A magnetization mineralization collaborative distributed processing method and system based on an internet of things

Abstract

This invention relates to the field of distributed control technology, specifically to an IoT-based collaborative distributed processing method and system for magnetization and mineralization. The method includes: real-time acquisition of multi-source sensing data from magnetization and mineralization processing units, and uploading this data to edge control nodes in a distributed control system via an IoT communication network; receiving the multi-source sensing data at the edge control nodes, generating local state vectors based on physical consistency constraint functions and physical feasible regions, and fusing and calculating these local state vectors through inter-node collaborative communication mechanisms to obtain the global water quality status and operational status evaluation results for each magnetization and mineralization processing unit. This invention combines distributed collaborative communication and fusion calculation between edge control nodes, and introduces a state correction mechanism based on physical consistency constraints and a gradient aging suppression strategy based on logical timestamps, enabling the distributed control system to achieve robust estimation of the global water quality status.

Description

Technical Field

[0001] This invention relates to the field of distributed control technology, specifically to a distributed processing method and system for magnetization and mineralization collaboration based on the Internet of Things. Background Technology

[0002] The existing magnetized mineralized water treatment systems have the following main technical problems: The systems generally adopt a centralized control architecture, with all sensing data aggregated to a single control center for global calculation and decision-making. This model lacks an effective distributed control mechanism. When the number of processing units increases, it is easy to generate communication bandwidth bottlenecks and single-point failure risks. Moreover, the strongly coupled parameters between units (magnetic field strength, flow velocity, residence time) cannot be quickly and collaboratively adjusted through local neighborhood information, resulting in poor robustness to sensor noise, communication delay and sudden changes in local operating conditions. It is difficult to achieve consistent estimation of the global state and balanced optimization of the operating performance of each unit under dynamic water quality conditions. Summary of the Invention

[0003] The purpose of this invention is to address the problems existing in the background technology by proposing a collaborative distributed processing method and system for magnetization and mineralization based on the Internet of Things.

[0004] The technical solution of this invention: A collaborative distributed processing method for magnetization and mineralization based on the Internet of Things, comprising: S1. Real-time acquisition of multi-source sensing data from the magnetization and mineralization processing unit, and uploading to the edge control node in the distributed control system via the Internet of Things communication network; S2. Receive multi-source sensing data through edge control nodes, introduce physical consistency constraint functions and physical feasible regions based on multi-source sensing data to generate local state vectors, and perform fusion calculation on local state vectors through inter-node collaborative communication mechanisms to obtain global water quality status and operational status evaluation results of each magnetization and mineralization treatment unit. S3. Based on the global state estimation results and the operation status evaluation results of each magnetization and mineralization processing unit, a gradient aging suppression mechanism based on logical timestamps is introduced to calculate the optimal adjustment set of each magnetization and mineralization processing unit and generate a distributed collaborative control strategy. S4. The distributed collaborative control strategy is decomposed into local control commands for the corresponding magnetization and mineralization processing units, and then sent to the corresponding edge control nodes through the Internet of Things communication network for collaborative adjustment of the magnetization and mineralization process.

[0005] As a further improvement to this technical solution, in step S1, multi-source sensing data from the magnetization and mineralization processing unit is collected in real time and uploaded to the edge control node in the distributed control system via the Internet of Things communication network, including the following steps: The system collects multi-source sensing data from the magnetization and mineralization processing unit in real time, preprocesses and standardizes the multi-source sensing data, encapsulates it according to a unified data structure to generate standardized data packets, encodes and encapsulates the standardized data packets according to the Internet of Things (IoT) communication protocol to generate data frames, and sends the data frames from the magnetization and mineralization processing unit to the corresponding edge control node in the distributed control system through the IoT communication network.

[0006] As a further improvement to this technical solution, in step S2, a local state vector is generated based on multi-source sensing data by introducing a physical consistency constraint function and a physical feasible region. This local state vector is then fused and calculated through an inter-node collaborative communication mechanism to obtain the global water quality status and the operational status evaluation results of each magnetization and mineralization treatment unit. This includes the following steps: S2.1 Utilize edge control nodes to receive standardized data packets; S2.2 Establish a communication channel between each edge control node and its topological neighbor nodes; S2.3 For each magnetization and mineralization processing unit, map the multi-source sensing data to a local state vector of a unified dimension. and the local state vector Standardize to generate standardized local state vectors. ; Based on the standardized local state vector Construct physical consistency constraint functions and define the physical feasible region, and then convert the local state vectors... The modified local state vector is obtained by projecting it onto the physically feasible region. And based on the modified local state vector Generate credibility factor ; S2.4, Based on the modified local state vector and credibility factor Local fusion vectors are generated through weighted fusion. ; S2.5. Aggregate the local fusion vectors of all edge control nodes globally through a distributed computing network to generate a global fusion vector. ; S2.6, The global fusion vector is obtained by pre-setting the water quality evaluation model. Mapped to global water quality status; S2.7 For each magnetization and mineralization processing unit, combine the local fusion vector With global fusion vector Generate operational status assessment results .

[0007] As a further improvement to this technical solution, in step S2.3, a physical consistency constraint function is constructed and a physical feasible region is defined, and the local state vector is... The modified local state vector is obtained by projecting it onto the physically feasible region. And based on the modified local state vector Generate credibility factor This includes the following steps: S2.31, Based on the standardized local state vector Construct physical consistency constraint function ; S2.32 Constructing the Physically Feasible Region Based on Physical Constraint Functions ; S2.33, Based on the standardized local state vector A weighted projection optimization model is constructed by introducing a neighborhood consistency constraint term, which transforms the local state vector... Mapping to the physical feasible region Within, generate a modified local state vector. ; S2.34 Calculate the standardized local state vector With the correction of the local state vector deviation ; S2.35, Based on deviation Constructing credibility factors using exponential functions .

[0008] As a further improvement to this technical solution, in S2.33, a neighborhood consistency constraint term is introduced to construct a weighted projection optimization model, which will transform the local state vector... Mapping to the physical feasible region Within, generate a modified local state vector. This includes the following steps: For edge control nodes Determine its set of neighboring nodes. For the set of neighboring nodes Local state vectors of each edge control node Perform mean aggregation to generate neighborhood reference states. ; To address the differences in multi-source sensing data, an adaptive weight matrix is ​​introduced. Construct a weighted L2 norm; By combining the weighted L2 norm and introducing a neighborhood-based reference state... Using the neighborhood consistency constraint, a weighted projection optimization model is constructed, which transforms the local state vector... Mapping to the physically feasible region generates a modified local state vector. .

[0009] As a further improvement to this technical solution, in step S3, a gradient aging suppression mechanism based on logical timestamps is introduced to calculate the optimal adjustment set for each magnetization and mineralization processing unit, and to generate a distributed collaborative control strategy, including the following steps: S3.1 Constructing a nonlinear mechanism model based on multi-source sensing data; S3.2 Based on the nonlinear mechanism model, obtain the current global water quality status results, and obtain the current operating parameters and operating status evaluation results of each magnetization and mineralization treatment unit. The decision variables are defined as the adjustment amounts of the control parameters of each magnetization and mineralization treatment unit; Using the adjustment amount of each magnetization and mineralization treatment unit as the optimization decision variable, combined with the global fusion vector Compared with the operational status assessment results Construct a multi-objective optimization problem and introduce constraints into the multi-objective optimization problem; S3.3. The multi-objective optimization problem is decomposed into sub-problems of each edge control node using a distributed gradient descent algorithm. During the decomposition process using the distributed gradient descent algorithm, a gradient aging suppression mechanism based on logical timestamps is introduced to generate the optimal adjustment amount for each magnetization and mineralization processing unit. S3.4. Encapsulate the optimal adjustment values ​​of each magnetization and mineralization processing unit to form a set of distributed collaborative control strategies.

[0010] As a further improvement to this technical solution, in step S3.3, during the decomposition process using the distributed gradient descent algorithm, a gradient aging suppression mechanism based on logical timestamps is introduced, including the following steps: S3.31, Each edge control node Calculate the gradient based on the current decision variables and the local sub-objective function. And generate a global logical timestamp. The gradient and global logical timestamp are encapsulated into a triple and broadcast to all neighboring nodes; S3.32. Set aging window parameters based on each edge control node: aging attenuation coefficient and maximum tolerable aging steps ; S3.33, When the edge control node Local time Received from edge control node When calculating the gradient, the number of aging steps is calculated. And based on aging steps Gradient aging suppression is performed in stages to generate effective gradients; S3.34. Aggregate the effective gradients using the distributed gradient descent algorithm and update the decision variables; S3.35 After the decision variables are updated, broadcast the update to neighboring nodes based on the aging window parameters and trigger the neighboring nodes to update the aging window parameters. S3.36. Repeat steps S3.31 to S3.35 until the global convergence condition is met, and output the optimal adjustment amount of each edge control node.

[0011] As a further improvement to this technical solution, in S3.35, broadcasting to neighboring nodes based on aging window parameters and triggering neighboring nodes to update parameters includes the following steps: After the decision variables are updated, the aging window parameters are encapsulated into an aging state vector based on the current edge control node, with a global logical timestamp attached, and broadcast to all neighboring nodes. The neighboring nodes receive the aging state vector and compare it with the current aging window parameters of their own node, triggering a parameter negotiation mechanism to make the aging window parameters of all edge control nodes converge to a consistent level.

[0012] As a further improvement to this technical solution, in step S4, the distributed collaborative control strategy is decomposed into local control commands for the corresponding magnetization and mineralization processing units, and then sent to the corresponding edge control nodes through the Internet of Things communication network, including the following steps: Based on the distributed collaborative control strategy, the optimal adjustment set in the distributed collaborative control strategy is parsed into local control instructions for a single edge control node. The local control instructions are encoded and encapsulated according to the preset IoT communication protocol to generate control frames. The control frames are then sent to the corresponding edge control nodes for collaborative adjustment of the magnetization and mineralization process.

[0013] On the other hand, the present invention provides an Internet of Things-based magnetization and mineralization collaborative distributed processing system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the above-mentioned Internet of Things-based magnetization and mineralization collaborative distributed processing method.

[0014] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: by deploying IoT sensing nodes in each magnetization and mineralization treatment unit, combining distributed collaborative communication and fusion computing between edge control nodes, and introducing a state correction mechanism based on physical consistency constraints and a gradient aging suppression strategy based on logical timestamps, the distributed control system can still achieve robust estimation of global water quality status and collaborative optimization of control parameters of each unit under non-ideal conditions such as communication delay, local anomalies or sudden changes in operating conditions. This significantly improves the processing efficiency and stability of the magnetization and mineralization process, while avoiding system runaway caused by single point failures, and provides highly reliable and adaptive distributed collaborative control capabilities for complex industrial water treatment scenarios. Attached Figure Description

[0015] Figure 1 This is a flowchart of the overall method of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1: Please refer to Figure 1 As shown, this embodiment provides a distributed magnetization and mineralization collaborative processing method based on the Internet of Things, including the following steps: S1. Deploy IoT sensing nodes in each magnetization and mineralization processing unit to collect multi-source sensing data from the magnetization and mineralization processing unit in real time, and upload the data to the edge control node in the distributed control system through the IoT communication network. At the same time, number the nodes and register the topology of each magnetization and mineralization processing unit to form a set of distributed processing nodes. In this embodiment, multi-source sensing data from the magnetization and mineralization processing unit is collected in real time and uploaded to the edge control node in the distributed control system via the Internet of Things communication network, including the following steps: Multiple types of sensors are deployed inside the magnetization and mineralization treatment unit, including magnetic field strength sensors, flow / velocity sensors, and residence time estimation modules, to collect multi-source sensing data (including at least magnetic field strength) from the magnetization and mineralization treatment unit in real time. Flow rate Duration of stay The system preprocesses and standardizes multi-source sensing data (performing noise reduction filtering, outlier removal, and dimensional unification), and encapsulates it according to a unified data structure to generate standardized data packets. (Time stamps are added to the standardized data packets, and time calibration is performed based on a local clock or network time synchronization mechanism to ensure time consistency between different parameters, providing a foundation for subsequent distributed fusion.) The standardized data packets are then encoded and encapsulated according to a preset IoT communication protocol to generate data frames conforming to network transmission formats. These frames contain device identifiers, parameter fields, and time stamp information. The data frames are then transmitted from the magnetization and mineralization processing unit to the corresponding edge control node in the distributed control system (DCS) via an IoT communication network (such as a wired industrial bus or wireless communication link), achieving device-to-control communication. The data uplink transmission is as follows: Pre-processed and standardized multi-source sensing data is organized into fields according to a preset data structure, and device identifiers, parameter data, and timestamp information are sequentially filled in to form a standardized data packet. The standardized data packet is encoded based on a preset Internet of Things (MQTT) protocol, including protocol header encapsulation, data payload packaging, and verification field generation, constructing a data frame that conforms to network transmission specifications. Subsequently, the data frame is sent through the corresponding communication interface (including a wired industrial bus interface or a wireless communication module), and the data frame is scheduled and sent according to the communication link's transmission mechanism. During the sending process, a link access and transmission confirmation mechanism is executed to reliably transmit the data frame from the magnetization and mineralization processing unit to the corresponding edge control node in the distributed control system.

[0018] S2. Receive multi-source sensing data through edge control nodes, introduce physical consistency constraint functions and physical feasible regions based on multi-source sensing data to generate local state vectors, and perform fusion calculation on local state vectors through inter-node collaborative communication mechanisms to obtain global water quality status and operational status evaluation results of each magnetization and mineralization treatment unit. In this embodiment, a local state vector is generated based on a physical consistency constraint function and a physical feasible region introduced from multi-source sensing data. The local state vector is then fused and calculated through an inter-node collaborative communication mechanism to obtain the global water quality status and the operational status evaluation results of each magnetization and mineralization treatment unit. The process includes the following steps: S2.1 Utilize the edge control node to receive standardized data packets from the magnetization and mineralization processing unit, and perform time synchronization verification on the received multi-source sensing data to ensure that data from different sensors correspond to the same time point; S2.2 Establish a communication channel between each edge control node and its topology neighbor nodes; specifically: for each edge control node, first identify its set of topology neighbor nodes and assign a unique communication identifier according to the configuration of the industrial network or wireless mesh network; then complete the interconnection between nodes by establishing physical links or virtual channels, including configuring network addresses, ports and encryption authentication information; after the link is established, initialize the transmission queue and buffer, set the data sending / receiving buffer and communication retry mechanism to ensure reliable message transmission; S2.3 For each magnetization and mineralization processing unit, map the multi-source sensing data to a local state vector of a unified dimension. ( The magnetic field strength, For flow rate, For the duration of stay, For transpose operation, (For edge control node indexing), and for local state vectors Standardize to generate standardized local state vectors. Specifically: First, the magnetic field strength Flow rate Duration of stay Normalization is performed according to their respective dimensions and measurement ranges to ensure that different physical quantities are on a uniform numerical scale. Then, a mean-removal method is used to eliminate the influence of dimensional differences and numerical fluctuations on subsequent calculations. Finally, a standardized local state vector is generated. This ensures that each component is within a unified range and can be directly used for physical feasible domain projection, neighborhood consistency constraints, and weighted fusion calculations, thereby improving the comparability and stability of local state vectors in distributed fusion and optimization. Based on the standardized local state vector Construct physical consistency constraint functions and define the physical feasible region, and then convert the local state vectors... The modified local state vector is obtained by projecting it onto the physically feasible region. And based on the modified local state vector Generate credibility factor ; This involves constructing a physical consistency constraint function and defining the physical feasible region, and then using the local state vector... The modified local state vector is obtained by projecting it onto the physically feasible region. And based on the modified local state vector Generate credibility factor This includes the following steps: S2.31, Based on the standardized local state vector Construct physical consistency constraint function (Based on the standardized local state vector) Based on the magnetic field strength during magnetization and mineralization Enhancement of ion migration and nucleation deposition, residence time The effect on reaction sufficiency and flow rate To influence the effects of material transport and reaction dilution, a physical consistency constraint function is constructed. In the formula, The normalized magnetic field strength, This refers to the standardized stay time. The standardized flow rate, magnetic field strength Duration of stay The overall gain coefficient on reaction effects (such as mineralization efficiency or ion migration rate) reflects the enhancing effect of the coupling between magnetic field action and material residence time on the reaction. For flow rate The attenuation or dilution effect coefficient of the reaction system is used to characterize the effect of high flow rates on reducing reaction sufficiency or disturbing reaction equilibrium. , (coefficients determined by process experience or experiments) are used to characterize the energy and reaction balance relationships between key physical variables in the local state vector; S2.32 Constructing the Physically Feasible Region Based on Physical Constraint Functions (In the formula, Physical consistency constraint function The minimum allowable value corresponds to the lower limit of the local state satisfying the basic physical laws during the magnetization and mineralization process, ensuring that the system will not experience excessively low magnetic fields, excessively low flow rates, or insufficient reactions. Physical consistency constraint function The maximum allowable value corresponds to the physically acceptable upper limit, to avoid non-ideal or dangerous working conditions caused by excessively strong magnetic fields, excessive flow rates, or excessively long residence times. It is used to constrain local states to meet basic physical laws, thereby filtering out states that do not conform to physical laws. S2.33, Based on the standardized local state vector A weighted projection optimization model is constructed by introducing a neighborhood consistency constraint term, which transforms the local state vector... Mapping to the physical feasible region Within, generate a modified local state vector. ; Furthermore, a neighborhood consistency constraint term is introduced to construct a weighted projection optimization model, which will then be used to optimize the local state vector. Mapping to the physical feasible region Within, generate a modified local state vector. This includes the following steps: For edge control nodes Determine its set of neighboring nodes. (its set of neighboring nodes) The location of the edge control node is determined based on the physical or logical topology in the distributed control system, i.e., by selecting the connection points in pipelines, communication links, or spatial layout. Edge control nodes that are directly connected or capable of stable data exchange serve as neighboring nodes, forming a set of neighboring nodes. To ensure that local state information can be efficiently shared among nodes, and to achieve distributed collaborative computing based on neighborhood consistency, the set of neighborhood nodes... Local state vectors of each edge control node Perform mean aggregation to generate neighborhood reference states. (This neighborhood reference state is used to characterize the consistency trend between local nodes and the overall topology, providing structural priors for subsequent constraints.) To address the differences in measurement accuracy and fluctuation characteristics of multi-source sensing data (such as magnetic field strength, flow velocity, and residence time), an adaptive weighting matrix is ​​introduced. (Adaptive weight matrix) The construction process is as follows: Multi-source sensing data from each edge control node includes A physical quantity (such as magnetic field strength, flow velocity, residence time, i.e.) First, we statistically analyze the physical quantities within the time window. ( Measurement variance within (sampling periods) ( (where the variance is the index of the physical quantity). A larger variance indicates more severe measurement noise or fluctuation in that physical quantity, and therefore a smaller weight should be assigned during state correction; a diagonal weight matrix is ​​constructed accordingly. (where the diagonal elements are the reciprocals of the variances of each physical quantity and normalized), constructing a weighted L2 norm. (Used to measure the deviation between the corrected state and the original local state, where,) Let be the current local state vector to be optimized. (For transpose operation) By combining the weighted L2 norm and introducing a neighborhood-based reference state... Neighborhood consistency constraint term (i.e. item, The weights of the neighborhood consistency constraint terms (set through cross-validation or empirical values) are used to construct a weighted projection optimization model, which then transforms the local state vectors... Mapping to the physically feasible region generates a modified local state vector. By simultaneously considering the original state of the node itself, the neighborhood consistency constraint, and the uncertainty of multi-source sensing data, abnormal or physically unreasonable local states are corrected to feasible regions that satisfy physical laws and topological consistency, thereby ensuring the accuracy and reliability of distributed fusion computing and global water quality status assessment. S2.34 Calculate the standardized local state vector With the correction of the local state vector deviation It is used to quantify the degree of deviation of the current local state from the physical consistency constraint, and to provide a basis for determining the credibility of the node state in the future. In this way, it can distinguish between reliable states and states affected by anomalies or measurement errors in fusion computing and control decision-making, and enhance the robustness of global water quality state estimation and distributed control strategies. S2.35, Based on deviation Using the exponential function (i.e.) Used to measure deviation Construct credibility factors by mapping them to credibility factors between 0 and 1. (In the formula, This is the credibility factor attenuation coefficient, used to adjust the degree of influence of the deviation on credibility. The larger the value, the stronger the inhibitory effect of the deviation on credibility. It is determined through expert experience or experimental calibration.

[0019] In this embodiment, steps S2.31 to S2.35 primarily address the issue of physical inconsistencies in the state of distributed sensing data caused by local sensor noise, communication anomalies, equipment malfunctions, or sudden changes in operating conditions. Specifically, it prevents non-physically reasonable or abnormal states of the magnetization and mineralization processing unit (such as combinations of magnetic fields, flow velocities, and residence times violating the fundamental physical laws of magnetization and mineralization) from "polluting" the global water quality assessment and subsequent collaborative control decisions during the distributed fusion process. This invention introduces a physical consistency constraint function and a physical feasible region driven by the magnetization and mineralization mechanism. It utilizes neighborhood consistency constraints to construct a weighted projection optimization model, projecting local state vectors into the physical feasible region to generate corrected state vectors. Simultaneously, it generates a reliability factor through exponential decay of the deviation, thereby achieving structured correction and reliability quantification of abnormal states before fusion calculation. This significantly improves the robustness of the distributed fusion system to abnormal nodes, noise, and sudden changes in local operating conditions, avoiding information waste caused by "directly discarding abnormal data" or control deviations caused by "indiscriminately fusing outliers" in traditional methods. Ultimately, it ensures the accuracy of the global water quality state estimation and the physical rationality of the distributed collaborative control strategy.

[0020] S2.4, Based on the modified local state vector and credibility factor Local fusion vectors are generated through weighted fusion. (In the formula, For edge control nodes The neighboring nodes, Neighboring nodes Credibility factor Neighboring nodes (corrected local state vector). S2.5. Aggregate the local fusion vectors of all edge control nodes globally through a distributed computing network to generate a global fusion vector. Specifically, firstly, each edge control node calculates its own local fusion vector. It includes a timestamp and is broadcast to all nodes in the network or a designated aggregation node via a distributed computing network (such as an industrial bus, wireless mesh network, or other edge collaboration network); after receiving the local fusion vector from its neighboring nodes, each node assembles the vector according to the total number of edge control nodes. Perform cumulative summation and divide the sum by . Averaging is performed to obtain the global fusion vector. (In the formula, (This refers to the total number of edge control nodes, i.e., the number of nodes participating in distributed fusion). During the calculation process, nodes ensure that each vector corresponds to the same time step through timestamp verification, and ensure the integrity and consistency of global aggregation through message confirmation or redundant transmission mechanisms, thereby generating a global fusion vector that can be used for global water quality status assessment. S2.6, The global fusion vector is obtained by pre-setting the water quality evaluation model. Mapped to an interpretable global water quality state; specifically: the global fusion vector Each element corresponds to a specific water quality indicator input variable. For example, the elements representing magnetic field strength, flow velocity, and residence time in the fusion vector are mapped to a water quality evaluation model (including nonlinear functional relationships of indicators such as pH, turbidity, and mineralization reaction efficiency) established from historical experimental or process data. Then, the global fusion vector is... The data is input into the water quality assessment model, which calculates and outputs numerical predictions for each water quality indicator. The prediction results are then subjected to dimensional conversion and normalization to ensure comparability between different indicators. Next, each water quality indicator is classified or scored based on preset thresholds or standard ranges (e.g., pH is classified as acidic, neutral, and alkaline; turbidity as clear, slightly turbid, and severely turbid), forming an interpretable water quality state vector. Finally, all classified or scored indicators are combined to generate a comprehensive global water quality state, which can be used for intuitive display or as input to distributed control strategies to ensure that control decisions are both consistent with process principles and operable. The water quality assessment model architecture is a multiple-input multiple-output (MIMO) nonlinear mapping model: the input layer receives the global fusion vector. The various elements in the model, including magnetic field strength, flow velocity, and residence time, are processed through two fully connected hidden layers (the first layer contains 64 neurons, and the second layer contains 32 neurons, both using the ReLU activation function). The output layer outputs three continuous values ​​through a linear activation function, corresponding to pH, turbidity (NTU), and mineralization reaction efficiency (%), respectively. During model training, measured water quality indicators from historical experimental data are used as supervision signals to minimize mean squared error loss. Dropout (with a dropout rate of 0.2) is used to prevent overfitting, thereby mapping the distributed fusion global state to interpretable water quality assessment results. S2.7 For each magnetization and mineralization processing unit, combine the local fusion vector With global fusion vector Generate operational status assessment results ( (For the magnetization and mineralization processing unit index); specifically: , the local fusion vector of the magnetization and mineralization processing unit. With global fusion vector The process involves matching local and global states using differences or relative deviations, such as calculating the differences in key parameters like magnetic field strength, flow velocity, and residence time. Next, based on the deviation and preset process standards or safety thresholds, each deviation indicator is quantified or categorized to form a local deviation assessment vector. Finally, this local deviation vector is combined with the global water quality state mapping results, and a comprehensive operational status assessment result is generated through rule-based combinations. This reflects the unit's treatment efficiency, deviations from key parameters, and contribution to overall water quality; finally, it will... It is stored at the edge control node and can be reported to the control center through a distributed communication network for subsequent control strategy optimization and fault diagnosis. By comparing the differences of corresponding process parameters in the local fusion vector and the global fusion vector, the deviation of flow rate, magnetic field strength and residence time are calculated, the difference between each unit and the global coordination target in actual operation is quantified, and the basis for distributed collaborative control strategy is provided to realize parameter adjustment and optimization of each magnetization and mineralization processing unit.

[0021] S3. Based on the global state estimation results and the operation status evaluation results of each magnetization and mineralization processing unit, a gradient aging suppression mechanism based on logical timestamps is introduced to calculate the optimal adjustment set of each magnetization and mineralization processing unit and generate a distributed collaborative control strategy. In this embodiment, a gradient aging suppression mechanism based on logical timestamps is introduced to calculate the optimal adjustment set for each magnetization and mineralization processing unit, and to generate a distributed cooperative control strategy, including the following steps: S3.1, Based on multi-source sensing data (including magnetic field strength) Flow rate Duration of stay A nonlinear mechanism model is constructed. Specifically, based on multi-source sensing data from each magnetization and mineralization treatment unit, corresponding historical water quality index data, such as pH, turbidity, and mineralization reaction efficiency, are obtained. These inputs are mapped to corresponding local water quality indexes, including pH, turbidity, and mineralization reaction efficiency, through a pre-set water quality prediction model (such as a nonlinear mechanism model or fitting function), thereby obtaining local water quality index data for each treatment unit. Then, combining the promoting effect of the magnetic field on ion migration, the crystal nucleation mechanism, and the deposition kinetics during the magnetization and mineralization process, the Sigmoid function is used to establish the relationship between input variables (i.e., multi-source sensing data) and output water quality indexes. The mapping relationship of (vectors), i.e., the nonlinear mechanism model. (In the formula, For the Sigmoid function, For bias vectors, As the magnetic field strength weight, For flow rate weighting, As a weight for the length of stay, For the first The magnetic field strength of each magnetization and mineralization treatment unit For the first The flow rate of each magnetization and mineralization treatment unit For the first Residence time of each magnetization and mineralization treatment unit Indicates the first (Water quality indicators of each magnetization and mineralization treatment unit are used to characterize the influence of control parameters on the treatment effect.) Then, the model parameters are trained or fitted based on historical experimental or monitoring data, so that the model can accurately predict the impact of changes in control parameters on water quality indicators, thereby providing mechanistic constraints and decision-making basis for subsequent distributed optimization calculations. S3.2. Based on the nonlinear mechanism model, obtain the current global water quality status results and the current operating parameters of each magnetization and mineralization treatment unit. , , (from local fusion vector) Extraction: Each edge control node The local fusion vector calculated in S2.4 It is a three-dimensional vector, whose three components correspond to the magnetic field strength, flow velocity, and residence time after neighborhood weighted fusion. Because By integrating the node's own corrected local state vector (projected onto the physically feasible region) and the weighted corrected states of its neighboring nodes, it can more robustly represent the true operating parameters of the magnetization and mineralization processing unit at the current moment. Therefore, for the One magnetization and mineralization processing unit (corresponding to the edge control node) Its current operating parameters are directly taken (three components) and operational status assessment results The decision variables are defined as the adjustment amounts of the control parameters of each magnetization and mineralization treatment unit. , , The actual control variables are: , , ; Adjustment amount of each magnetization and mineralization treatment unit To optimize decision variables, a global fusion vector is combined. Compared with the operational status assessment results Construct a multi-objective optimization problem and introduce constraints into the multi-objective optimization problem; The multi-objective optimization problem is as follows: ; In the formula, This represents the total number of magnetization and mineralization treatment units. The objective function is used to optimize the overall system performance (the smaller the value, the better). The normalized (interval linear normalization) global water quality state vector (derived from the global fusion vector) (Obtained through mapping via a water quality evaluation model), including indicators such as pH, turbidity, and mineralization efficiency. This is the normalized (interval linear normalization) target water quality state vector (the water quality standard that the system expects to achieve). This represents the weight of the global water quality target, used to adjust the importance of this factor in the optimization process. The normalized (interval linear normalization) th The operational status evaluation results of each magnetization and mineralization treatment unit Optimize weights for the operating status of the magnetization and mineralization treatment unit. To control the penalty weight of the adjustment (to suppress excessive actions); This term minimizes the sum of squared deviations of the operating parameters of each unit from the global average level, driving the operating state of each magnetization and mineralization treatment unit to approach the overall coordination goal of the system. When the magnetic field strength, flow rate, or residence time of a certain unit deviates significantly from the global average, this term will impose a large penalty, thereby guiding the adjustment amount to change in the direction of reducing deviation during the optimization process. This can effectively suppress unbalanced operation between units, avoid some units being overloaded while others are idle, and reduce the overall water quality fluctuations caused by local abnormal conditions. The smaller the value, the closer the operating state of the unit is to the global average level, and the better the overall coordination of the system. Global water quality state vector Control variables of each unit The relationship is defined as follows: based on the actual topology (series or parallel) of the magnetization and mineralization treatment units. Water quality indicators of each unit (Given from the nonlinear mechanism model in S3.1) Weighted according to process sequence or flow rate: If the unit is a series structure, then... (Water quality of the last unit); if it is a parallel structure, then ,in, For the first The flow rate of each magnetization and mineralization treatment unit. This embodiment defaults to a series topology, meaning the magnetization and mineralization treatment units are connected sequentially, with the effluent from one unit serving as the influent for the next. At this time, the overall water quality status... Defined as the last unit (the The effluent water quality indicators of the unit, namely ,and There is a recursive relationship between this and the water quality in the previous unit: And the input water quality index for this unit is (The first unit input is the raw water quality index), where the recursive formula indicates that: if the intermediate unit Operating parameters , , If it deviates from the standard value, the effluent water quality will be affected. This will produce a deviation, which will then serve as the input for the next unit, affecting all subsequent units until the final water quality index is reached. Therefore, in multi-objective optimization problems The term already indirectly includes the impact of intermediate unit deviations; Constraints include: equipment operating range constraints (i.e., magnetic field strength, flow velocity, and residence time are within their respective allowable upper and lower limits). ; ; In the formula, Indicates the first The lower limit of the unit magnetic field strength (the minimum magnetic field strength allowed by the equipment, determined by the structure of the electromagnet or permanent magnet). For the first The upper limit of the unit magnetic field strength (the maximum magnetic field strength allowed by the equipment, which is limited by excitation current, heat dissipation, etc.). For the first The lower limit of the unit flow rate (to prevent the flow rate from being too low, which could lead to deposition or blockage). For the first The upper limit of the unit flow rate (to prevent excessive flow rate from causing scouring or insufficient reaction). For the first The lower limit of the unit residence time (the shortest contact time required to ensure the mineralization reaction). For the first The upper limit of unit residence time (to avoid excessively large reactor volume or excessively low treatment efficiency) and the constraint of mechanism consistency (i.e., the combination of magnetic field strength, flow rate and residence time of each magnetization and mineralization treatment unit should conform to the physical equilibrium relationship of the magnetization and mineralization process, and avoid non-physical or unattainable states: ; ; In the formula, This is the maximum allowable absolute value of the magnetic field strength adjustment within a single optimization cycle (to prevent sudden changes in the magnetic field from impacting equipment and processes). This is the maximum allowable absolute value of the flow rate adjustment within a single optimization cycle (to avoid drastic flow fluctuations affecting upstream and downstream units). The constraints include the maximum allowable absolute value of residence time adjustment within a single optimization cycle (to ensure a smooth transition of the reaction process), flow conservation constraints (i.e., the sum of the flow rates of all magnetization and mineralization treatment units should equal the total system flow rate to ensure overall water balance), and time-series dependence constraints between magnetization and mineralization treatment units (i.e., the influent or treatment start time of subsequent magnetization and mineralization treatment units should not be earlier than the effluent or treatment completion time of the preceding unit to ensure the sequentiality and continuity of the treatment process in the topology). S3.3. The distributed gradient descent algorithm is used to decompose the multi-objective optimization problem into sub-problems for each edge control node (the global multi-objective optimization problem is transformed into a solvable single-objective optimization function through weighted summation or scalarization methods, and the multi-objective optimization problem is decomposed into local sub-objective functions for each node according to the magnetization and mineralization processing unit). In the process of decomposition using the distributed gradient descent algorithm, a gradient aging suppression mechanism based on logical timestamps is introduced to generate the optimal adjustment amount for each magnetization and mineralization processing unit. Each magnetization and mineralization treatment unit Each is configured with a corresponding edge control node; In this embodiment, the gradient aging suppression mechanism based on logical timestamps is introduced to address the problem that gradient information "aging" (i.e., the received gradient corresponds to an earlier iteration) caused by communication delays, data packet loss, or asynchronous execution in edge control nodes under the Internet of Things environment, which leads to slower convergence, oscillations, or even non-convergence in distributed optimization. This invention attaches a global logical timestamp to each gradient, which is exponentially decayed or directly discarded at the receiving end according to the number of aging steps. Furthermore, aging window parameters and periodic broadcast negotiation are introduced to enable all network nodes to form a consistent judgment on the timeliness of gradients. This significantly improves the robustness to network delays and non-ideal communication conditions while ensuring the convergence accuracy of distributed optimization. Furthermore, during the decomposition process using the distributed gradient descent algorithm, a gradient aging suppression mechanism based on logical timestamps is introduced, including the following steps: S3.31, Each edge control node Based on the current local decision variables Calculate the gradient using the current adjustment amount and the local sub-objective function. (Based on the current adjustment amount) Substitute into the local sub-objective function (Local sub-objective function) The acquisition process is as follows: The global multi-objective optimization problem is... Global water quality deviation term The contribution of each magnetization and mineralization treatment unit to the overall indicators is broken down, and the results of the operation status assessment specific to that node are also considered. And the penalty term for controlling the adjustment amount; then, these sub-terms are integrated using a weighted sum method to form the first... Local sub-objective function of each edge control node This means that it not only includes the node's own control constraints and state deviations, but also incorporates neighborhood consistency terms or global index gradient terms (such as those derived through the chain rule). This decomposition method, which uses local functions to handle the pressure of global optimization, ensures that each node minimizes its local function. During this process, the global objective function can be ultimately achieved through the collaborative computation of the distributed gradient descent algorithm. To achieve uniform convergence, first calculate the influence of each term in the objective function (such as water quality index deviation term, parameter constraint term, etc.) on the control variables; then... Taking the partial derivatives for each control variable separately, we obtain the components of the gradient. and at the current iteration point Numerical calculations are performed at the local sub-objective function; if the local sub-objective function contains a nonlinear mechanism model (such as a sigmoid mapping), the derivatives of the model output with respect to the input variables are calculated layer by layer using the chain rule; finally, the partial derivatives are combined to form the gradient vector. This is used to characterize the direction and magnitude of the change in the objective function caused by the current adjustment of the control variable, thereby supporting subsequent distributed gradient updates, and to generate a global logical timestamp. (Each edge control node) During the first When performing local gradient calculations or parameter updates, the current iteration round is used. As the basic time identifier, it is combined with a local counter to generate the corresponding logical timestamp. The gradient and the global logical timestamp are encapsulated into a triple. And broadcast it to all neighboring nodes; S3.32. Set local aging window parameters based on each edge control node: aging attenuation coefficient (Controlling the exponential decay rate of the historical gradient) and the maximum tolerable number of aging steps (Gradients exceeding this maximum tolerable aging steps will be discarded). S3.33, When the edge control node Local time (i.e., the current iteration) receives data from the edge control node. When calculating the gradient, the number of aging steps is calculated. And based on aging steps Graded aging suppression is performed to generate effective gradients: like If the timestamp is abnormal, the gradient is discarded; if Then, an exponentially decaying weight is applied to the gradient to obtain the effective gradient. (The decayed gradient still retains directional information, but its amplitude decreases as the aging process progresses); if If a gradient is not found to be a valid historical gradient, it is discarded and will not be used for parameter updates in this round. Finally, all valid gradients that meet the conditions are used for subsequent distributed gradient aggregation and control variable updates. S3.34, Based on effective gradients (including its own local gradient) And from the neighbors Aggregate the data using the distributed gradient descent algorithm and update the decision variables: ; In the formula, For the first Each edge control node in the iteration round Updated decision variables, For the first Each edge control node in the iteration round The decision variables, namely the control parameter vector of the current edge control node. For the first Local sub-objective function of each edge control node In the current decision variables The gradient at a given point reflects the sensitivity of the local target to the current parameters. For the first Neighborhood of each edge control node Internal nodes Gradient information (gradient after aging suppression or weighting) The accumulation of ) is used to introduce the synergistic effect of the neighborhood. For the first The set of neighboring nodes of an edge control node, including those of the edge control node. All nodes that engage in information exchange or collaborative optimization The first after gradient aging suppression treatment The gradients of each neighboring node are used to adjust the weights or effectiveness of the neighboring gradients in the update. The learning rate; S3.35 After the decision variables are updated, broadcast the update to neighboring nodes based on the aging window parameters and trigger the neighboring nodes to update the aging window parameters. The process of broadcasting aging window parameters to neighboring nodes and triggering them to update their parameters includes the following steps: After the decision variables are updated, according to the preset synchronization cycle (e.g., every 10 iterations or a fixed time interval of 5 seconds), the aging window parameters (including the aging decay coefficient) used locally by the current edge control node are updated. and maximum tolerable aging steps It is encapsulated as an aging state vector and accompanied by a global logical timestamp, and broadcast to all neighboring nodes through the IoT communication link; The node receives an aging state vector with a global logical timestamp from its neighboring nodes and compares it with the current aging window parameters of its current node, triggering a parameter negotiation mechanism (i.e., if the difference between the two exceeds a preset threshold (such as the aging decay coefficient)). Relative deviation greater than 5% or maximum tolerable aging steps If the difference is greater than 2, the aging window parameters of this node are updated using a weighted average or median fusion method. After the update is completed, this node does not broadcast immediately, but waits for the next preset synchronization cycle before continuing to propagate the updated aging window parameters to downstream nodes to avoid the parameters oscillating in the network. Each node repeats the above broadcast and parameter negotiation process according to the preset synchronization cycle (such as every 10 iterations or a fixed time interval of 5 seconds) to make the aging window parameters of all edge control nodes in the network converge to a consistent level, ensuring the collaborative convergence of the distributed gradient descent algorithm in non-ideal network environments. S3.36. Repeat steps S3.31 to S3.35 until the global convergence condition is met (e.g., the maximum number of iterations m is reached), and output the optimal adjustment amount for each edge control node. ( Indicates the first The optimal adjustment of the magnetic field strength of each processing unit Indicates the first Optimal flow rate adjustment for each processing unit Indicates the first (Optimal adjustment of dwell time for each processing unit). S3.4 Adjust the optimal amount of each magnetization and mineralization processing unit (corresponding to the edge control node). The corresponding absolute values ​​of control parameters are encapsulated to form a set of distributed collaborative control strategies. Specifically, using the number of each edge control node as an index, the optimal adjustment amount and the corresponding absolute value of control parameters of each node are organized into structured data entries, including the target node identifier, the optimal adjustment amount of magnetic field strength, the optimal adjustment amount of flow velocity, the optimal adjustment amount of residence time, and other necessary information such as timestamps. Then, each control entry is encoded and encapsulated according to a preset Internet of Things communication protocol (such as MQTT) to generate control frames that conform to the network transmission format. Next, all control frames are aggregated into a set of distributed collaborative control strategies and stored or distributed in the distributed control system for each edge control node to receive, unpack, verify, and execute, thereby realizing the collaborative adjustment and precise control of the magnetization and mineralization processing unit.

[0022] S4. Decompose the distributed collaborative control strategy into local control commands for the corresponding magnetization and mineralization processing units, and send them to the corresponding edge control nodes through the Internet of Things communication network for collaborative adjustment of the magnetization and mineralization process. In this embodiment, the distributed collaborative control strategy is decomposed into local control commands for the corresponding magnetization and mineralization processing units, and then sent to the corresponding edge control nodes through the Internet of Things communication network, including the following steps: Based on a distributed collaborative control strategy, using the edge control node number of each magnetization and mineralization processing unit as an index, the optimal adjustment set in the distributed collaborative control strategy (including the optimal adjustment of magnetic field strength, flow rate, and residence time) is parsed into local control instructions for a single edge control node. This involves locating the corresponding optimal adjustment (including the optimal adjustment of magnetic field strength, flow rate, and residence time) from the distributed collaborative control strategy set; then extracting these parameters node by node and mapping them to the fields of the local control instructions (such as target node identifier, instruction type, timestamp, and adjustment amount), completing the conversion from a global parameter set to executable instructions for a single node, enabling each instruction to directly drive the edge control node to perform the corresponding adjustment operation. Relevant information (at least the target node identifier, instruction type field, and timestamp) is added to each local control instruction. The local control instructions are then encoded and encapsulated according to a preset IoT communication protocol (such as MQTT) to generate a control frame conforming to the network transmission format. This involves adding relevant information (at least the target node identifier, instruction type field, and timestamp) to each local control instruction. The node identifier, instruction type, timestamp, and adjustment amount are mapped according to the message structure of a preset IoT communication protocol (such as MQTT), and the instruction content is filled into the message payload. Then, according to the protocol requirements, necessary headers, topics, QoS levels, and verification information are added, and the message is serialized and encoded to generate a binary control frame that conforms to the network transmission format. The control frame is then sent to the corresponding edge control node (the distributed control system operates through an industrial wired bus or wireless communication link (such as 5G)) for coordinated adjustment of the magnetization and mineralization process. After receiving the control frame, the edge control node performs unpacking, timestamp verification, and instruction validity verification (such as checking whether the adjustment amount is within the preset safety threshold). If the verification is successful, the local actuator is driven according to the local control instruction content: the electromagnetic adjustment module adjusts the excitation current in real time according to the optimal adjustment amount of the magnetic field strength to change the magnetic field strength, the valve control module adjusts the opening of the electric valve according to the optimal adjustment amount of the flow rate to change the flow rate and flow distribution, and at the same time, it adjusts the residence time of the medium in the magnetization and mineralization treatment unit according to the optimal adjustment amount of the residence time to achieve precise mineralization reaction control.

[0023] Example 2: This example provides an IoT-based magnetization and mineralization collaborative distributed processing system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the computer program to implement the IoT-based magnetization and mineralization collaborative distributed processing method described in Example 1 above.

[0024] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A distributed processing method for magnetization and mineralization based on the Internet of Things, characterized in that, include: S1. Real-time acquisition of multi-source sensing data from the magnetization and mineralization processing unit, and uploading to the edge control node in the distributed control system via the Internet of Things communication network; S2. Receive multi-source sensing data through edge control nodes, introduce physical consistency constraint functions and physical feasible regions based on multi-source sensing data to generate local state vectors, and perform fusion calculation on local state vectors through inter-node collaborative communication mechanisms to obtain global water quality status and operational status evaluation results of each magnetization and mineralization treatment unit. S3. Based on the global state estimation results and the operation status evaluation results of each magnetization and mineralization processing unit, a gradient aging suppression mechanism based on logical timestamps is introduced to calculate the optimal adjustment set of each magnetization and mineralization processing unit and generate a distributed collaborative control strategy. S4. The distributed collaborative control strategy is decomposed into local control commands for the corresponding magnetization and mineralization processing units, and then sent to the corresponding edge control nodes through the Internet of Things communication network for collaborative adjustment of the magnetization and mineralization process.

2. The Internet of Things-based collaborative distributed processing method for magnetization and mineralization as described in claim 1, characterized in that, In step S1, multi-source sensing data from the magnetization and mineralization processing unit is collected in real time and uploaded to the edge control node in the distributed control system via the Internet of Things communication network, including the following steps: The system collects multi-source sensing data from the magnetization and mineralization processing unit in real time, preprocesses and standardizes the multi-source sensing data, encapsulates it according to a unified data structure to generate standardized data packets, encodes and encapsulates the standardized data packets according to the Internet of Things (IoT) communication protocol to generate data frames, and sends the data frames from the magnetization and mineralization processing unit to the corresponding edge control node in the distributed control system through the IoT communication network.

3. The Internet of Things-based collaborative distributed processing method for magnetization and mineralization as described in claim 1, characterized in that, In step S2, a local state vector is generated based on multi-source sensing data by introducing a physical consistency constraint function and a physical feasible region. The local state vector is then fused and calculated through an inter-node collaborative communication mechanism to obtain the global water quality status and the operational status evaluation results of each magnetization and mineralization treatment unit. This includes the following steps: S2.1 Utilize edge control nodes to receive standardized data packets; S2.2 Establish a communication channel between each edge control node and its topological neighbor nodes; S2.3 For each magnetization and mineralization processing unit, map the multi-source sensing data to a local state vector of a unified dimension. and the local state vector Standardize to generate standardized local state vectors. ; Based on the standardized local state vector Construct physical consistency constraint functions and define the physical feasible region, and then convert the local state vectors... The modified local state vector is obtained by projecting it onto the physically feasible region. And based on the modified local state vector Generate credibility factor ; S2.4, Based on the modified local state vector and credibility factor Local fusion vectors are generated through weighted fusion. ; S2.

5. Aggregate the local fusion vectors of all edge control nodes globally through a distributed computing network to generate a global fusion vector. ; S2.6, The global fusion vector is obtained by pre-setting the water quality evaluation model. Mapped to global water quality status; S2.7 For each magnetization and mineralization processing unit, combine the local fusion vector With global fusion vector Generate operational status assessment results .

4. The Internet of Things-based collaborative distributed processing method for magnetization and mineralization according to claim 3, characterized in that, In step S2.3, a physical consistency constraint function is constructed and a physical feasible region is defined, and the local state vector is... The modified local state vector is obtained by projecting it onto the physically feasible region. And based on the modified local state vector Generate credibility factor This includes the following steps: S2.31, Based on the standardized local state vector Construct physical consistency constraint function ; S2.32 Constructing the Physically Feasible Region Based on Physical Constraint Functions ; S2.33, Based on the standardized local state vector A weighted projection optimization model is constructed by introducing a neighborhood consistency constraint term, which transforms the local state vector... Mapping to the physical feasible region Within, generate a modified local state vector. ; S2.34 Calculate the standardized local state vector With the correction of the local state vector deviation ; S2.35, Based on deviation Constructing credibility factors using exponential functions .

5. The Internet of Things-based collaborative distributed processing method for magnetization and mineralization according to claim 4, characterized in that, In S2.33, a neighborhood consistency constraint term is introduced to construct a weighted projection optimization model, which will transform the local state vector... Mapping to the physical feasible region Within, generate a modified local state vector. This includes the following steps: For edge control nodes Determine its set of neighboring nodes. For the set of neighboring nodes Local state vectors of each edge control node Perform mean aggregation to generate neighborhood reference states. ; To address the differences in multi-source sensing data, an adaptive weight matrix is ​​introduced. Construct a weighted L2 norm; By combining the weighted L2 norm and introducing a neighborhood-based reference state... Using the neighborhood consistency constraint, a weighted projection optimization model is constructed, which transforms the local state vector... Mapping to the physically feasible region generates a modified local state vector. .

6. The Internet of Things-based collaborative distributed processing method for magnetization and mineralization according to claim 1, characterized in that, In step S3, a gradient aging suppression mechanism based on logical timestamps is introduced to calculate the optimal adjustment set for each magnetization and mineralization processing unit, and a distributed collaborative control strategy is generated, including the following steps: S3.1 Constructing a nonlinear mechanism model based on multi-source sensing data; S3.2 Based on the nonlinear mechanism model, obtain the current global water quality status results, and obtain the current operating parameters and operating status evaluation results of each magnetization and mineralization treatment unit. The decision variables are defined as the adjustment amounts of the control parameters of each magnetization and mineralization treatment unit; Using the adjustment amount of each magnetization and mineralization treatment unit as the optimization decision variable, combined with the global fusion vector Compared with the operational status assessment results Construct a multi-objective optimization problem and introduce constraints into the multi-objective optimization problem; S3.

3. The multi-objective optimization problem is decomposed into sub-problems of each edge control node using a distributed gradient descent algorithm. During the decomposition process using the distributed gradient descent algorithm, a gradient aging suppression mechanism based on logical timestamps is introduced to generate the optimal adjustment amount for each magnetization and mineralization processing unit. S3.

4. Encapsulate the optimal adjustment values ​​of each magnetization and mineralization processing unit to form a set of distributed collaborative control strategies.

7. The Internet of Things-based collaborative distributed processing method for magnetization and mineralization according to claim 6, characterized in that, In step S3.3, during the decomposition process using the distributed gradient descent algorithm, a gradient aging suppression mechanism based on logical timestamps is introduced, including the following steps: S3.31, Each edge control node Calculate the gradient based on the current decision variables and the local sub-objective function. And generate a global logical timestamp. The gradient and global logical timestamp are encapsulated into a triple and broadcast to all neighboring nodes; S3.

32. Set aging window parameters based on each edge control node: aging attenuation coefficient and maximum tolerable aging steps ; S3.33, When the edge control node Local time Received from edge control node When calculating the gradient, the number of aging steps is calculated. And based on aging steps Gradient aging suppression is performed in stages to generate effective gradients; S3.

34. Aggregate the effective gradients using the distributed gradient descent algorithm and update the decision variables; S3.35 After the decision variables are updated, broadcast the update to neighboring nodes based on the aging window parameters and trigger the neighboring nodes to update the aging window parameters. S3.

36. Repeat steps S3.31 to S3.35 until the global convergence condition is met, and output the optimal adjustment amount of each edge control node.

8. The Internet of Things-based collaborative distributed processing method for magnetization and mineralization according to claim 7, characterized in that, In step S3.35, broadcasting the aging window parameters to neighboring nodes and triggering them to update their parameters includes the following steps: After the decision variables are updated, the aging window parameters are encapsulated into an aging state vector based on the current edge control node, with a global logical timestamp attached, and broadcast to all neighboring nodes. The neighboring nodes receive the aging state vector and compare it with the current aging window parameters of their own node, triggering a parameter negotiation mechanism to make the aging window parameters of all edge control nodes converge to a consistent level.

9. The Internet of Things-based collaborative distributed processing method for magnetization and mineralization according to claim 1, characterized in that, In step S4, the distributed collaborative control strategy is decomposed into local control commands for the corresponding magnetization and mineralization processing units, and then sent to the corresponding edge control nodes through the Internet of Things communication network, including the following steps: Based on the distributed collaborative control strategy, the optimal adjustment set in the distributed collaborative control strategy is parsed into local control instructions for a single edge control node. The local control instructions are encoded and encapsulated according to the preset IoT communication protocol to generate control frames. The control frames are then sent to the corresponding edge control nodes for collaborative adjustment of the magnetization and mineralization process.

10. A distributed processing system for magnetization and mineralization based on the Internet of Things, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: The processor executes a computer program to implement the Internet of Things-based magnetization and mineralization collaborative distributed processing method as described in any one of claims 1-9.

Images