, is the separation degree, , are the average positions of the two materials, , is the standard deviation of the position. The sorting kinetic model describes the relationship between the material trajectory and the separation efficiency. The material balance equation is established by coupling the output parameters of the crushing kinetic model and the input parameters of the screening efficiency model. The material flow follows the continuity principle in each link and is expressed by the following formula: ; in, and are the input and output flow rates of each node respectively. The above equations are solved by iteration method to obtain the material balance constraint relationship and form a complete material balance equation set. On this basis, the energy balance equation is established by combining the energy consumption parameters of the crushing, screening and sorting models. Assume that the input energy of the system is , the output energy is , the intermediate energy loss is , then the energy conservation relationship is: ; The equations are solved by the matrix method, the energy transfer efficiency of each process is calculated, and a complete energy balance equation is formed. The objective function is constructed by combining the material balance equation and the energy balance equation. The objective function is optimized by the genetic algorithm with the crushing ratio, screening efficiency and sorting accuracy as optimization variables. The genetic algorithm finds the optimal solution through selection, crossover and mutation operations. After optimization, the least squares method is used to fit the parameter optimization solution to generate the final model parameter identification result, and the dynamic behavior of the crushing, screening and sorting links is recalculated based on the updated model parameters.

[0026] In one example, an autoregressive analysis is performed based on the process mathematical model, and the current process parameter combination and historical quality data are input into the ε-constrained optimizer to obtain the Pareto optimal solution of resource conversion rate and energy consumption, including: Based on the mathematical model of the process, the resource conversion rate objective function and the energy consumption objective function are calculated, and weighted combination is performed to obtain a multi-objective optimization function; Based on the current process parameter combination, a p-order autoregressive model is constructed, and the optimal order is determined by the AIC criterion to obtain the autoregressive model structure; Input the historical quality data into the autoregressive model structure, calculate the autoregressive coefficient and white noise variance, obtain the autoregressive prediction model, and make rolling predictions on the process parameters based on the autoregressive prediction model to obtain the process parameter prediction sequence; The process parameter prediction sequence is input into the ε-constrained optimizer, the resource conversion rate constraint interval and the energy consumption constraint interval are set, and a grid search is performed to obtain the constraint matrix; The constraint matrix is ​​segmented and optimized to obtain the segmented optimal solution set, and the segmented optimal solution set is non-dominated sorted, and the solution with larger congestion degree is selected as the elite individual to obtain the non-dominated solution set; The non-dominated solution set is locally searched and optimized, and the simulated annealing strategy is used to find the optimal solution in the neighborhood of the solution to obtain the Pareto optimal solution of resource conversion rate and energy consumption.

[0027] In this example, the resource conversion rate objective function and the energy consumption objective function are defined. These two objective functions describe the economic efficiency and energy consumption level of the process respectively, and have the characteristics of mutual restriction. The resource conversion rate objective function is expressed by the following formula: ; in, is the resource conversion rate, and are the mass of output and input materials, respectively, To optimize the variables, process parameters include crushing ratio, screening efficiency and sorting accuracy. The energy consumption objective function is expressed as: ; in, is the total energy consumption, For the The power of each process step, is the corresponding running time, is the total number of links. By weighted combination of these two objective functions, a multi-objective optimization function is constructed: ; in, and is the weight coefficient, which is used to adjust the balance between resource efficiency and energy consumption. Based on the current process parameter combination, a p-order autoregressive model is constructed to analyze the time series characteristics of the process parameters. The expression of the autoregressive model is: ; in, For the time series at time The value of is the autoregression coefficient, is white noise, is the model order. The optimal order is determined by the Akaike Information Criterion (AIC): ; in, is the likelihood function value of the model. The order with the smallest AIC value is selected as the optimal order to determine the autoregressive model structure. After obtaining the autoregressive model structure, the historical quality data is input into the model, and the autoregressive coefficient and white noise variance are calculated by the least squares method to generate an autoregressive prediction model. Based on this model, a rolling prediction of future process parameters is performed to generate a process parameter prediction sequence: ; in, For the moment The predicted process parameter sequence is input into -Constraint optimizer, set the constraint intervals of resource conversion rate and energy consumption respectively, and perform grid search to generate the constraint matrix. Grid search discretizes the variable space, covers all possible constraint combinations, and constructs constraints in the form of a multi-dimensional matrix. The constraint matrix is ​​segmented and optimized to obtain a series of segmented optimal solution sets. The segmented optimal solution set is screened by non-dominated sorting, and those solutions that have no obvious disadvantages in resource conversion rate and energy consumption targets are given priority. For areas with high congestion, more solutions are selected as elite individuals to generate non-dominated solution sets: ; Among them, F is the non-dominated solution set, and are the solution vectors, symbols Represents the dominance relationship. The simulated annealing strategy is used to perform local search optimization on the non-dominated solution set. Simulated annealing uses the following form of acceptance probability Hop search within a neighborhood: ; in, is the difference in objective function value between the current solution and the new solution, is the current temperature. By gradually lowering the temperature, the algorithm gradually converges to the Pareto optimal solution of resource conversion rate and energy consumption.

[0028] In one example, the process parameter prediction sequence is input into the ε-constrained optimizer, the resource conversion rate constraint interval and the energy consumption constraint interval are set, and a grid search is performed to obtain a constraint matrix, including: The process parameter prediction sequence is input into the ε-constrained optimizer for normality test to obtain the parameter distribution characteristics; Based on the parameter distribution characteristics, the process parameter prediction sequence is normalized to obtain a standardized prediction sequence, and the constraint boundary calculation is performed based on the standardized prediction sequence to obtain the resource conversion rate constraint interval. At the same time, the energy consumption analysis of the standardized prediction sequence is performed to obtain the energy consumption constraint interval; Generate grid points based on resource conversion rate constraint interval and energy consumption constraint interval to obtain discrete constraint point set, and input the discrete constraint point set into ε vector generator for equal-space sampling to obtain ε constraint vector; The constraint matrix is ​​obtained by combining the ε constraint vector and the discrete constraint point set.

[0029] In this example, a normality test is performed on the process parameter prediction sequence to obtain the distribution characteristics of the parameters. Suppose the process parameter prediction sequence is , the normality test is performed by the Shapiro-Wilk test, and the statistical calculation formula is: ; in, is the sorted sample value, is the sample mean, is a constant generated based on normal distribution. The test results are compared The significance level is used to determine whether the parameter follows a normal distribution. If it does not conform to a normal distribution, it is transformed into a normal distribution through methods such as Box-Cox transformation. The formula is: ; in, is the transformation parameter, which is determined by maximum likelihood estimation. After obtaining the parameter distribution characteristics, the process parameter prediction sequence is normalized to eliminate the scale effect of the data and improve the optimization efficiency. Normalization is achieved through the minimum-maximum normalization formula: ; in, is the normalized value, and are the minimum and maximum values ​​of the sequence respectively. After normalization, the standardized prediction sequence is obtained , whose range is unified as [0,1]. Based on the standardized prediction sequence, the constraint boundary calculation of resource conversion rate and energy consumption is performed. The constraint interval of resource conversion rate is determined by the mean of the prediction sequence. and standard deviation Determine, the constraint interval is expressed as: ; in, is the interval expansion factor, which is usually 2 or 3 to cover more than 95% of the sample values. Similarly, the energy consumption constraint interval is calculated, and the energy consumption interval is: ; in, and are the mean and standard deviation of the energy consumption prediction values, respectively. After obtaining the constraint intervals of resource conversion rate and energy consumption, the intervals are discretized by the grid point generation method to obtain a discrete constraint point set. Assume that the interval is divided into uniform grid, the discrete points are expressed as: ; in, The discrete point set is all possible The discrete constraint point set is input into the ε vector generator for equal-spaced sampling to generate the ε constraint vector. Let the ε vector be ,in For equally spaced sampling points, the calculation formula is: ; in, , and is the minimum and maximum value of the ε vector. The ε constraint vector and the discrete constraint point set are combined to generate the constraint matrix. It is expressed as: ; Each matrix element represents a combination of a specific resource conversion rate, energy consumption, and ε value.

[0030] In one example, a control strategy is generated based on the Pareto optimal solution, and the control amount of the actuator is calculated through the model predictive controller. The process parameters are adjusted by combining feedforward compensation and feedback compensation to generate the optimal process parameter combination, including: The Pareto optimal solution is input into the model predictive controller, the control performance index is constructed using the quadratic objective function, and the step response analysis is performed to obtain the control increment sequence; Based on the control increment sequence, a feedforward compensation model is constructed, and the speed, vibration frequency and sorting ratio parameters of the crusher, screen and sorting device are calculated for feedforward compensation to obtain the feedforward compensation parameter matrix. Input the feedforward compensation parameter matrix into the feedback compensator, calculate the feedback compensation amount of each actuator, and obtain the feedback compensation parameter matrix; Perform compensation control calculation based on the feedforward compensation parameter matrix and the feedback compensation parameter matrix to obtain a compensation control parameter matrix; The compensation control parameter matrix is ​​optimized, and the constraint conditions are established by setting the crusher speed range, the screen vibration frequency range and the sorting ratio range of the sorting device to obtain the parameter optimization range. The optimization solution is performed based on the compensation control parameter matrix and the parameter optimization interval to obtain the optimal process parameter combination.

[0031] In this example, the Pareto optimal solution is passed as input to the model predictive controller to achieve multi-objective control. The model predictive controller constructs a control performance indicator through a quadratic objective function to minimize the sum of the squares of the target deviation and the control increment. The objective function is expressed as: ; in, is the objective function value, For the moment The system output, is the target reference value, To control the increment, To control the weight factor of the increment, is the prediction step length. By optimizing the objective function, the control increment sequence is calculated. , which describes the adjustment path of the system from the current state to the target state step by step. Based on the control increment sequence, a feedforward compensation model is constructed to perform feedforward compensation calculations on the key parameters of the crusher, screen and sorting device (such as speed, vibration frequency and sorting ratio). Assume that the input disturbance is , the current process parameters are The target process parameters are , the feedforward compensation relationship is expressed as: ; in, is the parameter value after feedforward compensation, is the feedforward compensation gain matrix. The feedforward compensation value of each device is calculated by combining the control increment sequence to generate the feedforward compensation parameter matrix , describes the pre-adjustment strategies for different devices. The feedforward compensation parameter matrix is ​​input into the feedback compensator to correct the system deviation in real time. The feedback compensator collects the actual output With the expected output The deviation between , calculate the feedback compensation. The feedback control relationship is expressed as: ; in, is the feedback compensation value, The feedback compensator generates the feedback compensation parameter matrix , which is used to correct the system deviation caused by nonlinear disturbances and modeling errors. After obtaining the feedforward compensation parameter matrix and the feedback compensation parameter matrix, the compensation control calculation is performed, and the two are combined to generate the final compensation control parameter matrix , the formula is: ; The matrix includes both the active adjustment of disturbances by feedforward compensation and the real-time correction of deviations by feedback compensation, which can ensure the stability and accuracy of the system under complex working conditions. The compensation control parameter matrix is ​​optimized to set the reasonable range of crusher speed, screen vibration frequency and sorting ratio of the sorting device, and to construct constraint conditions. Assume that the crusher speed range is , the vibration frequency range of the screening machine is , the sorting ratio range is , then the constraint condition is expressed as: ; Combined with these constraints, the parameter optimization range is determined. Based on the compensation control parameter matrix and the parameter optimization range, the optimization algorithm is used to find the optimal process parameter combination. The objective function of the optimization problem is expressed as: ; in, is the resource conversion rate, For energy consumption, and is the weight factor. Through the constrained optimization algorithm (such as genetic algorithm or particle swarm optimization algorithm), the solution space is searched within the parameter optimization interval, and finally the optimal process parameter combination that meets the constraint conditions is obtained.

[0032] Reference Figure 2 This embodiment provides a real-time monitoring and optimization system for resource utilization of construction waste, including: The acquisition module 1 is used to process the construction waste resources based on the current process parameter combination, and collect the vibration frequency, material image and weight data of the crusher, screener and sorting device through a multi-source sensor network to obtain the original parameter data set; Feature extraction module 2, used for inputting the original parameter data set into the multidimensional feature extraction model to perform feature extraction and feature standardization processing to obtain a standardized feature data set; Establishing module 3, which is used to establish a crushing dynamics model, a screening efficiency model and a sorting process model according to the standardized characteristic data set, coupling the material balance equation and the energy balance equation, using the least square method and genetic algorithm to identify the model parameters, and obtaining a mathematical model of the process; Analysis module 4 is used to perform autoregressive analysis based on the mathematical model of the process, and input the current process parameter combination and historical quality data into the ε-constrained optimizer to obtain the Pareto optimal solution of resource conversion rate and energy consumption; The generation module 5 is used to generate a control strategy based on the Pareto optimal solution, calculate the control amount of the actuator through the model predictive controller, adjust the process parameters in combination with feedforward compensation and feedback compensation, and generate the optimal process parameter combination.

[0033] In this embodiment, for the specific implementation of each unit in the above system embodiment, please refer to the above method embodiment, which will not be repeated here.

[0034] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0035] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0036] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0037] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0038] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0039] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A real-time monitoring and optimization method for resource utilization of construction waste, characterized in that: The following steps are involved: Based on the current process parameter combination, construction waste is processed for resource use, and the vibration frequency, material image and weight data of the crusher, screener and sorting device are collected through a multi-source sensor network to obtain the original parameter data set; Inputting the original parameter data set into a multidimensional feature extraction model to perform feature extraction and feature standardization processing to obtain a standardized feature data set; A crushing dynamics model, a screening efficiency model and a sorting process model are established according to the standardized characteristic data set, and the material balance equation and the energy balance equation are coupled to identify the model parameters by using the least square method and the genetic algorithm to obtain a mathematical model of the process; Performing autoregressive analysis based on the mathematical model of the process, and inputting the current process parameter combination and historical quality data into an ε-constrained optimizer to obtain a Pareto optimal solution for resource conversion rate and energy consumption; A control strategy is generated based on the Pareto optimal solution, and the control amount of the actuator is calculated through a model predictive controller. The process parameters are adjusted in combination with feedforward compensation and feedback compensation to generate an optimal process parameter combination.

2. The real-time monitoring and optimization method for resource utilization of construction waste according to claim 1 is characterized in that: The construction waste resource processing is carried out based on the current process parameter combination, and the vibration frequency, material image and weight data of the crusher, the screening machine and the sorting device are collected through the multi-source sensor network to obtain the original parameter data set, including: Based on the current process parameter combination, construction waste is processed for resource use. Vibration sensors are installed at the main bearing seats of crushers, screeners, and sorting devices to collect vibration signals and obtain equipment vibration frequency data. An image sensor is installed at the material conveying channel to collect material images and obtain material shape data, and a weight sensor is installed on the conveyor belt to collect material weight and obtain material flow data; Based on the equipment vibration frequency data, the material shape data and the material flow data, a sensor calibration curve database is established, and calibration parameter calculation is performed to obtain a calibration parameter matrix; Based on the calibration parameter matrix, compensation calculation and data integration are performed on the equipment vibration frequency data, the material shape data and the material flow data to obtain an original parameter data set.

3. The real-time monitoring and optimization method for resource utilization of construction waste according to claim 2 is characterized in that: The step of inputting the original parameter data set into a multidimensional feature extraction model to perform feature extraction and feature standardization processing to obtain a standardized feature data set includes: Inputting the device vibration frequency data in the original parameter data set into the first graph convolution layer in the multidimensional feature extraction model, the first graph convolution layer comprising 8 5×5 convolution kernels, to obtain a vibration feature mapping matrix; Inputting the material morphology data in the original parameter data set into the second graph convolution layer in the multidimensional feature extraction model, the second graph convolution layer comprising 16 3×3 convolution kernels, to obtain an image feature mapping matrix; Inputting the material flow data in the original parameter data set into the third graph convolution layer in the multidimensional feature extraction model, the third graph convolution layer comprising 32 7×7 convolution kernels, to obtain a weight feature mapping matrix; Input the vibration feature mapping matrix, the image feature mapping matrix and the weight feature mapping matrix into the feature fusion layer in the multi-dimensional feature extraction model, calculate the feature weights through a multi-head attention mechanism, and obtain a multimodal feature fusion matrix; Input the multimodal feature fusion matrix into the three fully connected layers in the multidimensional feature extraction model, wherein the number of neurons in the first fully connected layer is 512, the number of neurons in the second fully connected layer is 256, and the number of neurons in the third fully connected layer is 128, and each layer adopts the LeakyReLU activation function to obtain the target feature vector; The target feature vector is standardized and decomposed by multi-scale wavelet, and a standardized feature data set is obtained by reconstructing the target feature vector.

4. The real-time monitoring and optimization method for resource utilization of construction waste according to claim 3 is characterized in that: The crushing dynamics model, screening efficiency model and sorting process model are established according to the standardized characteristic data set, and the material balance equation and the energy balance equation are coupled, and the least square method and genetic algorithm are used to identify the model parameters to obtain the process mathematical model, including: Performing time series analysis on the crushing process data in the standardized feature data set, establishing a crushing dynamics equation based on the material particle size distribution function and the crushing power function, calculating the crushing energy consumption per unit time using the Bond power calculation formula, and obtaining a crushing dynamics model; Probabilistic analysis is performed on the screening process data in the standardized characteristic data set, the material passing rate is calculated based on the Rosin-Rammler particle size distribution function, and a screening kinetic equation is established in combination with the material stacking thickness function to obtain a screening efficiency model; Performing trajectory analysis on the sorting process data in the standardized feature data set, calculating the material sorting trajectory based on the multi-body system motion equation, constructing the sorting dynamics equation through the separation function, and obtaining the sorting process model; Establishing a material balance equation according to the output parameters of the crushing dynamics model and the input parameters of the screening efficiency model, calculating the material flow rate of each process and solving the material balance constraint relationship by an iterative method to obtain a material balance equation group; An energy balance equation is established according to the energy consumption parameters of the crushing dynamics model, the screening efficiency model and the sorting process model, the energy transfer efficiency of each process is calculated and the energy balance constraint relationship is solved by using a matrix method to obtain an energy balance equation group; The material balance equation group and the energy balance equation group are used to construct an objective function, and optimization variables including crushing ratio, screening efficiency and sorting accuracy are set, and parameter optimization is performed through a genetic algorithm to obtain a model parameter optimization solution; The least squares method is used to perform parameter fitting on the model parameter optimization solution to obtain model parameter identification results, and based on the model parameter identification results, the crushing dynamics model, the screening efficiency model and the sorting process model are updated. The coupling relationship between the process parameters is calculated through the material balance equation group and the energy balance equation group to obtain the mathematical model of the process.

5. The real-time monitoring and optimization method for resource utilization of construction waste according to claim 4 is characterized in that: The method of performing autoregressive analysis based on the mathematical model of the process and inputting the current process parameter combination and historical quality data into the ε-constrained optimizer to obtain the Pareto optimal solution of resource conversion rate and energy consumption includes: Based on the mathematical model of the process, the resource conversion rate objective function and the energy consumption objective function are calculated, and weighted combination is performed to obtain a multi-objective optimization function; Based on the current process parameter combination, a p-order autoregressive model is constructed, and the optimal order is determined by the AIC criterion to obtain the autoregressive model structure; Inputting historical quality data into the autoregressive model structure, calculating the autoregressive coefficient and the white noise variance, obtaining the autoregressive prediction model, and performing rolling prediction on the process parameters based on the autoregressive prediction model to obtain a process parameter prediction sequence; Input the process parameter prediction sequence into the ε-constraint optimizer, set the resource conversion rate constraint interval and the energy consumption constraint interval, perform grid search, and obtain the constraint condition matrix; The constraint matrix is ​​solved by piecewise optimization to obtain a piecewise optimal solution set, and the piecewise optimal solution set is sorted by non-dominated sorting, and solutions with a larger congestion degree are selected as elite individuals to obtain a non-dominated solution set; The non-dominated solution set is locally searched and optimized, and a simulated annealing strategy is used to find the optimal solution in the neighborhood of the solution to obtain the Pareto optimal solution of resource conversion rate and energy consumption.

6. The real-time monitoring and optimization method for resource utilization of construction waste according to claim 5 is characterized in that: The process parameter prediction sequence is input into the ε-constraint optimizer, the resource conversion rate constraint interval and the energy consumption constraint interval are set, and a grid search is performed to obtain a constraint condition matrix, including: Inputting the process parameter prediction sequence into the ε-constrained optimizer for normality test to obtain parameter distribution characteristics; Based on the parameter distribution characteristics, the process parameter prediction sequence is normalized to obtain a standardized prediction sequence, and constraint boundary calculation is performed based on the standardized prediction sequence to obtain a resource conversion rate constraint interval, and at the same time, energy consumption analysis is performed on the standardized prediction sequence to obtain an energy consumption constraint interval; Generate grid points based on the resource conversion rate constraint interval and the energy consumption constraint interval to obtain a discrete constraint point set, and input the discrete constraint point set into an ε vector generator for equal-spaced sampling to obtain an ε constraint vector; The ε constraint vector and the discrete constraint point set are combined to obtain a constraint condition matrix.

7. The real-time monitoring and optimization method for resource utilization of construction waste according to claim 6 is characterized in that: The control strategy is generated based on the Pareto optimal solution, and the control amount of the actuator is calculated by the model predictive controller, and the process parameters are adjusted in combination with feedforward compensation and feedback compensation to generate the optimal process parameter combination, including: The Pareto optimal solution is input into a model predictive controller, a quadratic objective function is used to construct a control performance index, and a step response analysis is performed to obtain a control increment sequence; Based on the control increment sequence, a feedforward compensation model is constructed, and feedforward compensation calculations are performed on the rotation speed, vibration frequency and sorting ratio parameters of the crusher, the screening machine and the sorting device to obtain a feedforward compensation parameter matrix; Inputting the feedforward compensation parameter matrix into the feedback compensator, calculating the feedback compensation amount of each actuator, and obtaining the feedback compensation parameter matrix; Perform compensation control calculation based on the feedforward compensation parameter matrix and the feedback compensation parameter matrix to obtain a compensation control parameter matrix; Optimizing the compensation control parameter matrix, constructing constraint conditions by setting the crusher speed range, the screening machine vibration frequency range and the sorting ratio range of the sorting device, and obtaining the parameter optimization interval; An optimization solution is performed based on the compensation control parameter matrix and the parameter optimization interval to obtain an optimal process parameter combination.

8. A real-time monitoring and optimization system for resource utilization of construction waste, characterized in that: For implementing the steps of the method according to any one of claims 1 to 7, the system comprises: The acquisition module is used to process the construction waste resources based on the current process parameter combination, and collect the vibration frequency, material image and weight data of the crusher, screener and sorting device through a multi-source sensor network to obtain the original parameter data set; A feature extraction module is used to input the original parameter data set into a multidimensional feature extraction model to perform feature extraction and feature standardization processing to obtain a standardized feature data set; Establishing a module for establishing a crushing dynamics model, a screening efficiency model and a sorting process model according to the standardized characteristic data set, coupling the material balance equation and the energy balance equation, using the least square method and genetic algorithm to identify the model parameters, and obtaining a process mathematical model; An analysis module, for performing autoregressive analysis based on the mathematical model of the process, and inputting the current process parameter combination and historical quality data into an ε-constrained optimizer to obtain a Pareto optimal solution of resource conversion rate and energy consumption; A generation module is used to generate a control strategy based on the Pareto optimal solution, calculate the control amount of the actuator through a model predictive controller, adjust the process parameters in combination with feedforward compensation and feedback compensation, and generate an optimal process parameter combination.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images