Solid waste resource collaborative disposal analysis method and system based on neural network optimization

Abstract

The embodiment of the application discloses a solid waste resource collaborative treatment analysis method and system based on neural network optimization, which comprises the following steps: collecting raw material component data of solid waste resources and performing sensing test analysis to generate sensing test analysis information; combining the sensing test analysis information and the raw material component data to generate solid waste collaborative treatment correlation data set, mining the correlation characteristics of solid waste components and collaborative treatment effects, and determining initial production process parameters; inputting the correlation data set and the initial production process parameters into a neural network intelligent decision model under the constraint of a preset maintenance condition, generating performance prediction results and harmful component leaching risk assessment results of a target product, and creating a multi-objective optimization function based on the performance prediction results and the harmful component leaching risk assessment results; iteratively optimizing the initial production process parameters through a particle swarm optimization algorithm to generate a plurality of sets of process parameter optimization candidate sets; and generating production process optimization parameters after verification by the neural network intelligent decision model, so that efficient solid waste resource collaborative treatment and multi-objective performance standard reaching are realized.

Description

Technical Field

[0001] This application relates to the field of computer data analysis technology, specifically to a method and system for the collaborative disposal analysis of solid waste resources based on neural network optimization. Background Technology

[0002] With the acceleration of urbanization and the popularization of waste-to-energy incineration technology, the amount of fly ash generated from municipal solid waste incineration is increasing year by year. Fly ash, containing dioxins and various heavy metals, is classified as hazardous waste, and its disposal is becoming increasingly serious. Currently, the mainstream disposal method is "cement solidification and stabilization + secure landfill," but this method suffers from high disposal costs, significant land resource consumption, and high long-term environmental risks. Meanwhile, industrial alkali slag and desulfurization gypsum, as common industrial solid wastes, also cause environmental pollution and resource waste due to their large-scale stockpiling.

[0003] Current technologies mostly utilize a combination of fly ash and slag or desulfurization gypsum for solid waste treatment, and generally rely on empirical ratios, lacking the ability to adaptively adjust to fluctuations in raw material composition. This results in poor product performance stability and difficulty in accurately controlling the risk of heavy metal leaching. Some existing technologies use both fly ash and steel slag as solid wastes, without introducing the synergistic activation effect of alkali slag and desulfurization gypsum, and rely on manually preset ratios, making dynamic optimization based on raw material composition impossible. Furthermore, alkali activation technology cannot integrate hazardous solid wastes like fly ash, and lacks an intelligent decision-making system, leading to long optimization cycles for process parameters.

[0004] Therefore, how to achieve efficient and collaborative disposal of solid waste resources and meet multi-objective performance standards is a technical problem that needs to be solved at present. Summary of the Invention

[0005] This application provides a method and system for the collaborative disposal analysis of solid waste resources based on neural network optimization.

[0006] This application provides a method for analyzing the collaborative disposal of solid waste resources based on neural network optimization, including:

[0007] Collect raw material composition data of solid waste resources and conduct sensor and chemical analysis to generate sensor and chemical analysis information containing the proportion and characteristics of solid waste components. By combining the sensor and laboratory analysis information with the raw material composition data, a solid waste co-processing correlation dataset is generated. Based on the correlation dataset, the correlation features between solid waste composition and co-processing effect are mined, and the initial production process parameters are determined according to the correlation features. Under the constraint of preset maintenance conditions, the associated dataset and the initial production process parameters are input into the neural network intelligent decision model. Cross-dimensional association processing is performed through the feature interaction layer, and the performance prediction results and leaching risk assessment results of the target product are generated through the decision output layer. A multi-objective optimization function is created by combining the mechanical performance characteristics and service durability characteristics in the performance prediction results and the leaching characteristics in the hazardous component leaching risk assessment results. The initial production process parameters are input as optimization variables into a preset particle swarm optimization algorithm. Guided by the multi-objective optimization function, the initial production process parameters are iteratively optimized and adjusted to generate multiple sets of process parameter optimization candidate sets. The neural network intelligent decision-making model is invoked to generate performance prediction optimization results and leaching risk assessment optimization results corresponding to each set of process parameter optimization candidate sets. Based on the performance prediction optimization results and the leaching risk assessment optimization results, production process optimization parameters including the target product ratio, optimized synergistic conditions, and material forming parameters are generated.

[0008] This application also provides a solid waste resource collaborative disposal analysis system, including: processor; Storage device, on which computer programs are stored, When the computer program is executed by the processor, the processor implements any of the described neural network-optimized solid waste resource collaborative disposal analysis methods.

[0009] This application embodiment also provides a readable storage medium storing a program or instructions, which, when executed by a processor, implement the steps of the neural network-optimized solid waste resource collaborative disposal analysis method.

[0010] This application's embodiments achieve a fundamental shift in the co-processing of solid waste resources from an experience-dependent model to a data-driven model. The overall innovation lies in the deep synergy and breakthroughs in the multi-stage technological logic: First, relying on high-precision solid waste component proportion and characteristic information generated by sensor analysis, it overcomes the blindness of process parameters caused by the lack of raw material characteristic data and insufficient precision in traditional solid waste treatment. Second, by mining the correlation features between solid waste components and co-processing effects through associated datasets and determining initial production process parameters, a direct mapping relationship between raw material characteristics and process parameters is established, avoiding the problem of disconnect between process parameters and raw material characteristics in traditional methods. Third, using preset maintenance conditions as constraints, through cross-dimensional correlation processing of the feature interaction layer of a neural network intelligent decision-making model, it simultaneously achieves target product performance prediction and hazardous component leaching risk assessment, breaking through the limitations of traditional single-viewpoint methods. This approach overcomes the limitations of standard prediction by achieving collaborative prediction of multi-dimensional indicators. Furthermore, it creates a multi-objective optimization function by combining performance prediction results with leaching risk assessment results, incorporating mechanical performance, durability, and environmental risk into a unified optimization framework. This solves the problem of neglecting certain aspects in traditional single-objective optimization. Finally, iterative optimization of initial production process parameters is achieved using a particle swarm optimization algorithm, generating multiple sets of candidate process parameters. These parameters are then verified by a neural network intelligent decision-making model to determine the final optimized production process parameters. This achieves globally optimal configuration of process parameters, ensuring that the target product meets mechanical performance and durability requirements while minimizing the risk of hazardous component leaching. Ultimately, this achieves a win-win situation of efficient collaborative disposal of solid waste resources, stable product performance compliance, and controllable environmental risks, improving the intelligence level and overall benefits of solid waste resource collaborative disposal. In this way, efficient collaborative disposal of solid waste resources and multi-objective performance compliance can be achieved. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating a method for co-processing solid waste resources based on neural network optimization, provided in an embodiment of this application.

[0012] Figure 2 This is a schematic diagram of the basic structure of a solid waste resource co-processing analysis system provided in an embodiment of this application.

[0013] Figure 3 This is a functional block diagram of a solid waste resource co-processing analysis device provided in an embodiment of this application. Detailed Implementation

[0014] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the embodiments of this application will be further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0015] See Figure 1As shown, this figure is a flowchart of a solid waste resource co-processing analysis method based on neural network optimization provided in an embodiment of this application. This method is executed through a solid waste resource co-processing analysis system. Figure 1 As shown, the method includes steps 110-160.

[0016] Step 110: Collect raw material composition data of solid waste resources and perform sensor analysis to generate sensor analysis information containing the proportion and characteristics of solid waste components.

[0017] In this embodiment, the solid waste resource collaborative disposal analysis system first collects batch samples of three types of solid waste raw materials—incineration fly ash, industrial alkali slag, and desulfurization gypsum—using an automated sampling device. Simultaneously, it collects real-time composition data of the raw materials using an integrated near-infrared spectroscopy sensing unit, X-ray fluorescence sensing unit, and ion-selective electrode sensing unit. The near-infrared spectroscopy sensing unit outputs a high-dimensional spectral data sequence containing organic component characteristics; the X-ray fluorescence sensing unit outputs a feature vector containing the types and corresponding proportions of inorganic elements; and the ion-selective electrode sensing unit outputs a bond-value pair array containing the concentration of soluble ions. The system then preprocesses the collected samples, removing impurities and homogenizing and crushing them. Finally, it uses analytical methods such as high-temperature incineration, acid-base digestion, and gas chromatography-mass spectrometry to determine indicators such as the proportion of combustible components, total heavy metal elements, and types and contents of organic pollutants in the solid waste raw materials, generating an analytical dataset. Next, the system uses a multi-source data fusion calibration algorithm to align and match the sensor detection data with the laboratory analysis data, eliminating the systematic errors of different detection methods. Finally, it generates sensor and laboratory analysis information containing the precise proportions, physicochemical properties, and occurrence forms of harmful components of incineration fly ash, industrial alkali residue, and desulfurization gypsum. This information is stored in the form of a multi-dimensional tensor, where each dimension of the tensor corresponds to a feature set of a type of solid waste raw material, and each element corresponds to the proportion or characteristic parameters of a specific component.

[0018] Step 120: Combine the sensor analysis information with the raw material composition data to generate a solid waste co-processing correlation dataset. Based on the correlation dataset, mine the correlation features between solid waste composition and co-processing effect, and determine the initial production process parameters based on the correlation features.

[0019] In this embodiment, the solid waste resource co-processing analysis system first aligns the sensor and laboratory analysis information output in step 110 with the original raw material composition data, and performs feature splicing along the solid waste type dimension to generate an initial dataset containing basic raw material attributes, component proportions, and characteristic parameters. Subsequently, the system employs an association rule mining algorithm to perform association analysis on key features in the initial dataset, such as the heavy metal content of incineration fly ash, the active ingredient content of industrial alkali slag, and the crystallization water content of desulfurization gypsum, with result features in historical co-processing data, such as compressive strength, durability, and heavy metal leaching concentration. This analysis uncovers a set of strongly correlated features between solid waste components and co-processing effects. This set is stored in the form of an adjacency list, where each node corresponds to a type of feature, and each edge corresponds to the association relationship and association strength between features. Next, based on the associated feature set and combined with the basic thermodynamic laws of solid waste synergistic reaction, the system determines the initial production process parameters, including the mixing ratio of incineration fly ash, industrial alkali residue and desulfurization gypsum, the synergistic activation temperature range, the synergistic activation time range, the mixing and molding pressure range, and the molding time range. The initial production process parameters are stored in the form of a structured dictionary, where each key corresponds to a type of process parameter and each value corresponds to the range of parameter values.

[0020] Step 130: Under the constraints of preset maintenance conditions, input the associated dataset and the initial production process parameters into the neural network intelligent decision model, perform cross-dimensional association processing through the feature interaction layer, and generate the performance prediction results and harmful component leaching risk assessment results of the target product through the decision output layer.

[0021] In this embodiment, the solid waste resource co-processing analysis system first matches corresponding preset curing condition boundary thresholds based on the component proportion characteristics of incineration fly ash, industrial alkali slag, and desulfurization gypsum, including curing temperature range, curing humidity range, and curing time range. These boundary thresholds are then converted into constraint parameters and embedded into the feature interaction layer of the neural network intelligent decision-making model. Subsequently, the system extracts solid waste characteristic data from the associated dataset, including the heavy metal content of incineration fly ash, the active ingredient content of industrial alkali slag, and the crystallization water content of desulfurization gypsum, as well as process data from the initial production process parameters, including synergistic excitation temperature and mixing molding pressure. The solid waste characteristic data and process data are dimensionally aligned to ensure feature dimension matching before being input into the feature interaction layer. The feature interaction layer employs a cross-dimensional association algorithm to fuse the solid waste characteristic data and process data, generating a fused feature vector containing component-process correlation relationships. Each element of this vector corresponds to a set of correlation features and correlation strength between solid waste components and process parameters. The system then inputs the fused feature vector into the decision output layer of the neural network intelligent decision-making model. The multi-task predictor built into the decision output layer generates performance prediction results and hazardous component leaching risk assessment results.

[0022] Step 131: Based on the compositional characteristics of incineration fly ash, industrial alkali slag, and desulfurization gypsum in the solid waste resources, match the corresponding preset maintenance condition boundary thresholds, and convert the preset maintenance condition boundary thresholds into constraint parameters and embed them into the feature interaction layer of the neural network intelligent decision-making model.

[0023] In this embodiment, the solid waste resource co-processing analysis system first extracts component proportion data such as the proportion of heavy metal occurrence in incineration fly ash, the active calcium oxide content in industrial alkali slag, and the calcium sulfate dihydrate content in desulfurization gypsum from the associated dataset. This data is stored in a multi-dimensional array, with each subarray corresponding to the component proportion characteristics of a type of solid waste. Subsequently, based on engineering practice standards for solid waste co-processing, the system matches corresponding preset curing condition boundary thresholds. Incineration fly ash corresponds to the upper temperature limit for high-temperature curing, industrial alkali slag to the upper humidity limit for high-humidity curing, and desulfurization gypsum to the lower time limit for room-temperature curing. These thresholds are represented as numerical ranges. Next, the system transforms the preset curing condition boundary thresholds into constraint parameters recognizable by the feature interaction layer. Through a dimension mapping algorithm, the threshold ranges are transformed into constraint tensors that match the dimensions of the solid waste characteristic data, and embedded into the feature fusion logic of the feature interaction layer to ensure that the feature fusion process meets the boundary requirements of the curing conditions.

[0024] Step 132: Extract solid waste characteristic data, including the heavy metal content of incineration fly ash, the active ingredient content of industrial alkali slag, and the crystal water content of desulfurization gypsum, from the associated dataset, as well as process data, including the synergistic excitation temperature and mixing molding pressure, from the initial production process parameters. After dimensional alignment processing of the solid waste characteristic data and the process data, input them into the feature interaction layer.

[0025] In this embodiment, the solid waste resource co-processing analysis system first extracts solid waste characteristic data from the associated dataset, including arrays of heavy metal content in incineration fly ash, active ingredient content in industrial alkali slag, and water of crystallization content in desulfurization gypsum. The dimensions of these arrays correspond to the number of solid waste samples. Simultaneously, the system extracts process data from the initial production process parameters, including arrays of synergistic excitation temperature ranges and mixing and molding pressure ranges. The dimensions of these arrays are consistent with the sample quantity dimension of the solid waste characteristic data. Subsequently, the system employs a dimension alignment algorithm to match the feature dimensions of the solid waste characteristic data with those of the process data. Features with inconsistent dimensions are either augmented or compressed to ensure that the number and dimensions of features corresponding to each sample of the two types of data are completely matched. Finally, the system concatenates the dimension-aligned solid waste characteristic data and the process data into an input tensor, which is then input into the feature interaction layer of the neural network intelligent decision-making model.

[0026] Step 133: The aligned solid waste characteristic data and process data are fused using the cross-dimensional association algorithm built into the feature interaction layer to generate a fused feature vector containing the component process relationship.

[0027] In this embodiment, the feature interaction layer of the solid waste resource collaborative disposal analysis system incorporates a cross-dimensional association algorithm. This algorithm first performs feature mapping on the aligned solid waste characteristic data and process data, mapping both types of data to the same high-dimensional feature space to eliminate dimensional differences between different features. Subsequently, the algorithm performs association calculations on the mapped features, identifying the association relationship between each component feature in the solid waste characteristic data and each process parameter feature in the process data, generating an association matrix. Each element of the matrix corresponds to a set of association strengths between components and process parameters. Next, the algorithm fuses the association matrix with the original solid waste characteristic data and process data through a feature weighted concatenation method, generating a fused feature vector. Each dimension of this vector corresponds to a set of component-process association features, and each element corresponds to the comprehensive strength value of the association features.

[0028] Step 134: Input the fused feature vector into the decision output layer of the neural network intelligent decision model, and generate mechanical performance features containing compressive strength data of the target product and service durability features containing impermeability and frost resistance data through the multi-task predictor built into the decision output layer as the performance prediction results, and generate leaching characteristic features containing heavy metal leaching concentration and leaching rate data as the leaching risk assessment results of the harmful components.

[0029] In this embodiment, the solid waste resource co-processing analysis system first inputs the fused feature vector output in step 133 into the feature enhancement module built into the decision output layer. The feature enhancement module constructs processing logic based on the correlation between components and processes during solid waste resource co-processing, performing dimensionality enhancement and noise filtering on the fused feature vector to generate an enhanced feature vector. This enhanced feature vector has a higher dimension than the fused feature vector and removes invalid noise components from the original features. Subsequently, the system synchronously inputs the enhanced feature vector into the mechanical performance prediction branch, durability prediction branch, and leaching risk prediction branch included in the multi-task predictor.

[0030] The mechanical performance prediction branch incorporates a compressive strength prediction sub-model built on a deep neural network. This sub-model is trained using over 100,000 sets of historical solid waste co-processing data. The training data input consists of the correlation features between solid waste components and process parameters from multiple past sets, and the output is the compressive strength data of the corresponding product. The sub-model optimizes network parameters using a backpropagation algorithm, using the mean absolute percentage error as an evaluation index to achieve early shutdown. During the inference phase, the sub-model, based on the component-process correlation information contained in the enhanced feature vector, outputs the compressive strength data of the target product under different curing cycles. This data is integrated into a mechanical performance feature, which is stored in a multi-dimensional array, with each element corresponding to the predicted compressive strength value for one curing cycle. The durability prediction branch incorporates a permeability-freeze-resistance coupled prediction sub-model. This sub-model introduces the correlation function between solid waste components and hydrological properties, and outputs the permeability coefficient and freeze-thaw cycle tolerance data corresponding to the permeability and freeze-thaw resistance of the target product based on the enhanced feature vector. This data is integrated into a durability feature, which is stored in a dictionary, with keys for permeability and freeze-thaw resistance, and values ​​for the corresponding data sequences.

[0031] The leaching risk prediction branch incorporates a sub-model for predicting heavy metal migration patterns. This sub-model combines the distribution characteristics of heavy metal speciation in incineration fly ash with the chemical mechanism of synergistic activation processes. Based on enhanced feature vectors, it outputs leaching concentration and leaching rate data of heavy metals in the target product under different environmental conditions. This data is integrated into leaching characteristic features, stored as multi-dimensional tensors, with each dimension corresponding to an environmental condition and each element corresponding to a set of heavy metal leaching concentration and leaching rate data. Finally, the system standardizes the data formats of mechanical performance characteristics, service durability characteristics, and leaching characteristic features. The standardized mechanical performance characteristics and service durability characteristics are integrated into the performance prediction results, while the standardized leaching characteristic features are identified as the hazardous component leaching risk assessment results.

[0032] Step 1341: Input the fused feature vector into the feature enhancement module built into the decision output layer. The feature enhancement module performs dimensionality enhancement and noise filtering on the fused feature vector to generate an enhanced feature vector. The processing logic built into the feature enhancement module is based on the correlation between components and processes in the co-processing of solid waste resources. It is used to enhance the correlation between the heavy metal content of incineration fly ash, the active ingredient content of industrial alkali slag, the crystal water content of desulfurization gypsum, and the co-excitation temperature and mixing molding pressure.

[0033] In this embodiment, the feature enhancement module of the solid waste resource co-processing analysis system first constructs a feature weight allocation rule based on the correlation between components and processes in the co-processing of solid waste resources. It assigns higher weights to features in the fused feature vector corresponding to the heavy metal content of incineration fly ash, the active ingredient content of industrial alkali slag, the crystal water content of desulfurization gypsum, and the co-excitation temperature and mixing pressure, while assigning lower weights to other secondary features. Subsequently, the module employs a feature dimensionality enhancement algorithm to map the fused feature vector to a higher-dimensional feature space, supplementing the potential correlation information between components and processes by introducing virtual correlation features. Next, the module uses a noise filtering algorithm to remove noise features in the feature vector that are irrelevant to the co-processing effect, retaining the core features that significantly affect performance prediction and risk assessment. Finally, the module concatenates the weighted, dimensionally enhanced, and noise-filtered features to generate an enhanced feature vector, where each dimension of the vector corresponds to a set of enhanced component-process correlation features.

[0034] Step 1342: Synchronously input the enhanced feature vector into the mechanical property prediction branch, durability prediction branch, and leaching risk prediction branch of the multi-task predictor; wherein, the mechanical property prediction branch has a built-in compressive strength prediction sub-model based on a deep neural network, which is trained through historical solid waste co-processing data, and is used to output the compressive strength data of the target product under different curing cycles based on the component process correlation information contained in the enhanced feature vector, and integrate the compressive strength data into mechanical property features; the durability prediction branch has a built-in impermeability-freeze-resistance coupled prediction sub-model, which introduces the impermeability-freeze-resistance coupled prediction sub-model into the multi-task predictor. The correlation function between solid waste composition and hydrological properties is used to output the impermeability coefficient and freeze-thaw cycle tolerance data of the target product based on the enhanced feature vector, and integrate the impermeability data and freeze-thaw resistance data into service durability characteristics. The leaching risk prediction branch has a built-in heavy metal migration law prediction sub-model. The heavy metal migration law prediction sub-model combines the distribution characteristics of heavy metal speciation in incineration fly ash with the chemical action mechanism of the synergistic activation process, and is used to output the leaching concentration and leaching rate data of heavy metals in the target product under different environmental conditions based on the enhanced feature vector, and integrate the heavy metal leaching concentration and leaching rate data into leaching characteristic characteristics.

[0035] In this embodiment, the solid waste resource co-processing analysis system first synchronously inputs the enhanced feature vector output in step 1341 into the three branches of the multi-task predictor. For the mechanical performance prediction branch, its built-in compressive strength prediction sub-model adopts a deep neural network architecture, containing multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layers are used to extract local correlation information of component-process related features, the pooling layers are used to reduce and compress features, and the fully connected layers are used to map features to the compressive strength prediction space. The training data of this sub-model comes from more than 100,000 historical solid waste co-processing samples. The input is the correlation features between solid waste components and process parameters, and the output is the compressive strength data of the corresponding product. During training, the Adam optimizer is used, the initial learning rate is preset to a minimum value, the batch size is preset to a fixed value, the number of training rounds is preset to a fixed number, and the mean absolute percentage error is used as the early termination index. During the inference phase, the sub-model outputs the compressive strength data of the target product under multiple curing cycles based on the component and process correlation information in the enhanced feature vector. The above data is integrated into mechanical performance features, which are stored in the form of a time series array, with each element corresponding to the predicted compressive strength value of a curing cycle.

[0036] For the durability prediction branch, its built-in impermeability-freeze-resistance coupled prediction sub-model introduces a correlation function between solid waste components and hydrophysiological properties. This function is constructed based on the correlation between hydrophysiological properties such as porosity and water absorption rate in solid waste and impermeability and freeze-thaw resistance. The sub-model first extracts the hydrophysiological feature parameters corresponding to solid waste components from the enhanced feature vector, and maps them to the impermeability and freeze-thaw resistance feature space through the correlation function. Then, it outputs the impermeability coefficient sequence and freeze-thaw cycle tolerance sequence of the target product. The above data are integrated into a durability feature, which is stored in dictionary form, with the keys being impermeability data and freeze-thaw resistance data, and the values ​​being the corresponding sequence arrays.

[0037] For the leaching risk prediction branch, its built-in heavy metal migration prediction sub-model combines the distribution characteristics of heavy metal speciation in incineration fly ash with the chemical mechanism of the synergistic activation process. This mechanism is based on the solidification and stabilization law of heavy metals under alkaline activation environment. The sub-model first extracts the distribution characteristics of heavy metal speciation in incineration fly ash and the synergistic activation process parameters from the enhanced feature vector. Through the chemical mechanism model, it predicts the migration law of heavy metals under different environmental conditions, outputting the leaching concentration sequence and leaching rate sequence of heavy metals in the target product. The above data are integrated into leaching characteristic features, which are stored in the form of multi-dimensional tensors. Each dimension corresponds to an environmental condition, and each element corresponds to a set of heavy metal leaching data.

[0038] Step 1343: Standardize the data format of the mechanical performance characteristics, service durability characteristics and leaching characteristics, integrate the standardized mechanical performance characteristics and service durability characteristics into the performance prediction results, and determine the standardized leaching characteristics as the leaching risk assessment results of the harmful components.

[0039] In this embodiment, the solid waste resource co-processing analysis system first standardizes the data format of mechanical performance characteristics, service durability characteristics, and leaching characteristics. Using the Z-Score standardization algorithm, the original data of each characteristic is transformed into standardized data with a mean of 0 and a standard deviation of 1, eliminating dimensional differences between different characteristics. Subsequently, the system concatenates the standardized mechanical performance characteristics with the service durability characteristics, merging the two types of features into a high-dimensional tensor along the feature dimensions. This tensor serves as the performance prediction result, where each dimension corresponds to a type of performance characteristic, and each element corresponds to a standardized performance parameter value. Simultaneously, the system identifies the standardized leaching characteristics as the hazardous component leaching risk assessment result. This result is stored in the form of a multi-dimensional tensor, where each dimension corresponds to a heavy metal element, and each element corresponds to a standardized leaching parameter value.

[0040] Step 140: Combine the mechanical performance characteristics and service durability characteristics in the performance prediction results with the leaching characteristics in the hazardous component leaching risk assessment results to create a multi-objective optimization function.

[0041] In this embodiment, the solid waste resource co-processing analysis system first uses the compressive strength target corresponding to the mechanical performance characteristics in the performance prediction results as the first optimization objective, the impermeability and frost resistance target corresponding to the durability characteristics as the second optimization objective, and the heavy metal leaching concentration limit corresponding to the leaching characteristics in the hazardous component leaching risk assessment results as the third optimization objective. Based on industry standards and engineering practice requirements for solid waste resource co-processing, the system uses the analytic hierarchy process (AHP) to rank the importance of the three optimization objectives and determines the weight coefficients corresponding to each optimization objective based on the ranking results. The sum of the three weight coefficients is 1. Subsequently, the system constructs the main body of the objective function based on the weight coefficients, using the synergistic activation conditions and mixing and molding parameters corresponding to the initial production process parameters as function variables, and defines the value range of each variable. This range is determined based on the thermodynamic conditions of the solid waste synergistic activation reaction and the operating limit parameters of the engineering equipment. Next, the system introduces the material consumption cost threshold and energy consumption threshold in the solid waste resource co-processing as constraints. The material consumption cost threshold corresponds to the upper limit of the sum of the products of the total consumption of the three types of solid waste raw materials and their corresponding unit prices, and the energy consumption threshold corresponds to the upper limit of the total energy consumption of the synergistic activation process and the mixing and molding process. Finally, the system integrates the three optimization objectives into the main body of the objective function using a linear weighting method. The optimization direction of the function is defined as maximizing the degree of compliance of compressive strength and durability, minimizing the probability of heavy metal leaching concentration exceeding the limit, and generating a multi-objective optimization function that includes the main body of the objective function, the range of variables, and the constraints.

[0042] Step 141: Take the compressive strength target value corresponding to the mechanical performance characteristics in the performance prediction results as the first optimization target, take the impermeability and frost resistance target values ​​corresponding to the durability characteristics as the second optimization target, and take the heavy metal leaching concentration limit value corresponding to the leaching characteristics in the harmful component leaching risk assessment results as the third optimization target, and determine the weight coefficient of each optimization target.

[0043] In this embodiment, the solid waste resource co-processing analysis system first determines the benchmark values ​​for compressive strength, impermeability, and frost resistance for the first optimization objective, and the heavy metal leaching concentration limit for the third optimization objective, based on industry standards for solid waste resource co-processing. These benchmark values ​​are expressed as threshold ranges specified in industry standards. Subsequently, the system employs the analytic hierarchy process (AHP) to invite experts in the solid waste disposal field to score the importance of the three optimization objectives, constructing a judgment matrix. Matrix operations are used to determine the importance ranking of each optimization objective, and the ranking results correspond to the allocation of weight coefficients. Next, the system combines the ranking results to determine the weight coefficients corresponding to each optimization objective. The sum of the weight coefficients corresponding to compressive strength, durability, and heavy metal leaching risk control is 1. The specific values ​​of these weight coefficients are determined based on the expert scores and the results of the AHP calculations.

[0044] Step 142: Construct the main body of the objective function based on the weight coefficients, and take the synergistic excitation conditions and mixing molding parameters corresponding to the initial production process parameters as function variables, and introduce the material consumption cost threshold and energy consumption threshold in the co-processing of solid waste resources as constraints.

[0045] In this embodiment, the solid waste resource co-processing analysis system first constructs the main structure of the objective function based on the weighting coefficients determined in step 141. The main structure is represented in a linear weighted form, with each weighting coefficient corresponding to a quantitative index of the optimization objective. Subsequently, the system uses the mixing ratio of incineration fly ash-industrial alkali residue-desulfurization gypsum, co-activation temperature, co-activation time, mixing and molding pressure, and molding time included in the initial production process parameters as function variables, defining the value range of each variable. The value range of the mixing ratio is determined based on the co-activation reaction requirements of the three types of solid waste; the value ranges of the co-activation temperature and time are determined based on the thermodynamic conditions of the solid waste activation reaction; and the value ranges of the mixing and molding pressure and time are determined based on the operating limit parameters of the engineering equipment. Next, the system introduces a material consumption cost threshold and an energy consumption threshold as constraints. The constraint corresponding to the material consumption cost threshold is that the sum of the products of the consumption amount and the corresponding unit price of the three types of solid waste raw materials does not exceed a preset threshold. The constraint corresponding to the energy consumption threshold is that the sum of the energy consumption of the co-activation process and the energy consumption of the mixing and molding process does not exceed a preset threshold. These constraints are embedded in the main body of the objective function in the form of inequalities.

[0046] Step 143: Integrate the first optimization objective, the second optimization objective, and the third optimization objective into the main body of the objective function using a linear weighting method. Define the optimization direction of the function as maximizing the degree of compliance of compressive strength and durability, and minimizing the probability of heavy metal leaching concentration exceeding the limit. Generate a multi-objective optimization function that includes the main body of the objective function, the range of variables, and the constraints.

[0047] In this embodiment, the solid waste resource co-processing analysis system first defines the quantitative index of the first optimization objective as the ratio of the actual compressive strength to the compressive strength compliance benchmark value. The quantitative index of the second optimization objective is defined as the average of the impermeability compliance coefficient and the frost resistance compliance coefficient, where the impermeability compliance coefficient is the ratio of the actual impermeability coefficient to the impermeability compliance benchmark value, and the frost resistance compliance coefficient is the ratio of the actual freeze-thaw cycle tolerance number to the frost resistance compliance benchmark value. The quantitative index of the third optimization objective is defined as the difference between 1 and the probability of heavy metal leaching concentration exceeding the limit. Subsequently, the system uses a linear weighting method to multiply the quantitative indexes of the three optimization objectives by their corresponding weighting coefficients and then sums them, integrating them into the main body of the objective function to obtain a comprehensive objective value. The system then defines the optimization direction of the function as maximizing the comprehensive objective value, that is, maximizing the compressive strength and durability compliance level while minimizing the probability of heavy metal leaching concentration exceeding the limit. Finally, the system integrates the objective function, the range of variable values, and the constraints to generate a complete multi-objective optimization function. This function is stored in a structured form, containing the objective function expression, the set of variable ranges, and the set of constraints.

[0048] Step 150: Input the initial production process parameters as optimization variables into the preset particle swarm optimization algorithm, and iteratively optimize and adjust the initial production process parameters guided by the multi-objective optimization function to generate multiple sets of process parameter optimization candidate sets.

[0049] In this embodiment, the solid waste resource co-processing analysis system first sets the initial parameters of a preset particle swarm optimization algorithm, including particle swarm size, maximum number of iterations, learning factor, and inertia weight. The particle swarm size is determined based on the number of optimization variables, the maximum number of iterations is determined based on the convergence speed requirement of the algorithm, the learning factor is used to adjust the learning ability of particles towards individual optimal solutions and global optimal solutions, and the inertia weight is used to adjust the global search and local search capabilities of particles. Subsequently, the system uses the mixing ratio of incineration fly ash-industrial alkali residue-desulfurization gypsum, co-excitation temperature, co-excitation time, mixing and molding pressure, and molding time included in the initial production process parameters as particle position vectors, with each particle's position vector corresponding to a set of process parameter combinations.

[0050] Next, the system uses the multi-objective optimization function generated in step 140 as the fitness function to calculate the fitness value of each particle in the initial particle population. It then selects the initial optimal particles whose fitness values ​​meet the preset fitness threshold. Based on these initial optimal particles, the system updates the velocity and position vectors of each particle in the population. During each iteration, the updated process parameters are substituted into the multi-objective optimization function to recalculate the fitness value, retaining the non-dominated solution particles from the iteration process. When the number of iterations reaches the maximum number of iterations or the change in fitness value is less than the preset convergence threshold, the system stops iterating. It then performs deduplication on the process parameters corresponding to the non-dominated solution particles retained during the iteration process, selecting process parameter combinations that meet the basic requirements for co-processing of solid waste resources, and generating multiple sets of process parameter optimization candidate sets.

[0051] Step 151: Set the initial parameters of the preset particle swarm optimization algorithm, including particle swarm size, maximum number of iterations, learning factor and inertia weight.

[0052] In this embodiment, the solid waste resource collaborative disposal analysis system first sets the particle swarm size based on the number of optimization variables. There are five groups of optimization variables, and the particle swarm size is set to a fixed value proportional to the number of variables. Then, based on the convergence speed requirements of the particle swarm optimization algorithm, the system sets the maximum number of iterations, which is determined based on the convergence results of historical optimization tasks. Next, the system sets learning factors, including individual learning factors and global learning factors. The individual learning factor adjusts the particle's ability to learn from its own historical best solution, while the global learning factor adjusts the particle's ability to learn from the global best solution of the swarm. The values ​​of both learning factors are determined based on the algorithm's search performance requirements. Finally, the system sets an inertia weight. The initial inertia weight is determined based on the algorithm's global search requirements and dynamically decays with the number of iterations to balance the algorithm's global and local search capabilities.

[0053] Step 152: The mixing ratio of incineration fly ash-industrial alkali residue-desulfurized gypsum, synergistic excitation temperature, synergistic excitation time, mixing and molding pressure and molding time included in the initial production process parameters are used as the position vectors of the particles.

[0054] In this embodiment, the solid waste resource co-processing analysis system first extracts five types of parameters from the initial production process parameters: the mixing ratio of incineration fly ash-industrial alkali residue-desulfurization gypsum, co-activation temperature, co-activation time, mixing and molding pressure, and molding time. Each type of parameter corresponds to an optimization variable. The system then maps the value range of each type of parameter to the dimensions of the particle position vector. Each particle's position vector is a five-dimensional vector, with each dimension corresponding to the value of a type of process parameter. For example, the first dimension of the position vector corresponds to the proportion of incineration fly ash in the mixing ratio, the second dimension to the co-activation temperature, the third dimension to the co-activation time, the fourth dimension to the mixing and molding pressure, and the fifth dimension to the molding time. The value range of each dimension is consistent with the value range of the corresponding process parameter.

[0055] Step 153: Using the multi-objective optimization function as the fitness function, calculate the fitness value of each particle in the initial particle population, select the initial optimal particles whose fitness values ​​meet the preset fitness threshold, update the velocity vector and position vector of each particle in the population based on the initial optimal particles, and in each iteration, substitute the process parameters corresponding to the updated particles into the multi-objective optimization function to recalculate the fitness value, and retain the non-dominated solution particles in the iteration process.

[0056] In this embodiment, the solid waste resource co-processing analysis system first transforms the position vector of each particle in the initial particle population into corresponding process parameters. These parameters are then substituted into a multi-objective optimization function to calculate the comprehensive fitness value for each particle. This value is the result of a linear weighted average of three optimization objective quantification indicators. Subsequently, the system determines a preset fitness threshold based on the minimum compliance requirements for solid waste resource co-processing, selecting particles with a comprehensive fitness value greater than or equal to this threshold as initial optimal particles. The position vector and fitness value of the initial optimal particles are recorded. Next, the system determines the individual optimal solution based on the position vector of the initial optimal particle. By comparing the fitness values ​​of all initial optimal particles, the system determines the global optimal solution. A preset velocity update formula and position update formula are used to update the velocity and position vectors of each particle in the population. The velocity update formula introduces a learning factor and inertia weight to adjust the particle's search capability. The inertia weight dynamically decays with the number of iterations to balance global and local search capabilities.

[0057] After each iteration update, the system extracts the process parameters corresponding to the updated position vectors of each particle, substitutes them into the multi-objective optimization function to recalculate the comprehensive fitness value, compares the fitness values ​​of each particle based on the Pareto dominance relationship, retains non-dominated solution particles that are not dominated by other particles, and stores their corresponding position vectors and fitness values ​​in an external file. At the same time, the external file is maintained. When the number of particles in the file exceeds the preset capacity, the crowding degree sorting method is used to delete particles with lower crowding degree to maintain diversity. The above steps are repeated until the current iteration round is completed.

[0058] Step 1531: Transform the position vector of each particle in the initial particle population into the corresponding process parameters of incineration fly ash-industrial alkali residue-desulfurization gypsum mixing ratio, synergistic excitation temperature, synergistic excitation time, mixing molding pressure and molding time. Substitute the process parameters into the multi-objective optimization function to calculate the comprehensive fitness value of each particle. The comprehensive fitness value is the result of linear weighting of the quantitative indicators of each optimization objective.

[0059] In this embodiment, the solid waste resource co-processing analysis system first transforms the five-dimensional position vector of each particle in the initial particle population into corresponding process parameters. The five dimensions of the position vector correspond to five core parameters: the proportion of incineration fly ash, the proportion of industrial alkali slag, the proportion of desulfurization gypsum, the co-activation temperature, the co-activation time, the mixing and molding pressure, and the molding time. The system then substitutes these process parameters into the multi-objective optimization function generated in step 140. The function first calculates the quantitative indicators of the three optimization objectives: the compressive strength compliance ratio, the average durability compliance value, and the heavy metal leaching risk control value. Then, it multiplies these indicators by their corresponding weighting coefficients and sums them to obtain the comprehensive fitness value for each particle. This value, expressed numerically, reflects the comprehensive optimization degree of the corresponding process parameter combination.

[0060] Step 1532: Determine the fitness preset threshold based on the minimum compliance requirements for co-processing of solid waste resources, select particles with a comprehensive fitness value greater than or equal to the fitness preset threshold as the initial optimal particles, and record the position vector and fitness value corresponding to the initial optimal particles.

[0061] In this embodiment, the solid waste resource co-processing analysis system first determines a preset fitness threshold based on the minimum engineering requirements for solid waste resource co-processing. This threshold corresponds to the comprehensive fitness value when the combination of process parameters meets the minimum compressive strength requirement, minimum durability requirement, and highest heavy metal leaching risk requirement. The system then iterates through all particles in the initial particle population, comparing the comprehensive fitness value of each particle with the preset fitness threshold, and selecting particles with a comprehensive fitness value greater than or equal to the threshold as the initial optimal particles. Next, the system records the position vector and fitness value corresponding to each initial optimal particle and stores them in a temporary array, providing a data foundation for subsequent determination of individual and global optimal solutions.

[0062] Step 1533: Determine the individual optimal solution based on the position vector corresponding to the initial optimal particle, determine the global optimal solution by comparing the fitness values ​​of all initial optimal particles, and update the velocity vector and position vector of each particle in the population using a preset velocity update formula and position update formula; wherein, the velocity update formula introduces a learning factor and inertia weight to adjust the particle's search capability, and the inertia weight dynamically decays with the number of iterations to balance the global search and local search capabilities of the algorithm.

[0063] In this embodiment, the solid waste resource collaborative disposal analysis system first determines the position vector of each initial optimal particle as its individual optimal solution and records the fitness value corresponding to each particle's individual optimal solution. Then, the system compares the fitness values ​​of all initial optimal particles and determines the position vector corresponding to the particle with the highest fitness value as the global optimal solution, recording the fitness value corresponding to the global optimal solution. Next, the system updates the velocity and position vectors of each particle in the population using preset velocity update and position update formulas. The velocity update formula is as follows: the particle's new velocity equals the inertia weight multiplied by the current velocity, plus the individual learning factor multiplied by the random coefficient multiplied by the difference between the individual optimal solution and the current position, plus the global learning factor multiplied by the random coefficient multiplied by the difference between the global optimal solution and the current position. The inertia weight dynamically decreases according to a preset ratio as the number of iterations increases, with a larger initial value to enhance global search capability and a smaller value in later iterations to enhance local search capability. The position update formula is as follows: the particle's new position equals the current position plus the new velocity, ensuring that the position update is within the range of variable values.

[0064] Step 1534: After each iteration update, extract the process parameters corresponding to the updated position vector of each particle, substitute the process parameters into the multi-objective optimization function to recalculate the corresponding comprehensive fitness value, and compare the fitness values ​​of each particle based on the Pareto dominance relationship.

[0065] In this embodiment, after each iteration update, the solid waste resource collaborative disposal analysis system first extracts the process parameters corresponding to the updated position vectors of each particle, substitutes these parameters into the multi-objective optimization function, and recalculates the comprehensive fitness value corresponding to each particle. Subsequently, the system compares the fitness values ​​of each particle based on the Pareto dominance relationship. If the comprehensive fitness value of particle A is not inferior to that of particle B in all optimization objectives, and is superior to that of particle B in at least one optimization objective, then particle A is considered to dominate particle B, particle B is a dominated solution, and particle A is a non-dominated solution.

[0066] Step 1535: Retain the non-dominated solution particles that are not dominated by other particles during the iteration process, store the position vector and fitness value corresponding to the non-dominated solution particles in the external file, and maintain the non-dominated solution particles in the external file. When the number of particles in the external file exceeds the preset capacity, use the crowding sorting method to delete particles with smaller crowding to maintain the diversity of particles in the file.

[0067] In this embodiment, the solid waste resource collaborative disposal analysis system first filters out non-dominated solution particles that are not dominated by other particles after each iteration. The position vectors and fitness values ​​corresponding to these particles are stored in an external archive, which stores all non-dominated solutions generated during the iteration process. The system then maintains the external archive. When the number of particles in the archive exceeds a preset capacity, the crowding degree ranking method is used to calculate the crowding degree of each particle in the archive. Crowding degree represents the density of surrounding particles on the Pareto front; the smaller the crowding degree, the denser the particles at the particle's location. Next, the system deletes particles with lower crowding degrees and retains particles with higher crowding degrees to maintain the diversity of particles in the external archive and ensure that the coverage of the Pareto front is sufficiently broad.

[0068] Step 154: Stop iterating when the number of iterations reaches the maximum number of iterations or the change in fitness value is less than the preset convergence threshold.

[0069] In this embodiment, after each iteration, the solid waste resource collaborative disposal analysis system first determines whether the current iteration count has reached the preset maximum iteration count. If it has, the iteration stops directly. If it has not, the system calculates the average change between the overall fitness value of all particles in the current iteration and the overall fitness value of the corresponding particles in the previous iteration. If this average is less than the preset convergence threshold, it indicates that the algorithm has entered the convergence state, and continuing the iteration will not significantly improve the optimization effect. At this point, the iteration stops.

[0070] Step 155: Deduplicate the process parameters corresponding to the non-dominated solution particles retained during the iteration process, screen out the process parameter combinations that meet the basic requirements for co-processing of solid waste resources, and generate multiple sets of process parameter optimization candidate sets.

[0071] In this embodiment, the solid waste resource co-processing analysis system first deduplicates the process parameters corresponding to the non-dominated solution particles retained during the iteration process. A feature matching algorithm is used to compare the feature similarity of process parameter combinations corresponding to different particles. Process parameter combinations with similarity higher than a preset threshold are considered duplicate combinations, and only one of them is retained. Subsequently, the system selects combinations from the deduplicated process parameter combinations that meet the basic requirements for solid waste resource co-processing. These basic requirements include that the mixing ratio meets the synergistic activation reaction requirements of the three types of solid waste, the process parameters are within the equipment operating limits, and minimum performance and environmental protection requirements are met. Finally, the system organizes the selected process parameter combinations into multiple candidate sets for process parameter optimization. Each candidate set corresponds to one set of process parameter combinations and is stored in the form of a structured dictionary.

[0072] Step 160: Call the neural network intelligent decision model to generate performance prediction optimization results and leaching risk assessment optimization results corresponding to each set of process parameter optimization candidate sets. Based on the performance prediction optimization results and the leaching risk assessment optimization results, generate production process optimization parameters that include the target product ratio and optimized synergistic conditions, and material forming parameters.

[0073] In this embodiment, the solid waste resource co-processing analysis system first aligns the process parameters corresponding to each candidate set of process parameters with the solid waste co-processing associated dataset generated in step 120, and inputs them into a neural network intelligent decision-making model. The model outputs corresponding performance prediction optimization results and leaching risk assessment optimization results. The performance prediction optimization results include mechanical performance characteristics and service durability characteristics, while the leaching risk assessment optimization results include leaching characteristic characteristics. Subsequently, the system extracts the actual predicted values ​​of compressive strength, durability, and heavy metal leaching corresponding to each candidate set, and constructs a parameter-performance-risk assessment matrix. Each row of the matrix corresponds to a candidate set, and each column corresponds to a type of assessment index.

[0074] Next, the system uses a multi-attribute decision-making method based on the evaluation matrix to comprehensively rank the candidate sets. This method calculates a comprehensive score for each candidate set based on the weights of each evaluation indicator, selecting the candidate set with the highest comprehensive score as the core candidate set, and retaining alternative candidate sets whose comprehensive scores fall within a preset range. Subsequently, the system verifies the process parameters corresponding to the core candidate sets, confirming that the mixing ratio meets the requirements for synergistic reaction activation, and that the temperature and time parameters in the synergistic effect conditions match the pressure and time parameters in the material forming parameters. Finally, the system determines the mixing ratio corresponding to the verified core candidate sets as the target ratio, determines the synergistic activation temperature and synergistic activation time as the optimized synergistic effect conditions, and determines the mixing and forming pressure and forming time as the material forming parameters, combining these three types of parameters to generate optimized production process parameters.

[0075] Step 161: Extract the actual predicted values ​​of compressive strength and durability from the performance prediction optimization results corresponding to each set of process parameter optimization candidates, and the actual predicted values ​​of heavy metal leaching from the leaching risk assessment optimization results, and construct a parameter-performance-risk assessment matrix.

[0076] In this embodiment, the solid waste resource co-processing analysis system first traverses each set of process parameter optimization candidates, extracting the actual predicted values ​​of compressive strength and durability from the corresponding performance prediction optimization results. The actual predicted values ​​of compressive strength are numerical sequences under multiple curing cycles, and the actual predicted values ​​of durability are numerical sequences of the impermeability coefficient and the number of freeze-thaw cycles. Simultaneously, the system extracts the actual predicted values ​​of heavy metal leaching from the leaching risk assessment optimization results. These values ​​are numerical sequences of leaching concentration and leaching rate under multiple environmental conditions. Subsequently, the system performs dimensionality compression on the above predicted values, converting the sequence data into comprehensive index values. For example, the actual predicted value of compressive strength is the average of multiple curing cycles, the actual predicted value of durability is the average of the impermeability and freeze-thaw resistance compliance coefficients, and the actual predicted value of heavy metal leaching is the ratio of the maximum leaching concentration to the limit under multiple environmental conditions. Finally, the system aligns the process parameters, compressed performance indicators, and compressed risk indicators corresponding to each candidate set to construct a parameter-performance-risk assessment matrix. Each row of the matrix corresponds to a candidate set, and the columns are the process parameter combination, compressive strength index, durability index, and heavy metal leaching risk index, respectively.

[0077] Step 162: Based on the evaluation matrix, a multi-attribute decision method is used to comprehensively sort the candidate sets for optimizing each process parameter, select the candidate set with the highest comprehensive score as the core candidate set, and retain the alternative candidate set whose comprehensive score is within a preset range.

[0078] In this embodiment, the solid waste resource collaborative disposal analysis system first employs a multi-attribute decision-making method. Based on the needs of solid waste resource collaborative disposal, it determines the weight coefficients of each evaluation indicator. The sum of the weight coefficients for compressive strength, durability, and heavy metal leaching risk is 1. Then, based on the data in the evaluation matrix, the system standardizes each evaluation indicator to eliminate dimensional differences. For example, the compressive strength indicator uses positive standardization, where a higher value results in a larger standardized value; the heavy metal leaching risk indicator uses negative standardization, where a higher value results in a smaller standardized value. Next, the system multiplies the standardized indicators by their corresponding weight coefficients and sums them to obtain a comprehensive score for each candidate set. A higher comprehensive score indicates better overall performance of the candidate set. The system then sorts all candidate sets in descending order of their comprehensive scores, selecting the candidate set with the highest comprehensive score as the core candidate set, and retaining candidate sets with comprehensive scores within a preset range as alternative candidate sets. The preset range corresponds to a fixed proportion of candidates with high comprehensive scores.

[0079] Step 163: Verify the process parameters corresponding to the core candidate set to confirm that the mixing ratio of incineration fly ash-industrial alkali residue-desulfurization gypsum corresponding to the process parameters meets the requirements for synergistic activation reaction, and that the temperature and time parameters in the synergistic effect conditions match the pressure and time parameters in the material forming parameters.

[0080] In this embodiment, the solid waste resource co-processing analysis system first verifies the mixing ratio corresponding to the core candidate set. Based on the thermodynamic laws of solid waste synergistic activation reactions, it checks whether the proportions of the three types of solid waste in the mixing ratio meet the reaction requirements of alkaline activation and sulfate activation. For example, whether the proportion of active ingredients in industrial alkali slag can provide a sufficient alkaline environment, and whether the crystal water content of desulfurized gypsum can provide sufficient sulfate ions. Subsequently, the system verifies the matching between the synergistic effect conditions and the material forming parameters, checking whether the synergistic activation temperature and time parameters match the mixing and forming pressure and time parameters. For example, when the synergistic activation temperature is high, whether the forming time is adjusted accordingly to ensure that the material's fluidity meets the forming requirements. If the process parameters are found to be unsatisfactory during the verification process, the system re-sends the core candidate set to the particle swarm optimization algorithm for secondary optimization until the verification is successful.

[0081] Step 164: Determine the mixing ratio corresponding to the verified core candidate set as the target ratio of the target product, determine the synergistic excitation temperature and synergistic excitation time corresponding to the core candidate set as the optimized synergistic conditions, and determine the mixing molding pressure and molding time corresponding to the core candidate set as the material molding parameters. Combine the target ratio, optimized synergistic conditions and material molding parameters to generate production process optimization parameters.

[0082] In this embodiment, the solid waste resource co-processing analysis system first determines the target ratio of the incineration fly ash-industrial alkali residue-desulfurization gypsum mixture corresponding to the verified core candidate set as the target ratio of the target product. This ratio is expressed as the mass percentage of the three types of solid waste. Subsequently, the system determines the synergistic activation temperature and synergistic activation time corresponding to the core candidate set as optimized synergistic conditions, expressed as temperature and time intervals. Next, the system determines the mixing and molding pressure and molding time corresponding to the core candidate set as material molding parameters, expressed as pressure and time intervals. Finally, the system structurally integrates the target ratio, optimized synergistic conditions, and material molding parameters to generate optimized production process parameters. These parameters are stored in a multi-level dictionary format, including a solid waste ratio sub-dictionary, a synergistic condition sub-dictionary, and a material molding parameter sub-dictionary.

[0083] In an optional embodiment, it further includes: Step 210: Perform data cleaning and dimension alignment on the actual process parameter data, actual product performance data, and actual hazardous component leaching data generated during the co-processing of different batches of solid waste resources to obtain transfer learning sample data.

[0084] In this embodiment, the solid waste resource co-processing analysis system first collects actual process parameter data generated during the co-processing of different batches of solid waste resources, including mixing ratio, co-activation temperature, molding pressure, etc.; actual product performance data, including compressive strength, impermeability, and freeze resistance, etc.; and actual hazardous component leaching data, including heavy metal leaching concentration and leaching rate, etc. The system then cleans the above data, using an outlier detection algorithm to identify and remove outliers, and using a missing value imputation algorithm to fill in missing data, ensuring data integrity and accuracy. Next, the system performs dimensional alignment processing on the cleaned actual process parameter data, actual product performance data, and actual hazardous component leaching data, performing feature matching along the batch dimension and feature dimension to ensure that the batch quantity and feature dimension of the three types of data are completely consistent. Finally, the system integrates the aligned data into transfer learning sample data, which is stored in the form of a multi-dimensional tensor, with each dimension corresponding to a feature set of one type of data.

[0085] Step 220: Perform feature distribution difference analysis on the transfer learning sample data and the training sample data of the neural network intelligent decision-making model to generate a feature distribution difference matrix.

[0086] In this embodiment, the solid waste resource collaborative disposal analysis system first extracts statistical distribution information of corresponding features, including mean, variance, and quantiles, from the transfer learning sample data and the training sample data of the neural network intelligent decision-making model. Then, the system uses a feature distribution difference analysis algorithm to compare the feature distributions of the two types of sample data dimension by dimension, calculating the distribution difference degree for each feature dimension. The distribution difference degree is determined based on the degree of deviation of the statistical distribution information of the two types of data. Next, the system organizes the distribution difference degree of each feature dimension into a feature distribution difference matrix. Each row of the matrix corresponds to a feature dimension, and each column corresponds to a specific value of the distribution difference degree. This matrix reflects the distribution difference between the transfer learning sample data and the original training sample data.

[0087] Step 230: Construct a transfer learning adaptation function based on the feature distribution difference matrix, and perform distribution correction processing on the transfer learning sample data through the transfer learning adaptation function to obtain the adapted transfer learning sample data.

[0088] In this embodiment, the solid waste resource collaborative disposal analysis system first constructs a transfer learning adaptation function based on the distribution difference degree in the feature distribution difference matrix. The function's logic is as follows: based on the distribution difference degree of each feature dimension, it transforms the distribution of the corresponding features in the transfer learning sample data to make its distribution consistent with the distribution of the original training sample data. Subsequently, the system inputs the transfer learning sample data into the transfer learning adaptation function, which transforms the data point-by-point for each feature dimension, for example, by using normalization, standardization, or other distribution transformation methods to eliminate the distribution differences between the two types of sample data. Finally, the system outputs the adapted transfer learning sample data, whose feature distribution is highly consistent with the feature distribution of the original training sample data, and can be used for fine-tuning the neural network intelligent decision-making model.

[0089] Step 240: Input the adapted transfer learning sample data into the neural network intelligent decision-making model, and use fine-tuning training to iteratively update the parameters of the feature interaction layer and decision output layer of the neural network intelligent decision-making model to generate the iteratively updated neural network intelligent decision-making model.

[0090] In this embodiment, the solid waste resource collaborative disposal analysis system first divides the adapted transfer learning sample data output in step 230 according to the ratio of training set to validation set. The training set is used for fine-tuning the model, and the validation set is used for performance verification. Subsequently, the system uses a fine-tuning training method, freezing the parameters of the underlying feature extraction layer of the neural network intelligent decision-making model and updating only the parameters of the feature interaction layer and the decision output layer. During fine-tuning training, the Adam optimizer is used, with the initial learning rate preset to a minimum value, the batch size preset to a fixed value, and the number of training epochs preset to a fixed number. The mean absolute percentage error on the validation set is used as the early stopping criterion. In each training epoch, the system inputs the training set data into the model, the model outputs the prediction result, and the backpropagation algorithm is used to optimize the parameters of the feature interaction layer and the decision output layer by calculating the error between the prediction result and the true label. When the number of training epochs reaches a preset number or the validation set error reaches a preset threshold, training stops, and an iteratively updated neural network intelligent decision-making model is generated.

[0091] Step 250: Input the associated dataset corresponding to the production process optimization parameters into the iteratively updated neural network intelligent decision-making model to obtain new performance prediction results and hazardous component leaching risk assessment results. By analyzing the deviation between the new performance prediction results and hazardous component leaching risk assessment results and the actual data, test the generalization ability of the iteratively updated neural network intelligent decision-making model. If the deviation is less than the preset deviation threshold, the iteratively updated neural network intelligent decision-making model is determined as the new neural network intelligent decision-making model. Otherwise, adjust the parameters of the transfer learning adaptation function and repeat the distribution correction, model fine-tuning and verification steps until the deviation is less than the preset deviation threshold.

[0092] In this embodiment, the solid waste resource co-processing analysis system first inputs the solid waste co-processing associated dataset corresponding to the production process optimization parameters into the iteratively updated neural network intelligent decision-making model. The model outputs new performance prediction results and hazardous component leaching risk assessment results. The system then performs a deviation analysis on the predicted results and the corresponding actual product performance data and actual hazardous component leaching data. The deviation is determined based on the mean of the absolute errors between the predicted and actual values. Next, the system compares the deviation with a preset deviation threshold. If the deviation is less than the preset deviation threshold, it indicates that the model's generalization ability meets the requirements, and the iteratively updated neural network intelligent decision-making model is determined as the new neural network intelligent decision-making model for subsequent solid waste resource co-processing analysis. If the deviation is greater than or equal to the preset deviation threshold, it indicates that the model's generalization ability is insufficient. In this case, the system adjusts the parameters of the transfer learning adaptation function, such as adjusting the intensity of the distribution transformation or changing the logic of the adaptation function, repeating steps 230 to 250 until the deviation is less than the preset deviation threshold.

[0093] In an optional embodiment, it further includes: Step 310: Extract the fitness value change curves corresponding to the production process optimization parameters, and determine the convergence and fluctuation characteristics of the fitness values ​​through curve fitting analysis.

[0094] In this embodiment, the solid waste resource co-processing analysis system first extracts the average comprehensive fitness value of all particles in each iteration of the particle swarm optimization algorithm during the generation of production process optimization parameters. The average values ​​are then arranged according to the iteration order to generate a fitness value change curve, with the iteration order as the horizontal axis and the average fitness value as the vertical axis. Subsequently, the system uses a curve fitting analysis algorithm to fit the fitness value change curve, determining its convergence and fluctuation characteristics. The convergence characteristics include the convergence speed and the stable value after convergence, while the fluctuation characteristics include the fluctuation amplitude and fluctuation frequency. The convergence speed is determined based on the number of iterations required for the curve to reach its stable value; the stable value after convergence is determined based on the average value of the later stages of the curve; the fluctuation amplitude is determined based on the difference between the peak and trough values ​​in the curve; and the fluctuation frequency is determined based on the number of fluctuation cycles in the curve.

[0095] Step 320: Construct dynamic adjustment rules for algorithm parameters based on convergence and fluctuation characteristics. These rules are used to correlate the change in fitness value with the adjustment range of the inertia weight and learning factor of the particle swarm optimization algorithm.

[0096] In this embodiment, the solid waste resource collaborative disposal analysis system first constructs dynamic adjustment rules for algorithm parameters based on the convergence and fluctuation characteristics of fitness values. If the convergence speed of the fitness value change curve is slow, it indicates that the algorithm's global search capability is insufficient. In this case, the dynamic adjustment rule increases the initial value of the inertia weight to enhance the algorithm's global search capability. If the stable value after convergence is high but the fluctuation amplitude is large, it indicates that the algorithm's local search capability is insufficient. In this case, the dynamic adjustment rule increases the value of the learning factor to enhance the algorithm's learning ability towards the optimal solution. If the fluctuation frequency is high, it indicates that the algorithm's search process is unstable. In this case, the dynamic adjustment rule increases the decay rate of the inertia weight to accelerate the algorithm's transition from global search to local search. The dynamic adjustment rules are stored in the form of conditional judgments, with each condition corresponding to a fitness value change state and each result corresponding to a set of algorithm parameter adjustment amplitudes.

[0097] Step 330: Based on the algorithm parameter dynamic adjustment rules, generate a dynamic parameter adjustment sequence on the basis of the initial parameters of the particle swarm optimization algorithm; embed the dynamic parameter adjustment sequence into the preset particle swarm optimization algorithm to generate a dynamic particle swarm optimization algorithm.

[0098] In this embodiment, the solid waste resource collaborative disposal analysis system first generates a dynamic parameter adjustment sequence based on the initial parameters of the particle swarm optimization algorithm, according to the algorithm parameter dynamic adjustment rules constructed in step 320. This sequence stores the adjustment values ​​of the inertia weight and learning factor corresponding to each iteration in the order of iteration rounds. For example, in the early stage of iteration, the inertia weight takes a larger value and the learning factor takes a smaller value; in the middle stage of iteration, the inertia weight dynamically decays with each round, and the learning factor gradually increases; in the later stage of iteration, the inertia weight takes a smaller value and the learning factor takes a larger value. Subsequently, the system embeds the dynamic parameter adjustment sequence into a preset particle swarm optimization algorithm, modifies the update logic of the inertia weight and learning factor in the algorithm, so that it updates according to the parameter values ​​corresponding to the dynamic parameter adjustment sequence in each iteration, generating a dynamic particle swarm optimization algorithm. This algorithm can dynamically adjust the search strategy according to the change in fitness value.

[0099] Step 340: Input the initial production process parameters as optimization variables into the dynamic particle swarm optimization algorithm again, and perform a second iteration optimization adjustment guided by the original multi-objective optimization function to generate a candidate set of secondary process parameter optimizations.

[0100] In this embodiment, the solid waste resource co-processing analysis system first uses the initial production process parameters determined in step 120 as optimization variables and inputs them into the dynamic particle swarm optimization algorithm generated in step 330. Then, the system uses the original multi-objective optimization function generated in step 140 as the fitness function and initiates a second iteration for optimization adjustment. In each iteration, the algorithm adjusts the inertia weight and learning factor according to the dynamic parameter adjustment sequence to optimize the combination of process parameters. The iteration stops when the number of iterations reaches the preset maximum number of iterations or the change in fitness value is less than the preset convergence threshold. Finally, the system deduplicates and filters the process parameters corresponding to the non-dominated solution particles retained during the iteration process, generating a secondary process parameter optimization candidate set. This candidate set contains more and better-performing combinations of process parameters.

[0101] Step 350: Call the neural network intelligent decision model to generate the performance prediction secondary optimization results and leaching risk assessment secondary optimization results corresponding to the candidate set of secondary process parameters; by comparing the comprehensive fitness values ​​corresponding to the production process optimization parameters and the candidate set of secondary process parameters, select the process parameter combination with the better comprehensive fitness value as the final production process optimization parameters.

[0102] In this embodiment, the solid waste resource co-processing analysis system first aligns the process parameters corresponding to the candidate set of secondary process parameters with the solid waste co-processing associated dataset in terms of dimensions, inputs them into a neural network intelligent decision-making model, and the model outputs the corresponding performance prediction secondary optimization results and leaching risk assessment secondary optimization results. Then, the system substitutes the process parameters corresponding to the candidate set of secondary process parameters into the original multi-objective optimization function, calculates the comprehensive fitness value of each candidate set, and extracts the comprehensive fitness value corresponding to the production process optimization parameters. Next, the system compares the comprehensive fitness values ​​of the production process optimization parameters with those of the candidate set of secondary process parameters, and selects the process parameter combination with the higher comprehensive fitness value as the final production process optimization parameters. If there is a combination in the candidate set of secondary process parameters with a comprehensive fitness value higher than the production process optimization parameters, then that combination is determined as the final production process optimization parameters; if the comprehensive fitness values ​​of all secondary candidate sets are not higher than the production process optimization parameters, then the original production process optimization parameters are retained as the final production process optimization parameters.

[0103] In an optional embodiment, it further includes: Step 410: Obtain the activation values ​​of neurons in each layer of the neural network intelligent decision-making model during the processing of associated datasets. Determine the feature contribution of each neuron through statistical analysis of the activation values. Set a pruning threshold based on the feature contribution and filter out redundant neurons and redundant connections whose feature contribution is lower than the pruning threshold.

[0104] In this embodiment, the solid waste resource collaborative disposal analysis system first acquires the activation values ​​of neurons in each layer in real time during the processing of associated datasets by the neural network intelligent decision-making model. The activation values ​​are stored in a multi-dimensional array, with each element corresponding to the activation level of a neuron during data processing. Subsequently, the system determines the feature contribution of each neuron through statistical analysis of the activation values. The feature contribution is determined based on the mean, variance, and correlation with the output results of the neuron's activation values. A higher mean, smaller variance, and stronger correlation with the output results indicate a greater feature contribution. Next, the system sets a pruning threshold based on the feature contribution. This threshold corresponds to the minimum contribution required by a neuron to the model's performance; neurons below this threshold are considered redundant neurons, and their corresponding connections are considered redundant connections. Finally, the system filters out redundant neurons and redundant connections with feature contributions below the pruning threshold and records the location information of these neurons and connections.

[0105] Step 420: Create a pruning constraint function, which limits the upper limit of the model accuracy loss during the pruning process.

[0106] In this embodiment, the solid waste resource collaborative disposal analysis system first sets an upper limit for model accuracy loss during pruning. This upper limit corresponds to the maximum allowable increment of the deviation between the performance prediction results before and after pruning and the actual data. Subsequently, the system creates a pruning constraint function based on this upper limit. The function's logic is as follows: the accuracy loss value of the model after pruning must not exceed the preset upper limit, and the accuracy loss value is determined based on the degree of deviation between the prediction results of the model after pruning and the prediction results of the model before pruning. The pruning constraint function is expressed in inequality form and is used to constrain the number of redundant neurons and redundant connections deleted during the pruning process, ensuring that the accuracy of the model after pruning meets the requirements.

[0107] Step 430: Under the constraints of the pruning constraint function, prune the redundant neurons and redundant connections in the feature interaction layer and decision output layer of the neural network intelligent decision model to obtain the pruned neural network intelligent decision model.

[0108] In this embodiment, the solid waste resource collaborative disposal analysis system first prunes the redundant neurons and connections selected in step 410 under the constraint of a pruning constraint function. The pruning process adopts an incremental pruning method, deleting a small number of redundant neurons and connections each time, and then testing the accuracy loss value of the pruned model. If the accuracy loss value does not exceed the preset loss limit, the next batch of redundant neurons and connections is deleted; if the accuracy loss value exceeds the preset loss limit, pruning stops, and the neurons and connections deleted in the previous batch are restored. The pruning process only targets the feature interaction layer and decision output layer of the neural network intelligent decision model, while retaining the parameters of the underlying feature extraction layer unchanged. Finally, the system obtains a pruned neural network intelligent decision model, which has fewer neurons and connections than the original model, but the accuracy loss is controlled within an allowable range.

[0109] Step 440: Input the associated dataset and initial production process parameters into the pruned neural network intelligent decision-making model to generate performance prediction results and leaching risk assessment results after pruning.

[0110] In this embodiment, the solid waste resource co-processing analysis system first aligns the dimensions of the solid waste co-processing associated dataset generated in step 120 and the initial production process parameters determined in step 120, and inputs them into the pruned neural network intelligent decision-making model. Subsequently, the model performs feature extraction, feature interaction, and multi-task prediction on the input data to generate pruned performance prediction results and pruned leaching risk assessment results. The pruned performance prediction results include mechanical performance characteristics and service durability characteristics, while the pruned leaching risk assessment results include leaching characteristic characteristics. The format of these results is consistent with the format of the original model output.

[0111] Step 450: Compare the prediction accuracy and computational efficiency of the model before and after pruning. If the loss in prediction accuracy is lower than the preset loss threshold and the improvement in computational efficiency is higher than the preset improvement threshold, the pruned neural network intelligent decision-making model is determined as the optimized neural network intelligent decision-making model. Otherwise, adjust the pruning threshold and the parameters of the pruning constraint function, and re-perform redundancy screening and pruning until an optimized neural network intelligent decision-making model that meets the requirements is obtained.

[0112] In this embodiment, the solid waste resource collaborative disposal analysis system first compares the prediction accuracy of the models before and after pruning. The prediction accuracy is determined based on the deviation between the predicted results output by the model and the actual data. The difference in deviation between the models before and after pruning is calculated to obtain the prediction accuracy loss. Next, the system compares the computational efficiency of the models before and after pruning. The computational efficiency is determined based on the time required for the model to process a unit of data. The difference in processing time between the models before and after pruning is calculated to obtain the computational efficiency improvement. Then, the system compares the prediction accuracy loss with a preset loss threshold and the computational efficiency improvement with a preset improvement threshold. If the prediction accuracy loss is lower than the preset loss threshold and the computational efficiency improvement is higher than the preset improvement threshold, it indicates that the pruned model meets the accuracy and efficiency requirements, and the pruned neural network intelligent decision-making model is determined as the optimized neural network intelligent decision-making model. If the conditions are not met, the system adjusts the pruning threshold and the parameters of the pruning constraint function, for example, increasing the pruning threshold to reduce the number of redundant neurons and decreasing the upper limit of accuracy loss to reduce the pruning amplitude. Steps 410 to 450 are repeated until a satisfactory optimized neural network intelligent decision-making model is obtained.

[0113] It should be understood that those skilled in the art, when implementing the aforementioned co-processing technology for solid waste resources, can effectively overcome computational problems related to parameters with different physical meanings and dimensions through adaptive normalization, ensuring the overall feasibility of the solution. In multiple stages such as solid waste composition analysis, process parameter optimization, and model prediction, various parameter types are involved, such as solid waste component proportion (dimensionless), co-excitation temperature (°C), molding pressure (MPa), compressive strength (MPa), and heavy metal leaching concentration (mg / L). These parameters have significantly different dimensions, and directly performing feature fusion, correlation calculation, or objective function construction may lead to dimensional mismatch issues. To address this, those skilled in the art can use conventional data preprocessing methods such as standardization and normalization to convert parameters with different dimensions to a unified or dimensionless range, eliminating the impact of dimensional differences.

[0114] Specifically, in the feature fusion and correlation analysis stage, for multi-source data such as near-infrared spectral high-dimensional data sequences, inorganic element proportion feature vectors, and soluble ion concentration bond-value pair arrays, as well as cross-dimensional correlation processing of solid waste characteristic data and process data, Z-Score standardization and Min-Max normalization can be used to map various types of data to the same numerical range, ensuring that the weight allocation of parameters of different dimensions is reasonable during feature fusion and that the correlation calculation results have practical physical meaning. In the construction of multi-objective optimization functions, for optimization objectives of different dimensions such as compressive strength compliance values, impermeability and frost resistance compliance values, and heavy metal leaching concentration limits, normalization processing can unify the quantitative indicators of each objective to a standard range such as [0, 1], making the comprehensive objective value after linear weighted integration comparable and ensuring the rationality of the optimization direction and the effectiveness of the optimization results.

[0115] It is important to emphasize that adaptive normalization of parameters with different physical meanings and dimensions to overcome dimensional errors is a conventional technique that can be implemented by those skilled in the art based on existing technology. In data-driven industrial process optimization, intelligent model training, and related fields, data preprocessing methods such as standardization and normalization have become fundamental and mature technical solutions, widely used to eliminate dimensional differences, improve data compatibility, and enhance model training effectiveness. Those skilled in the art can select appropriate normalization methods based on the characteristics of specific parameters and computational requirements, completing the relevant processing without additional creative effort. This ensures the smooth implementation of the aforementioned solid waste resource co-processing technology, achieving accurate correlation between solid waste components and process parameters, efficient optimization of process parameters, and reliable assessment of target product performance and environmental risks.

[0116] Therefore, although the above technical solutions involve parameters with different dimensions, the potential dimension mismatch problem can be effectively solved by the conventional adaptive normalization process in the field, ensuring the operability and implementation effect of the solution. Such normalization is not an innovative means that requires breaking through the existing technology, but a routine operation for those skilled in the art when implementing similar data-driven technical solutions.

[0117] This application's embodiments achieve a fundamental shift in the co-processing of solid waste resources from an experience-dependent model to a data-driven model. The overall innovation lies in the deep synergy and breakthroughs in the multi-stage technological logic: First, relying on high-precision solid waste component proportion and characteristic information generated by sensor analysis, it overcomes the blindness of process parameters caused by the lack of raw material characteristic data and insufficient precision in traditional solid waste treatment. Second, by mining the correlation features between solid waste components and co-processing effects through associated datasets and determining initial production process parameters, a direct mapping relationship between raw material characteristics and process parameters is established, avoiding the problem of disconnect between process parameters and raw material characteristics in traditional methods. Third, using preset maintenance conditions as constraints, through cross-dimensional correlation processing of the feature interaction layer of a neural network intelligent decision-making model, it simultaneously achieves target product performance prediction and hazardous component leaching risk assessment, breaking through the limitations of traditional single-viewpoint methods. This approach overcomes the limitations of standard prediction by achieving collaborative prediction of multi-dimensional indicators. Furthermore, it creates a multi-objective optimization function by combining performance prediction results with leaching risk assessment results, incorporating mechanical performance, durability, and environmental risk into a unified optimization framework. This solves the problem of neglecting certain aspects in traditional single-objective optimization. Finally, iterative optimization of initial production process parameters is achieved using a particle swarm optimization algorithm, generating multiple sets of candidate process parameters. These parameters are then verified by a neural network intelligent decision-making model to determine the final optimized production process parameters. This achieves globally optimal configuration of process parameters, ensuring that the target product meets mechanical performance and durability requirements while minimizing the risk of hazardous component leaching. Ultimately, this achieves a win-win situation of efficient collaborative disposal of solid waste resources, stable product performance compliance, and controllable environmental risks, improving the intelligence level and overall benefits of solid waste resource collaborative disposal. In this way, efficient collaborative disposal of solid waste resources and multi-objective performance compliance can be achieved.

[0118] See Figure 2 As shown in the figure, this is a schematic diagram of the basic structure of a solid waste resource co-processing analysis system 20 provided in an embodiment of this application. The solid waste resource co-processing analysis system 20 includes: Processor 201; Storage device 202, on which computer program 2020 is stored; When the computer program 2020 is executed by the processor 201, the processor 201 implements any of the described neural network-optimized solid waste resource collaborative disposal analysis methods.

[0119] Based on the above, a readable storage medium is provided, on which a program or instructions are stored, and when the program or instructions are executed by a processor, the steps of the above method are implemented.

[0120] See Figure 3 As shown in the figure, this is a functional block diagram of a solid waste resource co-processing analysis device provided in an embodiment of this application. The solid waste resource co-processing analysis device includes: The component data acquisition module is used to collect raw material component data of solid waste resources and perform sensor and chemical analysis to generate sensor and chemical analysis information containing the proportion and characteristics of solid waste components. The process parameter determination module is used to combine the sensor and laboratory analysis information with the raw material composition data to generate a solid waste co-processing correlation dataset, mine the correlation features between solid waste composition and co-processing effect based on the correlation dataset, and determine the initial production process parameters based on the correlation features. The intelligent decision output module is used to input the associated dataset and the initial production process parameters into the neural network intelligent decision model under the constraints of preset maintenance conditions. The model performs cross-dimensional association processing through the feature interaction layer and generates the performance prediction results and harmful component leaching risk assessment results of the target product through the decision output layer. The optimization function creation module is used to create a multi-objective optimization function by combining the mechanical performance characteristics and service durability characteristics in the performance prediction results and the leaching characteristic characteristics in the hazardous component leaching risk assessment results. The process parameter optimization module is used to input the initial production process parameters as optimization variables into a preset particle swarm optimization algorithm, and to iteratively optimize and adjust the initial production process parameters guided by the multi-objective optimization function to generate multiple sets of process parameter optimization candidate sets. The optimization parameter generation module is used to call the neural network intelligent decision model to generate performance prediction optimization results and leaching risk assessment optimization results corresponding to each set of process parameter optimization candidate sets. Based on the performance prediction optimization results and the leaching risk assessment optimization results, production process optimization parameters including the target product ratio and optimized synergistic conditions and material forming parameters are generated.

[0121] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0122] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0123] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0124] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0125] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for co-processing solid waste resources based on neural network optimization, characterized in that, The method includes: Collect raw material composition data of solid waste resources and conduct sensor and chemical analysis to generate sensor and chemical analysis information containing the proportion and characteristics of solid waste components. By combining the sensor and laboratory analysis information with the raw material composition data, a solid waste co-processing correlation dataset is generated. Based on the correlation dataset, the correlation features between solid waste composition and co-processing effect are mined, and the initial production process parameters are determined according to the correlation features. Under the constraint of preset maintenance conditions, the associated dataset and the initial production process parameters are input into the neural network intelligent decision model. Cross-dimensional association processing is performed through the feature interaction layer, and the performance prediction results and leaching risk assessment results of the target product are generated through the decision output layer. A multi-objective optimization function is created by combining the mechanical performance characteristics and service durability characteristics in the performance prediction results and the leaching characteristics in the hazardous component leaching risk assessment results. The initial production process parameters are input as optimization variables into a preset particle swarm optimization algorithm. Guided by the multi-objective optimization function, the initial production process parameters are iteratively optimized and adjusted to generate multiple sets of process parameter optimization candidate sets. The neural network intelligent decision-making model is invoked to generate performance prediction optimization results and leaching risk assessment optimization results corresponding to each set of process parameter optimization candidate sets. Based on the performance prediction optimization results and the leaching risk assessment optimization results, production process optimization parameters including the target product ratio, optimized synergistic conditions, and material forming parameters are generated.

2. The method according to claim 1, characterized in that, The process involves inputting the associated dataset and the initial production process parameters into a neural network intelligent decision-making model, constrained by preset maintenance conditions. This model undergoes cross-dimensional correlation processing via a feature interaction layer and generates performance prediction results and hazardous component leaching risk assessment results for the target product through a decision output layer. The process includes: Based on the compositional characteristics of incineration fly ash, industrial alkali slag, and desulfurization gypsum in the solid waste resources, a corresponding preset maintenance condition boundary threshold is matched, and the preset maintenance condition boundary threshold is transformed into a constraint parameter and embedded into the feature interaction layer of the neural network intelligent decision-making model. Extract solid waste characteristic data, including the heavy metal content of incineration fly ash, the active ingredient content of industrial alkali residue, and the crystal water content of desulfurization gypsum, from the associated dataset, as well as process data, including synergistic excitation temperature and mixing molding pressure, from the initial production process parameters. After dimensional alignment processing of the solid waste characteristic data and the process data, input them into the feature interaction layer. The cross-dimensional association algorithm built into the feature interaction layer is used to perform feature fusion on the aligned solid waste characteristic data and process data to generate a fused feature vector containing the component process association relationship. The fused feature vector is input into the decision output layer of the neural network intelligent decision model. The multi-task predictor built into the decision output layer generates mechanical performance features containing compressive strength data of the target product, service durability features containing impermeability and frost resistance data as the performance prediction results, and generates leaching characteristic features containing heavy metal leaching concentration and leaching rate data as the leaching risk assessment results of the harmful components.

3. The method according to claim 2, characterized in that, The process involves inputting the fused feature vector into the decision output layer of the neural network intelligent decision-making model. The multi-task predictor built into the decision output layer generates mechanical performance characteristics including compressive strength data of the target product, and service durability characteristics including impermeability and freeze-thaw resistance data, as the performance prediction results. It also generates leaching characteristic characteristics including heavy metal leaching concentration and leaching rate data as the leaching risk assessment results for the harmful components. This includes: The fused feature vector is input into the feature enhancement module built into the decision output layer. The feature enhancement module performs dimensionality enhancement and noise filtering on the fused feature vector to generate an enhanced feature vector. The processing logic built into the feature enhancement module is based on the correlation between components and processes in the co-processing of solid waste resources. It is used to enhance the correlation between the heavy metal content of incineration fly ash, the active ingredient content of industrial alkali slag, the crystal water content of desulfurization gypsum, and the co-excitation temperature and mixing molding pressure. The enhanced feature vector is synchronously input into the mechanical performance prediction branch, durability prediction branch, and leaching risk prediction branch of the multi-task predictor. The mechanical performance prediction branch incorporates a compressive strength prediction sub-model built on a deep neural network. This sub-model is trained using historical solid waste co-processing data and is used to output compressive strength data of the target product under different curing cycles based on the component-process correlation information contained in the enhanced feature vector. The compressive strength data is then integrated into mechanical performance features. The durability prediction branch incorporates a permeability-freeze-resistance coupled prediction sub-model, which incorporates solid waste composition... The correlation function between the component and the hydrophysical properties is used to output the permeability coefficient and freeze-thaw cycle tolerance data of the target product based on the enhanced feature vector, and integrate the permeability and freeze-thaw resistance data into the service durability characteristics; the leaching risk prediction branch has a built-in heavy metal migration law prediction sub-model, which combines the distribution characteristics of heavy metal speciation in incineration fly ash with the chemical action mechanism of the synergistic activation process, and is used to output the leaching concentration and leaching rate data of heavy metals in the target product under different environmental conditions based on the enhanced feature vector, and integrate the heavy metal leaching concentration and leaching rate data into the leaching characteristic characteristics; The mechanical performance characteristics, service durability characteristics, and leaching characteristics are standardized in data format. The standardized mechanical performance characteristics and service durability characteristics are integrated into the performance prediction results, and the standardized leaching characteristics are determined as the leaching risk assessment results of the harmful components.

4. The method according to claim 1, characterized in that, The method combines the mechanical performance characteristics and service durability characteristics from the performance prediction results with the leaching characteristic characteristics from the hazardous component leaching risk assessment results to create a multi-objective optimization function, including: The first optimization objective is to take the compressive strength corresponding to the mechanical performance characteristics in the performance prediction results as the first optimization objective, the second optimization objective is to take the impermeability and frost resistance corresponding to the service durability characteristics as the second optimization objective, and the third optimization objective is to take the heavy metal leaching concentration limit corresponding to the leaching characteristic characteristics in the harmful component leaching risk assessment results as the third optimization objective, and determine the weight coefficient of each optimization objective. The objective function is constructed based on the weighting coefficients, and the synergistic activation conditions and mixing molding parameters corresponding to the initial production process parameters are used as function variables. The material consumption cost threshold and energy consumption threshold in the co-processing of solid waste resources are introduced as constraints. The first, second, and third optimization objectives are integrated into the main body of the objective function using a linear weighting method. The optimization direction of the function is defined as maximizing the degree of compliance of compressive strength and durability, and minimizing the probability of heavy metal leaching concentration exceeding the limit. This generates a multi-objective optimization function that includes the main body of the objective function, the range of variables, and the constraints.

5. The method according to claim 4, characterized in that, The first, second, and third optimization objectives are integrated into the main body of the objective function using a linear weighting method. The optimization direction of the function is defined as maximizing the degree of compliance of compressive strength and durability, and minimizing the probability of heavy metal leaching concentration exceeding the limit. This generates a multi-objective optimization function containing the main body of the objective function, the range of variables, and constraints, including: Based on industry standards and engineering practice requirements for the collaborative disposal of solid waste resources, the following benchmark values ​​were determined: the first optimization objective (compressive strength), the second optimization objective (permeability and frost resistance), and the third optimization objective (heavy metal leaching concentration limit). The analytic hierarchy process (AHP) was used to rank the importance of these three optimization objectives, and the weight coefficients for each objective were determined based on the ranking results. The sum of the weight coefficients for compressive strength compliance, durability compliance, and heavy metal leaching risk control was 1. Based on the weighting coefficients, the main structure of the objective function is constructed. The mixing ratio of incineration fly ash-industrial alkali residue-desulfurization gypsum, synergistic activation temperature, synergistic activation time, mixing and molding pressure, and molding time included in the initial production process parameters are used as function variables. The value range of each function variable is defined. The value range is determined based on the thermodynamic conditions of the synergistic activation reaction of solid waste and the operating limit parameters of the engineering equipment. A linear weighted method is used to integrate the quantitative indicators corresponding to the first, second, and third optimization objectives into the main structure of the objective function. Among them, the quantitative indicator of the first optimization objective is the ratio of the actual compressive strength to the compressive strength benchmark value; the quantitative indicator of the second optimization objective is the average of the impermeability compliance coefficient and the frost resistance compliance coefficient; and the quantitative indicator of the third optimization objective is the difference between 1 and the probability of heavy metal leaching concentration exceeding the limit. The optimization direction of the function is defined as maximizing the comprehensive objective value after linear weighting. At the same time, the material consumption cost threshold and energy consumption threshold in the co-processing of solid waste resources are introduced as constraints to construct constraint equations. Among them, the constraint equation corresponding to the material consumption cost threshold is that the sum of the products of the consumption amount of each solid waste raw material and the corresponding unit price does not exceed the preset cost threshold. The constraint equation corresponding to the energy consumption threshold is that the total energy consumption of the co-activation process and the mixing and molding process does not exceed the preset energy consumption threshold. The main structure of the objective function, function variables, optimization direction, and constraint equations are integrated to generate a complete multi-objective optimization function, which is used to reflect the correlation between each optimization objective and process parameter variables.

6. The method according to claim 1, characterized in that, The step involves inputting the initial production process parameters as optimization variables into a preset particle swarm optimization algorithm, and iteratively optimizing and adjusting the initial production process parameters guided by the multi-objective optimization function to generate multiple sets of process parameter optimization candidate sets, including: The initial parameters of the preset particle swarm optimization algorithm are set, including particle swarm size, maximum number of iterations, learning factor, and inertia weight. The mixing ratio of incineration fly ash-industrial alkali residue-desulfurized gypsum, synergistic activation temperature, synergistic activation time, mixing and molding pressure and molding time included in the initial production process parameters are used as the position vector of the particles. Using the multi-objective optimization function as the fitness function, the fitness value of each particle in the initial particle population is calculated, and the initial optimal particles whose fitness values ​​meet the preset fitness threshold are selected. Based on the initial optimal particles, the velocity vector and position vector of each particle in the population are updated. In each iteration, the process parameters corresponding to the updated particles are substituted into the multi-objective optimization function to recalculate the fitness value, and the non-dominated solution particles in the iteration process are retained. The iteration stops when the number of iterations reaches the maximum number of iterations or the change in fitness value is less than the preset convergence threshold. The process parameters corresponding to the non-dominated solution particles retained during the iteration process are deduplicated, and the process parameter combinations that meet the basic requirements for the collaborative disposal of solid waste resources are selected to generate multiple sets of process parameter optimization candidate sets.

7. The method according to claim 6, characterized in that, The process involves using the multi-objective optimization function as the fitness function to calculate the fitness value of each particle in the initial particle population, selecting the initial optimal particles whose fitness values ​​meet a preset fitness threshold, updating the velocity and position vectors of each particle in the population based on the initial optimal particles, and recalculating the fitness value by substituting the updated particle's corresponding process parameters into the multi-objective optimization function during each iteration, while retaining non-dominated solution particles from the iteration process. This includes: The position vector corresponding to each particle in the initial particle population is transformed into the corresponding process parameters of incineration fly ash-industrial alkali residue-desulfurization gypsum mixing ratio, synergistic excitation temperature, synergistic excitation time, mixing molding pressure and molding time. The process parameters are substituted into the multi-objective optimization function to calculate the comprehensive fitness value corresponding to each particle. The comprehensive fitness value is the result of linear weighting of the quantitative indicators of each optimization objective. Based on the minimum compliance requirements for the collaborative disposal of solid waste resources, a fitness preset threshold is determined, and particles with a comprehensive fitness value greater than or equal to the fitness preset threshold are selected as the initial optimal particles. The position vector and fitness value corresponding to the initial optimal particles are recorded. The individual optimal solution is determined based on the position vector corresponding to the initial optimal particle. The global optimal solution is determined by comparing the fitness values ​​of all the initial optimal particles. The velocity vector and position vector of each particle in the population are updated using a preset velocity update formula and position update formula. The velocity update formula introduces a learning factor and inertia weight to adjust the search ability of the particles. The inertia weight dynamically decays with the number of iterations to balance the global search and local search capabilities of the algorithm. After each iteration update, the process parameters corresponding to the updated position vector of each particle are extracted, and the process parameters are substituted into the multi-objective optimization function to recalculate the corresponding comprehensive fitness value. The fitness values ​​of each particle are compared based on the Pareto dominance relationship. Non-dominated solution particles that are not dominated by other particles during the iteration process are retained. The position vector and fitness value corresponding to the non-dominated solution particles are stored in the external file. At the same time, the non-dominated solution particles in the external file are maintained. When the number of particles in the external file exceeds the preset capacity, the crowding degree sorting method is used to delete particles with lower crowding degree in order to maintain the diversity of particles in the file. Repeat the steps of iterative updates, fitness value recalculation, and retention of non-dominated solution particles until the current iteration round is completed.

8. The method according to claim 1, characterized in that, The production process optimization parameters generated based on the performance prediction optimization results and the leaching risk assessment optimization results include the target formulation of the target product, optimized synergistic conditions, and material forming parameters, including: Extract the actual predicted values ​​of compressive strength and durability from the performance prediction optimization results corresponding to each set of process parameter optimization candidates, and the actual predicted values ​​of heavy metal leaching from the leaching risk assessment optimization results, and construct a parameter-performance-risk assessment matrix; Based on the evaluation matrix, a multi-attribute decision-making method is used to comprehensively sort the candidate sets for optimizing each process parameter, select the candidate set with the highest comprehensive score as the core candidate set, and retain the alternative candidate set whose comprehensive score is within a preset range. The process parameters corresponding to the core candidate set were verified to confirm that the mixing ratio of incineration fly ash-industrial alkali residue-desulfurized gypsum corresponding to the process parameters met the requirements for synergistic activation reaction, and that the temperature and time parameters in the synergistic effect conditions matched the pressure and time parameters in the material forming parameters. The mixing ratio corresponding to the core candidate set that has passed the verification is determined as the target ratio of the target product. The synergistic excitation temperature and synergistic excitation time corresponding to the core candidate set are determined as the optimized synergistic conditions. The mixing molding pressure and molding time corresponding to the core candidate set are determined as the material molding parameters. Combining the target ratio, the optimized synergistic conditions and the material molding parameters, the production process optimization parameters are generated.

9. A solid waste resource co-processing analysis system, characterized in that, include: processor; A storage device storing a computer program, which, when executed by the processor, causes the processor to implement the neural network-optimized solid waste resource co-processing analysis method as described in any one of claims 1-8.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions, which, when executed by a processor, implement the neural network-optimized solid waste resource co-disposal analysis method as described in any one of claims 1-8.

Images