Recycling whole-process monitoring method and system for fluorine-containing rubber waste

Abstract

The invention discloses a full-process monitoring method and system for recovery of fluorine-containing rubber waste, and relates to the technical field of rubber waste recovery, and the method comprises the following steps: carrying out multi-mode sensing analysis on the fluorine-containing rubber waste to generate a waste characteristic parameter set; constructing a recovery process database, and obtaining real-time recovery state data; carrying out online detection according to the real-time recovery state data, and extracting a quality abnormity early warning signal; the method comprises the following steps: establishing a recovery full-process data chain, carrying out dynamic monitoring, generating recovery correction parameters, updating the recovery full-process data chain through the recovery correction parameters, and obtaining a recovery full-process optimized data chain. According to the method, the technical problems of low recovery efficiency and unstable product quality caused by monitoring lag and inaccurate regulation of the fluorine-containing rubber waste recovery process in the prior art are solved, and the technical effects of realizing real-time monitoring and accurate regulation of the whole fluorine-containing rubber waste recovery process and improving the recovery efficiency and the product quality are achieved.

Description

Technical Field

[0001] This invention relates to the field of rubber waste recycling technology, specifically to a method and system for monitoring the entire recycling process of fluorinated rubber waste. Background Technology

[0002] With the rapid development of industry, fluororubber is widely used in many fields, resulting in a growing amount of fluororubber waste. Improper handling of fluororubber waste not only wastes resources but may also cause environmental pollution, making its recycling crucial. However, current fluororubber waste recycling processes face numerous problems. Traditional monitoring methods rely on manual experience and lack precise analysis of waste characteristics, making real-time, dynamic monitoring throughout the entire process difficult. This leads to low recycling efficiency, unstable product quality, and an inability to meet the requirements of modern green, environmentally friendly, and efficient production.

[0003] Existing technologies suffer from technical problems such as lagging monitoring and inaccurate control in the recycling process of fluorinated rubber waste, resulting in low recycling efficiency and unstable product quality. Summary of the Invention

[0004] This application provides a method and system for monitoring the entire recycling process of fluororubber waste, which is intended to address the technical problems of low recycling efficiency and unstable product quality caused by lagging monitoring and inaccurate control in the existing fluororubber waste recycling process.

[0005] In view of the above problems, this application provides a method and system for monitoring the entire recycling process of fluororubber waste.

[0006] The first aspect of this application provides a method for monitoring the entire recycling process of fluororubber waste, the method comprising: Multimodal sensing analysis is performed on fluororubber waste to generate a set of waste characteristic parameters. A recycling process database is constructed, and recycling records are made according to the waste characteristic parameter set based on the recycling process database to obtain real-time recycling status data. Online detection is performed according to the real-time recycling status data, and a product quality evaluation matrix is ​​constructed based on the detection dataset to extract quality anomaly early warning signals. A recycling full-process data chain is established, and the quality anomaly early warning signals are mapped to the recycling full-process data chain for dynamic monitoring to generate recycling correction parameters. The recycling full-process data chain is updated through the recycling correction parameters to obtain an optimized recycling full-process data chain.

[0007] In a possible implementation, a multimodal sensor array is used to collaboratively detect fluororubber waste to obtain a multimodal sensing dataset; the multimodal sensing dataset is spatiotemporally aligned, and principal component analysis is performed on the initially aligned multimodal dataset to obtain a multimodal aligned dataset; confidence is evaluated based on the multimodal aligned dataset, and anomalies are removed from the multimodal aligned dataset according to the evaluation results to generate a multimodal dataset; composite feature analysis is performed based on the multimodal dataset to generate the waste feature parameter set.

[0008] In a possible implementation, a recycling process database is constructed, which includes a material property sub-database, a process parameter sub-database, and an environmental standard sub-database. The waste characteristic parameter set is dynamically matched with the material property sub-database, the process parameter sub-database, and the environmental standard sub-database to generate target recycling process data. Based on the target recycling process data, equipment control commands are activated and recorded according to the recycling time sequence to obtain a recycling record data stream. The recycling record data stream is encrypted and stored to generate a recycling process traceability chain, and the real-time recycling status data is extracted based on the recycling process traceability chain.

[0009] In a possible implementation, the molecular structure spectral features of fluorinated rubber waste are extracted based on the waste feature parameter set, and a fingerprint library of fluorinated rubber materials is constructed based on the molecular structure spectral features; the degradation degree of the material is evaluated based on the fingerprint library of fluorinated rubber materials, and a degradation degree score is generated; the thermal stability level of the fluorinated rubber waste is calculated based on the degradation degree score, and the material properties are integrated based on the thermal stability level and the fingerprint library of fluorinated rubber materials to construct the material property sub-library.

[0010] In one possible implementation, historical recycling data logs are retrieved, which contain multiple recycling process parameters and multiple waste recycling parameters, and the multiple recycling process parameters and the multiple waste recycling parameters have a corresponding relationship; a parameter association matrix is ​​constructed by associating the multiple recycling process parameters and the multiple waste recycling parameters; process control is performed using digital twins according to the parameter association matrix to generate a virtual process parameter set; and the process parameter sub-library is constructed by supplementing data blind spots based on the virtual process parameter set.

[0011] In a possible implementation, real-time online detection of fluororubber waste is performed according to the real-time recycling status data to obtain the detection dataset; a multi-dimensional quality evaluation index is constructed based on the environmental protection standard sub-library, and the detection dataset is mapped to the multi-dimensional quality evaluation index to generate a real-time quality parameter vector; dynamic weight allocation is performed based on the real-time quality parameter vector to construct a product quality evaluation matrix; matrix outlier evaluation is performed on the product quality evaluation matrix to obtain multiple outlier data points, anomaly identification is performed based on the multiple outlier data points, and quality anomaly feature vectors are extracted according to the anomaly pattern; a historical anomaly database is retrieved, and the quality anomaly feature vectors are matched with the historical anomaly database to generate a graded early warning signal, and the graded early warning signal is added to the quality anomaly early warning signal.

[0012] In a possible implementation, the product quality evaluation matrix is ​​dynamically decomposed to obtain a decomposition result, which includes an outlier residual matrix; the outlier residual matrix is ​​subjected to multidimensional spatial outlier calculation to extract multiple outlier data points; the multiple outlier data points are mapped to the spatiotemporal coordinate system of the recycling process to construct an outlier spatiotemporal correlation map; the outlier spatiotemporal correlation map is traversed to extract outlier pattern features for analysis, and the quality anomaly feature vector is generated.

[0013] In a possible implementation, the quality anomaly warning signal is mapped to the recycling end-process data chain to construct a spatiotemporal correlation graph; multiple anomaly source location labels are determined according to the spatiotemporal correlation graph; the multiple anomaly source location labels are traversed to perform multi-dimensional impact analysis on the anomaly sources, generating recycling correction parameters; the recycling correction parameters are backtracked to the recycling end-process data chain for online learning to generate data learning results; and closed-loop verification testing is performed on the recycling end-process data chain based on the data learning results to construct the recycling end-process optimized data chain.

[0014] In one possible implementation, multiple industrial IoT nodes are deployed based on the recycling process flow to collect data in real time and obtain equipment operating parameters; process parameter correlation analysis is performed according to the recycling process flow to construct a process parameter correlation network, and the coupling relationship of process parameters is identified based on the process parameter correlation network; the equipment operating parameters are synchronously simulated according to the process parameter coupling relationship to construct the full recycling process data chain.

[0015] A second aspect of this application provides a monitoring system for the entire recycling process of fluororubber waste, the system comprising: The system includes a multimodal sensing and analysis module for performing multimodal sensing and analysis on fluororubber waste and generating a set of waste characteristic parameters; a recycling record module for constructing a recycling process database and recording recycling data based on the waste characteristic parameter set to obtain real-time recycling status data; an anomaly warning signal extraction module for online detection based on the real-time recycling status data, constructing a product quality evaluation matrix based on the detection dataset, and extracting quality anomaly warning signals; and a dynamic monitoring module for establishing a full-process recycling data chain, dynamically monitoring the full-process recycling data chain by mapping the quality anomaly warning signals to the data chain, generating recycling correction parameters, and updating the full-process recycling data chain using the correction parameters to obtain an optimized full-process recycling data chain.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages: Multimodal sensing analysis is performed on fluororubber waste to generate a set of waste characteristic parameters; a recycling process database is constructed to obtain real-time recycling status data; online detection is performed according to the real-time recycling status data, and a product quality evaluation matrix is ​​constructed based on the detection dataset to extract quality anomaly early warning signals; a full-process recycling data chain is established, recycling correction parameters are generated, and the full-process recycling data chain is updated using the recycling correction parameters to obtain an optimized full-process recycling data chain. This achieves the technical effect of real-time monitoring and precise control of the entire fluororubber waste recycling process, improving recycling efficiency and product quality. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic diagram of the whole-process monitoring method for recycling fluororubber waste provided in the embodiments of this application; Figure 2 This is a schematic diagram of the entire process monitoring system for the recycling of fluorinated rubber waste provided in an embodiment of this application.

[0019] Figure labeling: Multimodal sensing analysis module 10, recovery recording module 20, abnormal early warning signal extraction module 30, dynamic monitoring module 40. Detailed Implementation

[0020] This application provides a method and system for monitoring the entire recycling process of fluororubber waste, which addresses the technical problems of low recycling efficiency and unstable product quality caused by lagging monitoring and inaccurate control in the recycling process of fluororubber waste in the prior art.

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0022] Example 1, as Figure 1 As shown, this application provides a method for monitoring the entire recycling process of fluororubber waste, the method comprising: Step S100: Perform multimodal sensing analysis on fluorinated rubber waste to generate a set of waste characteristic parameters.

[0023] Specifically, a multimodal sensor group consisting of a near-infrared spectrometer, a 3D vision system, an X-ray fluorescence spectrometer, and a microwave moisture sensor is used to collaboratively detect the chemical composition (near-infrared spectrometer, X-ray fluorescence spectrometer), 3D morphology (3D vision system), and moisture content (microwave moisture sensor) of fluororubber waste, obtaining a multimodal sensing dataset containing spectral feature data, 3D image data, elemental composition data, and moisture content data. This dataset is then spatiotemporally aligned to unify time and space benchmarks, and redundant information is removed through principal component analysis to obtain a multimodal aligned dataset. Based on the aligned dataset, confidence assessment is performed to remove low-confidence outliers, generating a multimodal dataset. Finally, through composite feature analysis, key information such as molecular structure spectral features, elemental composition ratios, 3D size parameters, and moisture content are extracted from the multimodal data to generate a waste feature parameter set characterizing the properties of fluororubber waste.

[0024] Step S200: Construct a recycling process database, and record recycling data according to the waste characteristic parameter set based on the recycling process database to obtain real-time recycling status data.

[0025] Specifically, a recycling process database is constructed, comprising a material properties sub-library, a process parameter sub-library, and an environmental standards sub-library. The material properties sub-library is formed by extracting the molecular structure and spectral characteristics of waste materials to construct a material fingerprint database and assessing degradation degree and thermal stability level. The process parameter sub-library is generated by constructing a parameter correlation matrix based on historical recycling data logs and supplementing data blind spots through digital twins. The environmental standards sub-library incorporates information such as pollutant emission standards. The waste characteristic parameter set is dynamically matched with the three sub-libraries to generate target recycling process data. Equipment control commands are activated and recorded according to the recycling time sequence to form a recycling record data stream. The data stream is encrypted and stored to generate a recycling process traceability chain, from which real-time recycling status data is extracted.

[0026] Step S300: Perform online detection according to the real-time recycling status data, construct a product quality evaluation matrix based on the detection dataset, and extract quality anomaly warning signals.

[0027] Specifically, based on real-time recycling status data, the recycling process of fluororubber waste is monitored online to obtain a test dataset containing indicators such as component purity and physical properties. An evaluation index covering environmental protection and quality dimensions is constructed based on an environmental standard sub-library, and the test data is mapped to the index to generate a real-time quality parameter vector. A dynamic weight allocation algorithm is used to assign weights to each parameter, constructing a product quality evaluation matrix. The matrix is ​​dynamically decomposed to calculate the multidimensional spatial outlier degree of the outlier residual matrix, extracting outlier data points and mapping them to the spatiotemporal coordinate system of the recycling process, constructing an outlier spatiotemporal correlation map, analyzing the outlier pattern characteristics in the map, and generating a quality anomaly feature vector. This vector is matched with a historical anomaly database to generate a graded early warning signal, which is then integrated into a quality anomaly early warning signal.

[0028] Step S400: Establish a recycling full-process data chain, dynamically monitor the recycling full-process data chain according to the quality anomaly warning signal, generate recycling correction parameters, update the recycling full-process data chain through the recycling correction parameters, and obtain an optimized recycling full-process data chain.

[0029] Specifically, industrial IoT nodes are deployed based on the recycling process flow to collect equipment operating parameters (such as temperature, pressure, and speed) in real time. A process parameter correlation network is constructed through process parameter correlation analysis to identify the coupling relationships between parameters. Equipment operating parameters are synchronously simulated according to their coupling relationships, building a data flow chain covering the entire recycling process—the full-process recycling data chain. When a quality anomaly warning signal is received, it is mapped to the data chain, constructing a spatiotemporal correlation map to locate the anomaly source. Multi-dimensional impact analysis is performed on each anomaly source (such as its impact on product quality, process efficiency, and environmental indicators), generating recycling correction parameters that include process parameter adjustments and equipment control strategy optimization. These correction parameters are then backtracked to the data chain for online learning, and the correction effect is verified through closed-loop validation testing. Finally, an optimized full-process recycling data chain is generated, enabling dynamic optimization and continuous monitoring of the recycling process.

[0030] In one possible implementation, step S100 further includes: Step S110: Cooperatively detect fluororubber waste using a multimodal sensor array to obtain a multimodal sensing dataset.

[0031] Step S120: Perform spatiotemporal alignment on the multimodal sensing dataset, and perform principal component analysis on the initial multimodal aligned dataset to obtain the multimodal aligned dataset.

[0032] Step S130: Calculate the confidence level based on the multimodal alignment dataset, and remove anomalies from the multimodal alignment dataset according to the evaluation results to generate a multimodal dataset.

[0033] Step S140: Perform composite feature analysis based on the multimodal dataset to generate the waste feature parameter set.

[0034] Specifically, a multimodal sensor array consisting of a near-infrared spectrometer, a 3D vision system, an X-ray fluorescence spectrometer, and a microwave moisture sensor is used to collaboratively detect fluororubber waste. The near-infrared spectrometer performs multi-band scanning of the waste surface to acquire molecular structural feature data of the fluororubber, which is used to analyze its chemical composition and molecular chain structure. The 3D vision system constructs a 3D point cloud model of the waste using optical imaging technology, calculating geometric parameters such as bulk density, particle size, and surface roughness. The X-ray fluorescence spectrometer scans the cross-section of the waste to generate distribution maps of heavy metal pollutants (such as iron, copper, and zinc), identifying the types and contents of impurity elements. The microwave moisture sensor uses the microwave reflection principle to perform non-contact measurement of the internal moisture content of the waste, acquiring dynamic moisture gradient data at different locations. Through synchronous data acquisition from multiple sensors, a multimodal sensing dataset containing molecular structural features, 3D geometric parameters, elemental distribution information, and moisture gradient data is formed, enabling multi-dimensional and high-precision detection of the physicochemical properties of fluororubber waste.

[0035] For the multimodal sensor dataset of fluorinated rubber waste collected by a multimodal sensor array including near-infrared spectrometer, 3D vision system, X-ray fluorescence spectrometer, and microwave moisture sensor, spatiotemporal alignment processing was first performed. By unifying the timestamps and spatial coordinates of the data from each sensor (e.g., using the coordinate system of the waste processing equipment as a reference), spatiotemporal deviations caused by differences in sensor sampling timing and physical location were eliminated, forming an initial multimodal aligned dataset. Subsequently, principal component analysis was used to reduce the dimensionality of the initial multimodal aligned dataset. Through linear transformation, high-dimensional data was mapped to a low-dimensional feature space, and redundant information (such as repeated background noise signals and non-critical feature parameters) was removed, while retaining the principal components that can characterize the core characteristics of the waste. This resulted in a multimodal aligned dataset with reduced dimensions and independent features, laying the foundation for subsequent data fusion and feature analysis.

[0036] For the multimodal aligned dataset after spatiotemporal alignment and principal component analysis, a confidence assessment method combining statistical analysis and machine learning is adopted. First, the Z-score algorithm is used to calculate the standard scores of data in each dimension to identify outliers that deviate from the mean by more than 3 standard deviations. Simultaneously, the Isolation Forest algorithm is used to detect outliers in the high-dimensional space. A binary tree is constructed to isolate each data point, and the path length of the isolated point is calculated to assess the outlier probability. A confidence threshold (e.g., 95%) is set, and low-confidence data points with an absolute Z-score > 3 or an outlier probability > 0.9 are marked as outliers and automatically removed by the data cleaning module, retaining high-confidence, valid data. Finally, a structurally clean multimodal dataset is generated to ensure the accuracy of subsequent feature analysis.

[0037] Based on the aforementioned multimodal dataset, Fourier transform was used to perform frequency domain analysis on near-infrared spectral data to extract molecular structure parameters such as CF bond vibration characteristic peaks and characteristic absorption spectra of fluorinated rubber molecules. A three-dimensional point cloud model combined with morphological algorithms was used to calculate three-dimensional geometric parameters of the waste, including bulk density, average particle size, and surface roughness. Partial least squares analysis was performed on X-ray fluorescence spectral data to determine the content of major elements such as fluorine, carbon, and hydrogen, as well as the concentration distribution of heavy metal impurities such as iron and copper. Based on time-domain reflectometry data from a microwave moisture sensor, a moisture diffusion model was constructed to obtain the internal moisture content and gradient distribution parameters of the waste. Finally, the above multi-dimensional characteristic parameters were normalized and integrated to generate a waste characteristic parameter set containing chemical composition, physical morphology, impurity content, and moisture state.

[0038] In one possible implementation, step S200 further includes: Step S210: Construct a recycling process database, which includes a material properties sub-database, a process parameter sub-database, and an environmental protection standard sub-database.

[0039] Step S220: Dynamically match the waste characteristic parameter set with the material property sub-library, the process parameter sub-library, and the environmental protection standard sub-library to generate target recycling process data.

[0040] Step S230: Based on the target recycling process data, activate the equipment control command and record it according to the recycling sequence to obtain the recycling record data stream.

[0041] Step S240: Encrypt and store the recycling record data stream, generate a recycling process traceability chain, and extract the real-time recycling status data based on the recycling process traceability chain.

[0042] Specifically, a recycling process database is constructed, comprising a material properties sub-database, a process parameter sub-database, and an environmental standards sub-database. The material properties sub-database stores the types of fluororubber (e.g., fluororubber 23, fluororubber 26, etc.), molecular weight distribution (number-average molecular weight and weight-average molecular weight data determined by gel permeation chromatography), and impurity characteristic data including heavy metal content and the proportion of foreign fibers mixed in. The process parameter sub-database records historical pyrolysis temperature curves, including heating rates at different stages, isothermal temperature values, catalyst ratios, and equipment control parameters (speed of the crushing equipment, screen aperture of the screening equipment). The environmental standards sub-database integrates compliance indicators such as national and industry-mandated fluoride emission limits and heavy metal residue thresholds, forming a comprehensive data support system covering material properties, process execution, and environmental compliance.

[0043] The waste's characteristic parameter set, including molecular structure spectral features, three-dimensional geometric parameters, elemental composition ratios, and moisture content, is dynamically matched with three sub-databases of the recycling process database. Using a spectral feature matching algorithm, the molecular structure data of the waste, such as CF bond vibration frequencies and characteristic absorption peak positions, is compared with the standard spectra of fluororubber 23 and fluororubber 26 in the material property sub-database using cosine similarity calculation. Material types with a matching degree >90% are selected. Combined with molecular weight distribution and impurity characteristics, the corresponding pyrolysis temperature range (e.g., 300-350℃ for fluororubber 26) is retrieved. Based on physical parameters such as waste bulk density and particle size, the pyrolysis temperature curves in the process parameter sub-database are matched using a parameter correlation matrix (a regression model trained on historical data). For example, when the average particle size of the waste... When the diameter is >5mm, the system automatically associates the curve with a heating rate of 10℃ / min and a constant temperature of 320℃, and increases the amount of sulfidation accelerator proportionally according to the moisture content (>15%) (the catalyst ratio increases by 0.5% for every 1% increase in moisture). The system inputs the concentration of heavy metals such as lead and cadmium in the waste and the potential release of fluorides into the compliance verification model of the environmental standard sub-library. If the lead content is >80ppm, a control command is triggered to increase the purification efficiency of the waste gas treatment equipment to 95%. Finally, through rule engine matching and model calculation of multi-dimensional data, the system integrates and generates target recycling process data that includes material degradation process parameters (pyrolysis temperature 320℃), equipment control parameters (crushing speed 1200r / min), and environmental thresholds (the frequency of real-time monitoring of fluoride emissions is increased to once per minute).

[0044] Based on target recovery process data (such as pyrolysis temperature of 320℃, crushing speed of 1200r / min, catalyst ratio adjustment coefficient, etc.), control commands are transmitted in real time to the controllers of each stage of equipment (such as crushing unit, pyrolysis reactor, catalyst automatic ratio system) through an industrial IoT platform, triggering the equipment to perform operations according to the preset recovery sequence (pretreatment, crushing, pyrolysis, purification); sensors deployed on the equipment collect operating parameters (actual temperature, speed, catalyst injection amount, etc.) in real time, and the data is packaged in timestamp order through an edge computing gateway to form a structured data stream containing equipment ID, parameter name, real-time value, and execution time; a message queue is used to realize real-time data transmission and buffering, ensuring the continuity and integrity of the recovery record data stream, and providing raw data support for subsequent traceability chain construction and real-time status analysis.

[0045] Using blockchain technology or hash encryption algorithms, information such as equipment operating parameters, process execution time, and control commands in the recycling record data stream is encrypted. Data blocks are linked chronologically through a distributed ledger or chain data structure to generate an immutable recycling process traceability chain. Smart contracts or data parsing engines are used to traverse the timestamps, equipment IDs, parameter sequences, and other information in the traceability chain to extract key status data of the current recycling stage in real time (such as real-time temperature of the cracking reactor, operating speed of the crushing equipment, and actual amount of catalyst injected). This data is then compared with preset thresholds in the target recycling process data to form real-time recycling status data reflecting process execution deviations. This data is used for quality anomaly analysis and early warning signal generation in the subsequent online detection module.

[0046] In one possible implementation, step S210 further includes: Step S211: Extract the molecular structure spectral features of fluorinated rubber waste based on the waste feature parameter set, and construct a fingerprint library of fluorinated rubber materials based on the molecular structure spectral features.

[0047] Step S212: Based on the fluorinated rubber material fingerprint library, evaluate the material degradation degree and generate a degradation degree score.

[0048] Step S213: Calculate the thermal stability level of the fluororubber waste based on the degradation score, and integrate the material properties based on the thermal stability level and the fluororubber material fingerprint library to construct the material property sub-library.

[0049] Specifically, near-infrared spectroscopic data of fluorinated rubber molecular structure (such as CF bond vibration frequency, characteristic absorption peak position and intensity) collected by a near-infrared spectrometer are extracted from the waste characteristic parameters. After standardization and normalization, baseline correction and noise interference are eliminated, a digital fingerprint template containing characteristic spectra of different types of fluorinated rubber is established to construct a fingerprint library of fluorinated rubber materials, thereby achieving accurate identification of the chemical composition of waste.

[0050] Based on the fingerprint database of fluororubber materials, the degree of molecular chain breakage and aging degradation of the material are evaluated by comparing the differences between the spectral characteristics of the waste and the standard fingerprint template (such as the degree of characteristic peak shift and the number of impurity peaks) and combining the functional group attenuation rate analysis of Fourier transform infrared spectroscopy. The degradation degree score of 0 to 100 is generated by using fuzzy comprehensive evaluation method or hierarchical analysis method to quantify the recyclability level of the waste.

[0051] Based on the degradation degree score, the thermal stability of fluororubber waste is divided into different levels (e.g., level 1 is high stability, level 2 is medium stability, and level 3 is low stability) according to preset threshold rules. For example, a degradation degree score ≥80 is set as level 1, 60~79 as level 2, and <60 as level 3. Subsequently, the thermal stability level is correlated and integrated with data such as molecular structure spectral characteristics and degradation degree score in the fluororubber material fingerprint database to form a multi-dimensional data set containing material type, spectral fingerprint, degradation degree, and thermal stability level. This data is structured and stored using a database management system to construct a material property sub-database, providing basic data support for the dynamic matching of subsequent recycling process parameters.

[0052] In one possible implementation, step S210 further includes: Step S214: Retrieve historical recycling data logs, which contain multiple recycling process parameters and multiple waste recycling parameters, and there is a corresponding relationship between the multiple recycling process parameters and the multiple waste recycling parameters.

[0053] Step S215: Construct a parameter correlation matrix by associating the multiple recycling process parameters with the multiple waste recycling parameters.

[0054] Step S216: Use digital twins to control the process according to the parameter association matrix, generate a virtual process parameter set, fill in the data blind spots based on the virtual process parameter set, and construct the process parameter sub-library.

[0055] Specifically, historical recycling data logs are retrieved from the recycling process database. These logs store multiple sets of data generated during previous recycling processes of fluororubber waste. The recycling process parameters include pyrolysis temperature, catalyst ratio, crushing equipment speed, and pyrolysis time. The waste recycling parameters include waste type degradation score, molecular structure spectral characteristics, impurity content (such as heavy metal concentration), and moisture content. After cleaning outliers and missing values ​​using data preprocessing techniques, a structured parameter correspondence dataset is formed. For example, in a certain set of data, fluororubber 26 waste (degradation score 75, moisture content 12%) corresponds to parameters such as pyrolysis temperature 320℃, catalyst ratio 5%, and crushing speed 1500 r / min. This clearly presents the actual correspondence between process parameters and waste parameters, providing a reliable data foundation for subsequent parameter correlation analysis.

[0056] The correlation coefficient between recycling process parameters (such as pyrolysis temperature, catalyst ratio, and crushing speed) and waste recycling parameters (such as degradation score, impurity content, and moisture content) in historical recycling data logs was calculated using the Pearson correlation coefficient algorithm. A threshold (such as |ρ|≥0.6) was set to screen strongly correlated parameter pairs. The Apriori algorithm was used to mine frequent itemsets between parameters and generate association rules (such as the confidence level of degradation score >70 when pyrolysis temperature >300℃ and catalyst ratio >4%). The above association results were structured to construct an m×n-dimensional parameter association matrix, where rows represent process parameters, columns represent waste parameters, and matrix elements are the correlation coefficients or confidence values ​​of parameter pairs. At the same time, the association strength was visualized through heatmaps to form an association matrix that intuitively reflects the parameter coupling relationship, providing an association model basis for digital twin simulation.

[0057] Based on the established parameter correlation matrix, a virtual simulation model of the fluororubber waste recycling process is created using digital twin technology. The parameter correlation rules in the matrix are transformed into constraints of the simulation model. By inputting key parameter combinations from historical recycling data (such as process parameter ranges corresponding to different waste types), multiple process scenarios are simulated in the virtual environment, generating a virtual process parameter set covering data blind spots. The virtual parameter set is then integrated with historical measured parameters to supplement process parameter ranges not covered by historical data (such as catalyst ratios under extreme humidity conditions). A database management system is used to store the parameter correlation matrix, historical parameters, virtual parameters, and simulation verification results, constructing a process parameter sub-library containing process parameter correlation relationships, parameter ranges, and simulation data, providing comprehensive data support for real-time recycling process parameter matching.

[0058] In one possible implementation, step S300 further includes: Step S310: Perform real-time online detection of fluororubber waste according to the real-time recycling status data to obtain the detection dataset.

[0059] Step S320: Construct a multidimensional quality evaluation index based on the environmental protection standard sub-library, map the detection dataset to the multidimensional quality evaluation index, and generate a real-time quality parameter vector.

[0060] Step S330: Based on the real-time quality parameter vector, perform dynamic weight allocation to construct the product quality evaluation matrix.

[0061] Step S340: Perform matrix outlier evaluation on the product quality evaluation matrix to obtain multiple outlier data points, identify anomalies based on the multiple outlier data points, and extract quality anomaly feature vectors according to the anomaly patterns.

[0062] Step S350: Retrieve the historical anomaly database, match the quality anomaly feature vector with the historical anomaly database, generate a graded early warning signal, and add the graded early warning signal to the quality anomaly early warning signal.

[0063] Specifically, based on real-time recycling status data, online detection equipment (such as near-infrared spectrometers, X-ray fluorescence spectrometers, microwave moisture sensors, and gas analyzers) deployed at key stages of the recycling production line dynamically collects physicochemical parameters of fluororubber waste during crushing, pyrolysis, and purification. Near-infrared spectrometers scan the waste surface in real time to obtain molecular structure spectral data to monitor the degree of material degradation; X-ray fluorescence spectrometers analyze the elemental distribution across the waste cross-section to track changes in heavy metal impurities; microwave moisture sensors non-contactly measure internal moisture content fluctuations; and gas analyzers monitor the emission concentrations of fluorides and volatile organic compounds during pyrolysis in real time. The data collected synchronously by each device is preprocessed (denoised and formatted) by an edge computing unit and integrated to form a structured detection dataset containing timestamps, process nodes, parameter types, and values, providing real-time data support for subsequent quality evaluation and anomaly early warning.

[0064] Compliance indicators such as fluoride emission limits, heavy metal residue thresholds, and energy consumption standards are extracted from the environmental standard sub-library. Combined with the physical performance requirements of the recycled products (such as particle size uniformity ≥90% and pyrolysis residue rate ≤5%), a multi-dimensional quality evaluation index system is constructed, encompassing chemical pollution, physical performance, and environmental compliance. Parameters such as fluoride ion concentration, heavy metal content, moisture content, and equipment energy consumption in the test dataset are mapped to corresponding evaluation indicators through data normalization processing (such as min-max normalization). Normalized values ​​in the 0-1 range are generated according to the indicator thresholds, and these values ​​are combined to form a real-time quality parameter vector (such as [0.85, 0.92, 0.78] corresponding to the compliance scores of fluoride emissions, heavy metal residues, and energy consumption, respectively), which intuitively reflects the current quality status of the recycled products and their compliance with environmental standards.

[0065] A fuzzy logic-based weight regulator is employed to assign real-time weights to multi-dimensional quality evaluation indicators based on the recycling stage (crushing, pyrolysis, purification) and the thermal stability level of the waste. For example, when the thermal stability level of the waste is detected as level 3 (low stability), the weight of the temperature control parameter in the pyrolysis stage is automatically increased to 35%, the weight of the fluoride emission parameter is adjusted to 30%, the weight of the heavy metal residue parameter is set to 25%, and the weight of the energy consumption parameter is set to 10%. The real-time quality parameter vector is multiplied by the weight matrix through matrix operations to generate a two-dimensional evaluation matrix containing each process node and quality indicator. For example, the matrix elements for the pyrolysis stage are [temperature compliance score 0.92 × 0.35, fluoride emission score 0.88 × 0.3, heavy metal residue score 0.95 × 0.25, energy consumption score 0.80 × 0.1], enabling quantitative assessment and visualization of quality risks at each stage.

[0066] Nonnegative matrix factorization is used to reduce the dimensionality of the product quality evaluation matrix, separating the basic matrix representing the normal mode and the outlier residual matrix containing abnormal information. By calculating the local outlier factor of each data point in the residual matrix, a threshold (e.g., outlier degree > 3σ) is set to extract outlier data points that significantly deviate from the mean. These outlier points are mapped to the spatiotemporal coordinate system of the recovery process (e.g., time axis and process node axis) to construct an outlier spatiotemporal correlation map. The abnormal trajectory formed by continuous outlier points is identified by the graph traversal algorithm (e.g., the pyrolysis temperature continuously deviates from the threshold and fluoride emissions exceed the standard within a certain period of time). Based on the trajectory features (duration, parameter fluctuation amplitude, number of related indicators), a quality anomaly feature vector containing anomaly type, severity, and occurrence node is extracted, such as [anomaly type - temperature runaway, severity - high, occurrence node - pyrolysis furnace, duration - 45 minutes].

[0067] Access the historical anomaly database, which stores feature vectors of various anomalies recorded during past recycling processes (such as fluoride leaks, abnormal catalyst ratios, equipment overheating, etc.) and their corresponding classification standards (Level I: Emergency, Level II: Warning, Level III: Alert). Using a cosine similarity algorithm, the real-time extracted quality anomaly feature vectors are matched with feature vectors in the historical database to calculate a matching score. Based on preset matching thresholds (e.g., matching score ≥85% for Level I, 60%~84% for Level II, <60% for Level III), a graded early warning signal is generated. For example, if a historical pyrolysis temperature runaway event is matched with a matching score of 92%, a Level I red warning is triggered. Finally, the graded early warning signal is integrated with the real-time anomaly features to form a complete quality anomaly early warning signal containing risk level, anomaly type, and location, which is then pushed to the system monitoring interface and relevant responsible persons.

[0068] In one possible implementation, step S340 further includes: Step S341: Perform dynamic matrix decomposition on the product quality evaluation matrix to obtain the decomposition result, which includes the outlier residual matrix.

[0069] Step S342: Calculate the outlier degree in the multidimensional space of the outlier residual matrix and extract multiple outlier data points.

[0070] Step S343: Map the multiple outlier data points to the spatiotemporal coordinate system of the recycling process to construct an outlier spatiotemporal correlation map.

[0071] Step S344: Traverse the outlier spatiotemporal correlation map to extract outlier pattern features for analysis, and generate the quality anomaly feature vector.

[0072] Specifically, dynamic mode decomposition is employed to orthogonally decompose the product quality evaluation matrix. By setting a sliding time window (e.g., 5 minutes) and process stage markers (crushing / pyrolysis / purification) as dynamic adjustment factors, the original matrix is ​​decomposed into a stable base matrix and an outlier residual matrix. The stable base matrix extracts the core feature vectors under normal operating conditions through non-negative matrix decomposition, representing the correlation pattern of each quality parameter under standard processes (e.g., the baseline mapping relationship between temperature and emission concentration during the pyrolysis stage). The outlier residual matrix is ​​obtained by calculating the difference between the original matrix and the stable base matrix, and its element values, after standardization, reflect the degree of abnormality in parameter fluctuations (e.g., the standard deviation of measured fluoride emission values ​​from the baseline value). This process is accelerated by GPU computing to achieve real-time decomposition, ensuring that single-batch matrix processing is completed within 200ms, providing real-time data support for subsequent anomaly detection.

[0073] For outlier residual matrices, a local outlier factor algorithm is employed for multidimensional spatial outlier quantification. First, the parameters of each dimension in the matrix (such as fluoride concentration, temperature, and energy consumption) are standardized to eliminate dimensional differences. Then, a local neighborhood is constructed using a sliding window (e.g., 10 consecutive sampling periods). The distance-density ratio (LOF) between each data point and other points in the neighborhood is calculated. By setting a threshold, such as LOF > 1.5, data points that significantly deviate from the local neighborhood are selected. For example, if the LOF value of the fluoride emission residual value in a certain pyrolysis stage is 2.1, exceeding the preset threshold, it is identified as an outlier data point. Finally, this method is used to extract multiple outlier data points in batches, providing precise coordinates for anomaly localization.

[0074] Multiple outlier data points were extracted and mapped to a two-dimensional spatiotemporal coordinate system with time as the horizontal axis and process node as the vertical axis, according to their corresponding collection timestamps and process node locations (e.g., equipment numbers in crushing, pyrolysis, and purification stages). Each outlier data point was marked with a different color and associated with its corresponding parameter type (e.g., fluoride concentration, equipment rotation speed) and degree of deviation. A spatiotemporal correlation graph of outliers was constructed using a graph database. Nodes represent outlier data points, and edges represent the time-series relationships or process logic connections between data points. This visually presents the persistence of anomalies in the time dimension (e.g., exceeding limits for three consecutive cycles) and the propagation path in the process space (e.g., from abnormal pyrolysis furnace temperature to increased impurity rate in the purification stage), providing visualization support for multi-dimensional anomaly analysis.

[0075] Using depth-first search, the spatiotemporal correlation graph of outliers is traversed to identify anomalous trajectories and multi-parameter correlation patterns formed by continuous outlier data points. For example, the co-occurrence pattern of excessive fluoride emissions and abnormal temperature parameters in the pyrolysis process within a certain period is extracted. The lag relationship between the two on the time axis (e.g., fluoride concentration begins to exceed the standard after 15 minutes of temperature increase) and the causal relationship in the process logic (excessive temperature leads to incomplete rubber pyrolysis) are analyzed. At the same time, the duration of the anomalous trajectory, the number of process nodes affected, the types of parameters involved (chemical indicators / physical indicators), and the degree of deviation from the threshold (e.g., residual value reaches 2.5σ) are statistically analyzed. Based on the above characteristics, a multi-dimensional quality anomaly feature vector containing anomaly type, occurrence location, time characteristics, parameter correlation, and severity level is generated, providing structured data support for subsequent anomaly matching and early warning.

[0076] In one possible implementation, step S400 further includes: Step S410: Map the quality anomaly early warning signal to the recycling full-process data chain to construct a spatiotemporal correlation map.

[0077] Step S420: Locate and identify multiple anomaly source locations according to the spatiotemporal correlation map.

[0078] Step S430: Traverse the multiple anomaly source location labels to perform multi-dimensional impact analysis on the anomaly sources and generate recovery correction parameters.

[0079] Step S440: The recycling correction parameters are traced back to the recycling full-process data chain for online learning to generate data learning results.

[0080] Step S450: Perform closed-loop verification test on the recycling process data chain based on the data learning results, and construct the optimized recycling process data chain.

[0081] Specifically, the key information carried in the quality anomaly warning signal, such as the anomaly type (e.g., excessive fluoride emissions), occurrence time stamp, and process node, is precisely matched with the equipment operating parameters (e.g., real-time temperature, rotation speed) and process correlation network recorded in the entire recycling process data chain in a spatiotemporal dimension. A two-dimensional spatiotemporal correlation map is constructed with the time axis as the horizontal axis and the process node (crushing / pyrolysis / purification) as the vertical axis. Each anomaly point in the map is marked with a diamond icon, and different colors are used to distinguish the warning level (e.g., red represents Level I warning, yellow represents Level II warning). At the same time, the deviation of the measured value of the associated anomaly parameter from the standard threshold is also shown (e.g., fluoride concentration exceeds the standard by 120%). The process flow diagram is embedded at the edge of the map to intuitively show the propagation path of the anomaly in physical space (equipment location) and time series. For example, a temperature anomaly in a pyrolysis furnace is represented by a red marker point for three consecutive sampling cycles in the spatiotemporal map, extending to the right along the time axis. At the same time, it is associated with the impurity rate anomaly point in the subsequent purification process through a dashed arrow, forming a complete anomaly spatiotemporal correlation visualization model.

[0082] Based on the constructed spatiotemporal correlation graph, node location algorithms in graph theory (such as centrality analysis and shortest path search) are used to trace the source of outlier data points. First, the earliest outlier data point in the graph is identified as an initial candidate for anomaly source. Then, by combining the logical relationships between processes in the process flow diagram (such as the material flow direction in crushing, pyrolysis, and purification), the starting node of anomaly propagation is determined. Next, by calculating the betweenness centrality of each node, the key nodes with the greatest impact on anomaly propagation are identified. For example, the temperature sensor node of a certain pyrolysis furnace has the highest betweenness centrality value, indicating that it is the core hub of anomaly propagation. Finally, these key nodes are marked with red triangles in the graph, generating multiple anomaly source location labels containing equipment number (e.g., pyrolysis furnace A-03), process stage (pyrolysis process), and anomaly type (temperature sensor drift), achieving a precise mapping from spatiotemporal distribution to physical entities.

[0083] For each anomaly source location tag, the coupling relationship between the anomaly source and upstream and downstream process parameters (such as the correlation coefficient between pyrolysis temperature and fluoride emission concentration) is analyzed by calling the parameter correlation matrix of the process parameter sub-library in the recycling process database. A digital twin model is then used to simulate the propagation path of the anomaly in the process chain and its impact on subsequent processes. Simultaneously, historical operating data of the anomaly source equipment (such as temperature fluctuation curves and motor load changes) is extracted from the equipment operation parameter sub-library of the database. The ARIMA algorithm is used to predict the impact of the anomaly on equipment lifespan or operational stability, and calibration, maintenance, or replacement instructions are generated based on the equipment maintenance manual. Furthermore, pollutant emission limits (such as gaseous fluoride ≤9 mg / m³) are matched to the environmental standards sub-library. 3The system calculates the excess levels and compliance risk levels caused by anomalies, and generates environmental compensation measures (such as increasing the operating power of waste gas treatment equipment by 10%). Finally, the system integrates the multi-dimensional analysis results, including process parameter adjustment suggestions (such as lowering the pyrolysis temperature by 5°C), equipment maintenance operation instructions, and environmental control measures, into structured recycling correction parameters, which are then pushed to each control node of the production line in real time via an industrial IoT interface for execution.

[0084] An artificial neural network algorithm is employed to learn the recycling correction parameters by backtracking them to the entire recycling process data chain online. Specifically, a multilayer perceptron model is constructed using process parameters (e.g., pyrolysis temperature, fluoride emission concentration), equipment status (e.g., sensor calibration records), and correction parameters (e.g., temperature adjustment amount, maintenance instructions) from the data chain as input layer neurons, and quality indicator optimization targets (e.g., fluoride emission compliance rate, product purity) as output layer neurons. The hidden layer weights are adjusted using a backpropagation algorithm. For example, a 5°C reduction in pyrolysis temperature is correlated with a 15% decrease in fluoride concentration in historical data. The mean square error is calculated, and the model parameters are iteratively optimized. During the learning process, the nonlinear mapping relationship between the correction parameters and the process status (e.g., the S-shaped curve characteristics of temperature adjustment magnitude and emission concentration change) is automatically extracted. Data learning results, including parameter correlation weight matrices and prediction error confidence intervals, are generated. For example, a quantification rule is output stating that when the pyrolysis temperature is ≥305°C and the sensor drift value is >0.8°C, the probability of a fluoride concentration decrease increases by 6% for every 1°C increase in temperature adjustment magnitude. This achieves adaptive learning and improved prediction capabilities of the recycling process model.

[0085] Based on the data learning results, a closed-loop verification test is initiated in the entire recycling process data chain: Optimized process parameters (such as the adjusted pyrolysis temperature threshold) and equipment control rules (such as shortening the sensor calibration cycle to 4 hours) are synchronized to the digital twin model to simulate the process operation status of 30 consecutive production cycles. The consistency between the predicted values ​​and historical measured data is compared (e.g., the prediction error rate of fluoride emission concentration is <5%). Simultaneously, key indicators of the production line under the action of the corrected parameters (such as the actual temperature fluctuation range of the pyrolysis furnace and the impurity rate in the purification section) are collected in real time through industrial IoT nodes to construct a verification dataset. Hypothesis testing (such as t-tests) is used to analyze the significant differences between the verification data and the learning results. If the measured values ​​all fall within the prediction confidence interval (e.g., 95% confidence level), the correction is deemed effective, and the optimized parameter combination (such as the updated value of the temperature-emission correlation matrix) is solidified into the process parameter sub-library of the data chain, forming an optimized recycling process data chain containing dynamic coupling relationships. If the verification fails, a relearning mechanism is triggered, returning to step S440 to adjust the machine learning model parameters until closed-loop optimization is completed.

[0086] In one possible implementation, step S450 further includes: Step S451: Deploy multiple industrial IoT nodes based on the recycling process to collect data in real time and obtain equipment operating parameters.

[0087] Step S452: Perform process parameter correlation analysis according to the recycling process flow, construct a process parameter correlation network, and identify the coupling relationship of process parameters based on the process parameter correlation network.

[0088] Step S453: Simulate the equipment operating parameters synchronously according to the coupling relationship of the process parameters to construct the entire recycling process data chain.

[0089] Specifically, industrial IoT nodes, including intelligent terminals such as temperature sensors, vibration sensors, and gas concentration detectors, are deployed on key equipment in the fluororubber waste recycling process, such as crushing, pyrolysis, and purification equipment (e.g., crushing motor speed, pyrolysis furnace pressure, and purification tower). These nodes collect equipment operating parameters (e.g., crushing motor speed, pyrolysis furnace pressure, and fluoride concentration in the purification section) in real time at a rate of seconds. Each node transmits the raw data to an edge computing unit via a wireless communication module (e.g., 5G, Wi-Fi). After filtering, noise reduction, and protocol conversion, a standardized set of equipment operating parameters (e.g., structured data containing timestamps, equipment numbers, parameter names, and values) is generated and encrypted and stored in the equipment operating parameter sub-database of the recycling process database, providing real-time data support for full-process monitoring.

[0090] Table 1: Example of real-time data collection for the recycling of fluororubber 26 waste: Table 1 Timestamp process nodes Equipment Number Collect parameters Measured value Standard threshold 2025 / 5 / 22 10:00 Crushing process B-001 motor speed 1500r / min 1450~1550r / min B-001 vibration amplitude 12.5mm / s ≤15mm / s 2025 / 5 / 22 10:05 Pyrolysis process R-003 Furnace temperature 320℃ 315~325℃ R-003 furnace pressure 101.3 kPa 100~103 kPa 2025 / 5 / 22 10:10 Purification process P-001 Fluoride concentration <![CDATA[8.2mg / m 3 ]]> <![CDATA[≤9mg / m 3 ]]> P-001 Product moisture content 0.80% ≤1.5% 2025 / 5 / 22 10:15 Central control system - Energy consumption of crushing process 250kWh - Following the logical sequence of the fluororubber waste recycling process (crushing-pyrolysis-purification), the Apriori algorithm was used to mine historical process parameters (such as crushing particle size, pyrolysis temperature, and fluoride emission concentration) stored in the recycling process database. The support and confidence levels between parameters were calculated (e.g., when the pyrolysis temperature is >300℃ and the crushing particle size is <2mm, the confidence level for fluoride emission exceeding the standard is 85%). A process parameter association network containing parameter association edges and weights was constructed. By analyzing the connection density and betweenness centrality of nodes in the network, strongly coupled parameter pairs (e.g., the correlation coefficient between pyrolysis temperature and catalyst activity reaches 0.92) and key pivotal parameters (e.g., the temperature parameter in the pyrolysis stage serves as the core node connecting the crushing and purification processes) were identified. The mutual influence paths and dependencies of each process parameter in the process were clarified, providing a structured association model for subsequent synchronous simulation and anomaly tracing.

[0091] Real-time equipment operating parameters (such as crusher motor speed, pyrolysis furnace temperature, and purification section pressure) collected through industrial IoT nodes are input into a pre-constructed process parameter association network. Dynamic synchronous simulation is performed based on the coupling relationships between parameters (e.g., a quantitative correlation showing that a 5°C increase in pyrolysis temperature leads to a 6% increase in fluoride emission concentration). A full-process virtual simulation model is constructed using digital twin technology, with equipment operating parameters for each process as input variables. The model is driven by parameter coupling relationships to simulate the cascading effects of parameter changes on upstream and downstream processes (e.g., larger crushing particle size leading to a 15-minute extension in pyrolysis time, resulting in a 3% increase in impurity rate in the purification section). During this process, spatiotemporal data on equipment status, process logic, and quality indicators are integrated to construct a comprehensive recycling data chain covering the entire process of crushing, pyrolysis, and purification. This data chain aligns the parameters of each process with a time axis as a reference, forming a dynamic digital mapping that includes parameter coupling relationships and anomaly propagation paths, providing real-time simulation support for full-process monitoring, anomaly location, and process optimization.

[0092] Example 2, based on the same inventive concept as the method for monitoring the entire recycling process of fluorinated rubber waste in the foregoing examples, such as... Figure 2 As shown, this application provides a monitoring system for the entire recycling process of fluororubber waste. The system and method embodiments in this application are based on the same inventive concept. The system includes: The multimodal sensing analysis module 10 is used to perform multimodal sensing analysis on fluorinated rubber waste and generate a set of waste characteristic parameters.

[0093] The recycling record module 20 is used to construct a recycling process database, and to record recycling data based on the waste characteristic parameter set according to the recycling process database to obtain real-time recycling status data.

[0094] The abnormal warning signal extraction module 30 is used to perform online detection according to the real-time recycling status data, construct a product quality evaluation matrix based on the detection dataset, and extract quality abnormal warning signals.

[0095] The dynamic monitoring module 40 is used to establish a recycling full-process data chain, dynamically monitor the recycling full-process data chain according to the quality anomaly early warning signal, generate recycling correction parameters, update the recycling full-process data chain through the recycling correction parameters, and obtain an optimized recycling full-process data chain.

[0096] Furthermore, the system is also used for the following functions: A multimodal sensor array is used to collaboratively detect fluororubber waste to obtain a multimodal sensor dataset. The multimodal sensor dataset is then spatiotemporally aligned, and principal component analysis is performed on the initially aligned multimodal dataset to obtain a multimodal aligned dataset. Confidence assessment is performed on the multimodal aligned dataset, and anomalies are removed based on the assessment results to generate a multimodal dataset. Composite feature analysis is then performed on the multimodal dataset to generate a set of feature parameters for the waste.

[0097] Furthermore, the system is also used for the following functions: A recycling process database is constructed, comprising a material property sub-database, a process parameter sub-database, and an environmental standard sub-database. The waste characteristic parameter set is dynamically matched with each of the three sub-databases to generate target recycling process data. Based on the target recycling process data, equipment control commands are activated and recorded according to the recycling sequence to obtain a recycling record data stream. The recycling record data stream is encrypted and stored to generate a recycling process traceability chain. Real-time recycling status data is extracted based on the recycling process traceability chain.

[0098] Furthermore, the system is also used for the following functions: Based on the waste feature parameter set, the molecular structure spectral features of fluorinated rubber waste are extracted, and a fingerprint library of fluorinated rubber materials is constructed based on the molecular structure spectral features. The degradation degree of the material is evaluated based on the fingerprint library of fluorinated rubber materials, and a degradation degree score is generated. The thermal stability level of the fluorinated rubber waste is calculated based on the degradation degree score, and the material properties are integrated based on the thermal stability level and the fingerprint library of fluorinated rubber materials to construct the material property sub-library.

[0099] Furthermore, the system is also used for the following functions: Retrieve historical recycling data logs, which contain multiple recycling process parameters and multiple waste recycling parameters, and there is a corresponding relationship between the multiple recycling process parameters and the multiple waste recycling parameters; construct a parameter association matrix based on the association between the multiple recycling process parameters and the multiple waste recycling parameters; use digital twins to control the process according to the parameter association matrix to generate a virtual process parameter set; supplement data blind spots based on the virtual process parameter set to construct the process parameter sub-library.

[0100] Furthermore, the system is also used for the following functions: Real-time online detection of fluororubber waste is performed based on the real-time recycling status data to obtain the detection dataset. A multi-dimensional quality evaluation index is constructed based on the environmental standard sub-library, and the detection dataset is mapped to the multi-dimensional quality evaluation index to generate a real-time quality parameter vector. Dynamic weight allocation is performed on the real-time quality parameter vector to construct a product quality evaluation matrix. Outlier evaluation is performed on the product quality evaluation matrix to obtain multiple outlier data points. Anomaly identification is performed based on these outlier data points, and quality anomaly feature vectors are extracted according to anomaly patterns. A historical anomaly database is retrieved, and the quality anomaly feature vectors are matched with the historical anomaly database to generate a graded early warning signal. The graded early warning signal is then added to the quality anomaly early warning signal.

[0101] Furthermore, the system is also used for the following functions: The product quality evaluation matrix is ​​subjected to dynamic matrix decomposition to obtain the decomposition result, which includes an outlier residual matrix. The outlier residual matrix is ​​subjected to multidimensional spatial outlier degree calculation to extract multiple outlier data points. The multiple outlier data points are mapped to the spatiotemporal coordinate system of the recycling process to construct an outlier spatiotemporal correlation map. The outlier spatiotemporal correlation map is traversed to extract outlier pattern features for analysis to generate the quality anomaly feature vector.

[0102] Furthermore, the system is also used for the following functions: The quality anomaly warning signal is mapped to the recycling end-process data chain to construct a spatiotemporal correlation graph; multiple anomaly source location labels are determined according to the spatiotemporal correlation graph; the multiple anomaly source location labels are traversed to perform multi-dimensional impact analysis on the anomaly sources, generating recycling correction parameters; the recycling correction parameters are backtracked to the recycling end-process data chain for online learning to generate data learning results; closed-loop verification testing is performed on the recycling end-process data chain based on the data learning results to construct the recycling end-process optimized data chain.

[0103] Furthermore, the system is also used for the following functions: Multiple industrial IoT nodes are deployed based on the recycling process flow to collect equipment operating parameters in real time; process parameter correlation analysis is performed according to the recycling process flow to construct a process parameter correlation network, and the coupling relationship of process parameters is identified based on the process parameter correlation network; the equipment operating parameters are synchronously simulated according to the process parameter coupling relationship to construct the entire recycling process data chain.

[0104] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0105] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0106] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A method for monitoring the entire recycling process of fluorinated rubber waste, characterized in that, The method includes: Multimodal sensing analysis was performed on fluorinated rubber waste to generate a set of waste characteristic parameters; A recycling process database is constructed, and recycling records are made based on the waste characteristic parameter set according to the recycling process database to obtain real-time recycling status data. Online detection is performed based on the real-time recycling status data, and a product quality evaluation matrix is ​​constructed based on the detection dataset to extract quality anomaly warning signals. A recycling end-to-end data chain is established. The quality anomaly warning signal is mapped to the recycling end-to-end data chain for dynamic monitoring. Recycling correction parameters are generated, and the recycling end-to-end data chain is updated using the recycling correction parameters to obtain an optimized recycling end-to-end data chain.

2. The method for monitoring the entire recycling process of fluorinated rubber waste as described in claim 1, characterized in that, Multimodal sensing analysis of fluororubber waste is performed to generate a set of waste characteristic parameters. The methods include: A multimodal sensing dataset was obtained by collaboratively detecting fluororubber waste using a multimodal sensor array. The multimodal sensing dataset is spatiotemporally aligned, and principal component analysis is performed on the initial multimodal aligned dataset to obtain the multimodal aligned dataset. Confidence assessment is performed on the multimodal aligned dataset, and anomalies are removed from the multimodal aligned dataset based on the assessment results to generate a multimodal dataset. Based on the multimodal dataset, composite feature analysis is performed to generate the waste feature parameter set.

3. The method for monitoring the entire recycling process of fluorinated rubber waste as described in claim 1, characterized in that, Constructing a recycling process database, and recording recycling data based on the database according to the waste characteristic parameter set to obtain real-time recycling status data, the method includes: A recycling process database is constructed, which includes a material property sub-database, a process parameter sub-database, and an environmental protection standard sub-database. The waste characteristic parameter set is dynamically matched with the material property sub-library, the process parameter sub-library, and the environmental protection standard sub-library to generate target recycling process data; Based on the target recycling process data, the activated equipment control command is recorded according to the recycling sequence to obtain a recycling record data stream; The recycling record data stream is encrypted and stored to generate a recycling process traceability chain. The real-time recycling status data is then extracted based on the recycling process traceability chain.

4. The method for monitoring the entire recycling process of fluorinated rubber waste as described in claim 3, characterized in that, The process and methods for constructing the material property sub-library include: Based on the set of waste characteristic parameters, the molecular structure spectral features of fluorinated rubber waste are extracted, and a fingerprint database of fluorinated rubber materials is constructed based on the molecular structure spectral features. The degradation degree of the fluororubber material is evaluated based on the fingerprint database of the fluororubber material, and a degradation degree score is generated. The thermal stability level of the fluororubber waste is calculated based on the degradation degree score. The material properties are then integrated with the fluororubber material fingerprint library based on the thermal stability level to construct the material property sub-library.

5. The method for monitoring the entire recycling process of fluorinated rubber waste as described in claim 3, characterized in that, The process of constructing the process parameter sub-library includes the following methods: Retrieve historical recycling data logs, which contain multiple recycling process parameters and multiple waste recycling parameters, and there is a corresponding relationship between the multiple recycling process parameters and the multiple waste recycling parameters; A parameter correlation matrix is ​​constructed by associating the multiple recycling process parameters with the multiple waste recycling parameters; The process is controlled using digital twins based on the parameter association matrix, generating a virtual process parameter set. Data blind spots are filled based on the virtual process parameter set, and the process parameter sub-library is constructed.

6. The method for monitoring the entire recycling process of fluorinated rubber waste as described in claim 3, characterized in that, Online detection is performed based on the real-time recovery status data. A product quality evaluation matrix is ​​constructed based on the detection dataset, and quality anomaly warning signals are extracted. The method includes: The fluororubber waste is subjected to real-time online detection of recycling based on the real-time recycling status data to obtain the detection dataset. Based on the environmental protection standard sub-library, a multidimensional quality evaluation index is constructed, and the detection dataset is mapped to the multidimensional quality evaluation index to generate a real-time quality parameter vector. Based on the real-time quality parameter vector, a dynamic weight allocation is performed to construct a product quality evaluation matrix; Matrix outlier evaluation is performed on the product quality evaluation matrix to obtain multiple outlier data points. Anomaly identification is performed based on the multiple outlier data points, and quality anomaly feature vectors are extracted according to the anomaly pattern. Retrieve the historical anomaly database, match the quality anomaly feature vector with the historical anomaly database, generate a graded early warning signal, and add the graded early warning signal to the quality anomaly early warning signal.

7. The method for monitoring the entire recycling process of fluorinated rubber waste as described in claim 1, characterized in that, The product quality evaluation matrix is ​​subjected to matrix outlier evaluation to obtain multiple outlier data points. Anomaly identification is performed based on the multiple outlier data points, and quality anomaly feature vectors are extracted according to the anomaly patterns. The method includes: The product quality evaluation matrix is ​​subjected to dynamic matrix decomposition to obtain the decomposition result, which includes the outlier residual matrix. The outlier residual matrix is ​​subjected to multidimensional spatial outlier calculation to extract multiple outlier data points; The multiple outlier data points are mapped to the spatiotemporal coordinate system of the recycling process to construct an outlier spatiotemporal correlation map; The outlier spatiotemporal correlation graph is traversed to extract outlier pattern features for analysis, thereby generating the quality anomaly feature vector.

8. The method for monitoring the entire recycling process of fluorinated rubber waste as described in claim 1, characterized in that, The method involves dynamically monitoring the recycling process data chain by mapping the quality anomaly warning signal to the recycling process data chain, generating recycling correction parameters, and updating the recycling process data chain using these correction parameters to obtain an optimized recycling process data chain. The quality anomaly early warning signal is mapped to the entire recycling process data chain to construct a spatiotemporal correlation map; Based on the spatiotemporal correlation map, the location and identification of multiple anomaly sources are determined; The multiple anomaly source location labels are traversed to perform a multi-dimensional impact analysis on the anomaly sources, and recovery correction parameters are generated. The recycling correction parameters are traced back to the entire recycling process data chain for online learning to generate data learning results. Based on the data learning results, a closed-loop verification test is performed on the recycling process data chain to construct an optimized recycling process data chain.

9. The method for monitoring the entire recycling process of fluorinated rubber waste as described in claim 1, characterized in that, The process of establishing a complete data chain for recycling includes the following methods: Multiple industrial IoT nodes are deployed based on the recycling process to collect data in real time and obtain equipment operating parameters. According to the recycling process flow, process parameter correlation analysis is performed to construct a process parameter correlation network, and the coupling relationship of process parameters is identified based on the process parameter correlation network. The equipment operating parameters are synchronously simulated according to the coupling relationship of the process parameters to construct the entire recycling process data chain.

10. A monitoring system for the entire recycling process of fluororubber waste, characterized in that, The system is used to implement the whole-process monitoring method for the recycling of fluorinated rubber waste according to any one of claims 1-9, and the system includes: The multimodal sensing analysis module is used to perform multimodal sensing analysis on fluororubber waste and generate a set of waste characteristic parameters. The recycling record module is used to construct a recycling process database, and to record recycling data based on the waste characteristic parameter set according to the recycling process database to obtain real-time recycling status data. The abnormal warning signal extraction module is used to perform online detection according to the real-time recycling status data, construct a product quality evaluation matrix based on the detection dataset, and extract quality abnormal warning signals. The dynamic monitoring module is used to establish a recycling process data chain, dynamically monitor the recycling process data chain according to the quality anomaly warning signal, generate recycling correction parameters, update the recycling process data chain through the recycling correction parameters, and obtain an optimized recycling process data chain.

Images