An electromagnetic casting cooperative control method and system based on multi-zone parameter linkage

Abstract

The application discloses a kind of electromagnetic casting collaborative control method and system based on multi-zone parameter linkage, it is related to electromagnetic casting and industrial control technical field, including the following steps: S1, acquisition multi-zone casting parameter, compensation is synchronized by narrowband internet of things, constructs multi-zone parameter cube;S2, establish digital twin state node, staggered slice and extract phase migration, energy offset and state residual;S3, construct zone parameter coupling token, get double-axis linkage tensor by double-axis propagation and residual competition;S4, to double-axis linkage tensor executes the ablation back-propagation of placeholder, filters main control chain and offset chain;S5, construct positive and negative double strategy field, calculate multi-zone control vector by recursive distribution and competition reduction;S6, industrial control system executes multi-zone control vector, and according to execution response updates state residual, connection weight and double strategy field parameter.The application realizes multi-zone parameter coupling identification and collaborative control, improves electromagnetic casting stability and control precision.

Description

Technical Field

[0001] This invention relates to the fields of electromagnetic casting and industrial control technology, and in particular to a collaborative control method and system for electromagnetic casting based on multi-zone parameter linkage. Background Technology

[0002] Electromagnetic casting is a continuous forming technology that uses electromagnetic fields to control the flow, surface state, and solidification process of molten metal. During casting, real-time monitoring and control of temperature, liquid level, flow rate, electromagnetic intensity, cooling intensity, and casting speed are typically required. With the increasing automation of casting equipment, existing technologies are gradually adopting sensor networks, industrial control systems, and wireless communication to collect process parameters from multiple areas, and then using closed-loop control to adjust the parameters of the electromagnetic coils, liquid supply devices, cooling devices, and traction devices.

[0003] Existing electromagnetic casting control methods typically employ independent feedback control for a single region or parameter, failing to adequately consider the time lag, parameter migration, and coupling propagation relationships between different casting regions. When the electromagnetic intensity or molten metal flow rate changes in the upstream region, parameter disturbances propagate along the casting direction, causing changes in downstream temperature, liquid level, and cooling status. Existing methods struggle to accurately identify the disturbance propagation path based on the sequential relationship of parameter changes across multiple regions, easily leading to the repeated adjustment of affected parameters as independent anomalies.

[0004] Meanwhile, in multi-parameter control processes, different control actions can have mutually reinforcing or mutually canceling relationships. Existing control methods typically calculate control quantities directly based on local deviations, lacking a joint analysis mechanism for regional propagation relationships and parameter interaction relationships, making it difficult to determine the dominant control chain and mutually canceling control chains. When multiple regions simultaneously perform electromagnetic intensity, flow rate, cooling intensity, and casting speed adjustments, control conflicts, parameter oscillations, and problems of local stability leading to overall imbalance can easily occur, affecting the stability of the casting process and the forming quality.

[0005] Therefore, how to provide a collaborative control method and system for electromagnetic casting based on multi-zone parameter linkage is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose an electromagnetic casting collaborative control method and system based on multi-zone parameter linkage. This invention fully utilizes sensor networks, digital twins, dual-axis linkage analysis, and distributed control technology to achieve multi-zone parameter coupling identification, control chain screening, and collaborative adjustment in casting. It has the advantages of strong linkage, high control accuracy, and good casting stability.

[0007] An electromagnetic casting collaborative control method based on multi-zone parameter linkage according to an embodiment of the present invention includes the following steps: S1. Collect temperature, liquid level, flow rate, electromagnetic intensity and cooling intensity of each casting area through sensor network, and perform same-period sampling alignment and frame loss recursive compensation through narrowband Internet of Things to construct a multi-zone parameter cube; S2. Based on the location of the casting area and the corresponding relationship of the equipment, establish digital twin state nodes, perform misaligned slicing of the multi-zone parameter cube according to the region axis and parameter axis, and calculate the phase shift, energy offset and state residual of adjacent slices. S3. Construct a region parameter coupling token based on the phase shift, energy offset and state residual, alternately perform forward propagation of the region axis and reverse migration of the parameter axis, update the token connection weight through residual competition, and obtain the dual-axis linkage tensor. S4. Perform placeholder ablation backtracking on the dual-axis linkage tensor, freeze the region-coupled tokens one by one and recalculate the corresponding digital twin state nodes, and select the master chain and the cancellation chain based on the sign shift and amplitude jump of the node residuals. S5. Construct a positive and negative dual-strategy field based on the main control chain and the offset chain. The control increment in the same direction is recursively allocated along the main control chain, and the control increment in the opposite direction is competitively reduced along the offset chain. Multi-zone control vectors are calculated through distributed control. S6. The industrial control system adjusts the electromagnetic intensity, molten metal flow rate, cooling intensity and casting speed according to the multi-zone control vector, and writes the execution response back to the digital twin state node, and corrects the connection weight and dual-strategy field parameters according to the updated state residual.

[0008] Optionally, S1 specifically includes: The sensor network collects the temperature, liquid level, flow rate, electromagnetic intensity and cooling intensity of each casting area, writes the area number, parameter category, sampling sequence number and local timestamp to each sampling channel, and establishes a global sampling benchmark based on the industrial control system control cycle. Narrowband IoT calculates the timestamp offset of adjacent control cycles according to the sampling sequence number. For channels with the same offset direction for three consecutive sampling cycles, the sampling position is progressively shifted. For channels with reversed offset direction, the sampling position is repositioned according to the local change slope before and after the reversal point to complete the sampling alignment in the same cycle. For channels with discontinuous sampling numbers, forward change trajectories and backward constraint trajectories are established respectively, and bidirectional recursion is performed based on the change direction, change amplitude and change speed before and after the missing position; At the missing position, the forward change trajectory and the backward constraint trajectory compete point by point, retain the trajectory segments with the same direction, and correct the directional conflict position according to the synchronous change direction of the same parameters in the adjacent area to complete the frame missing recursive compensation. Establish region axis, parameter axis and time axis according to casting area, parameter type and continuous control cycle respectively. Write the sampled values ​​that have completed the same period sampling alignment and missing frame recursive compensation into the corresponding positions, and record the alignment displacement mark and compensation source mark to construct a multi-region parameter cube.

[0009] Optionally, S2 specifically includes: A digital twin state node is established based on the arrangement of the casting area, the installation position of the sensor, and the corresponding relationship of the actuator. The parameter values ​​and data processing markers are written into the multi-zone parameter cube according to the continuous control cycle. Digital twin status nodes refer to data nodes that establish a fixed mapping with the actual casting area and corresponding equipment, and are used to record the status of area parameters; Extract continuous parameter windows along the region axis, and shift the parameter windows of adjacent casting regions forward or backward by one control cycle along the time axis to obtain a region misalignment slice; The parameter window is cross-transformed along the parameter axis according to the interaction between electromagnetic intensity, temperature, liquid level, flow rate and cooling intensity to obtain parameter misalignment slices. Region misalignment slices and parameter misalignment slices are collectively referred to as misalignment slices. In the misaligned slice, the locations of reversed change direction and abrupt change in local amplitude are extracted as migration anchors. Forward pairing and reverse verification are performed based on the original control cycle index, and migration anchors with consistent bidirectional matching are retained. The phase migration amount is obtained based on the control period displacement, change direction and duration period of the migration anchor point in the adjacent misaligned slice, and the energy offset is obtained based on the difference in local fluctuation contribution between adjacent migration anchor points. The node reference state is obtained by rearranging the previous control cycle state of the same area, the synchronous state of adjacent areas, and the misaligned slices of associated parameters. The current node state is compared with the node reference state parameter by parameter to obtain the state residual including the deviation direction, the number of consecutive deviations, and the deviation magnitude. According to the digital twin state node number, parameter category and control cycle, the phase shift, energy offset and state residual are written back to the corresponding digital twin state node.

[0010] Optionally, S3 specifically includes: Based on the digital twin state node number, parameter category, and control cycle, the phase shift, energy offset, and state residual are jointly encoded and written into the regional location index, parameter category index, and control cycle index to construct a regional parameter coupling token. The zone parameter coupling token refers to a feature unit that uses a single parameter in a single casting region as the object, jointly encodes the phase shift, energy offset and state residual, and retains the correspondence between the region location, parameter type and control cycle. Establish a regional axis adjacency chain along the direction of molten metal flow, connect regional parameter coupling tokens of the same parameter category in sequence, and perform connection retention, disconnection and cross-level jump according to the phase migration direction and control cycle displacement; Starting with the pre-connected pre-connected region parameter coupling token, the phase migration direction, energy offset amplitude, and state residual are transmitted node by node along the region axis, and the forward propagation of the region axis is completed based on the consistency of direction, energy continuity, and number of residual continuations. Establish a parameter axis connection chain based on state parameters, adjustment parameters, and action parameters. Starting from the state residual after the forward propagation of the region axis, match the region parameter coupling tokens in the reverse order from state parameters to adjustment parameters and from adjustment parameters to action parameters to complete the reverse migration of the parameter axis. Perform residual competition on the region propagation component and parameter migration component of the coupling tokens that connect to the same target area, update the connection weights according to the residual direction, number of consecutive deviations and residual magnitude, and record the cancellation weights; The region axis adjacency chain and parameter axis connection chain are rearranged according to the updated connection weights. Continuously decaying connections are deleted, continuously gaining connections are retained, and bidirectional binding is established for connections that gain simultaneously during region propagation and parameter migration. Based on the region location index, parameter category index, and control cycle index, write the region parameter coupling token, connection weight, cancellation weight, and bidirectional binding flag to the corresponding positions to obtain the dual-axis linkage tensor.

[0011] Optionally, S4 specifically includes: Read the region parameter coupling token, connection weight and bidirectional binding mark in the dual-axis linkage tensor, and build the original residual snapshot according to the region location index and parameter category index of the digital twin state node; The placeholder ablation pushback is performed on the dual-axis linkage tensor. The ablation order is determined according to the number of bidirectional bindings and the connection weight. Placeholder freezing is performed on the regional parameter coupling tokens in sequence, the token index and connection topology are retained, and the placeholder state is obtained by rearranging the state of the corresponding node in the previous control cycle and the synchronization state of the adjacent nodes. The placeholder ablation pushback method sequentially performs placeholder freezing, dual-axis propagation recalculation, residual comparison and original state restoration on the regional parameter coupling token, and tracks the processing of changes in the response of the digital twin state node along the connection direction of the regional parameter coupling token. Starting from the placeholder frozen region parameter coupling token, the process is re-propagated along the region axis adjacency chain and parameter axis connection chain. The state residuals of the associated digital twin state nodes are recalculated based on the original connection weights to obtain the ablation residual snapshot. Align the ablation residual snapshot with the original residual snapshot node by node, record the changes in the direction of the state residual and the increase or decrease of the residual amplitude, and register the position where the direction of the residual amplitude change of adjacent nodes switches as the amplitude transition point. Restore the current region parameter coupling token and continue freezing the next token according to the ablation order. After traversing the dual-axis linkage tensor, obtain the ablation response trajectory corresponding to each region parameter coupling token. Cross-matching is performed on the ablation response trajectory, and the region parameter coupling tokens that have continuously decayed residuals of associated nodes after freezing and whose residual direction migrates towards the reference position are connected to the main control chain. Connect the region parameter coupling tokens that continuously enhance the residuals of the associated nodes after freezing and whose residual directions continuously deviate from the reference position into an offset chain, and write the main control chain, the offset chain and the corresponding amplitude transition point back to the dual-axis linkage tensor.

[0012] Optionally, S5 specifically includes: Read the main control chain, cancellation chain, connection weight, cancellation weight and amplitude transition point in the dual-axis linkage tensor, establish link control slots according to region location, parameter type and control period, and write them into the forward policy layer and reverse policy layer respectively to construct a positive and negative dual policy field; A link control slot refers to a data location that corresponds to a single area, a single control parameter, and a single control cycle, and is used to record the control direction, control increment, link weight, and transmission order. A dual-strategy field refers to a control relationship structure formed by superimposing the forward strategy layer and the reverse strategy layer according to the same region location index and parameter category index; The control direction is determined by the residual deviation direction of the coupled token in the starting area of ​​the main control chain. The first-level control increment is determined according to the connection weight, residual amplitude and amplitude transition point, and then passed down level by level along the main control chain. During the transmission of the main control chain, the control increment is adjusted according to the continuous increase or decrease of the residual amplitude. At the amplitude transition point, the previous increment is frozen and the subsequent control increment is redefined to complete the recursive allocation of the same-direction control increment. The reverse control increment is generated using the residual deviation direction and offset weight corresponding to the offset chain. The reverse control increment and the same-direction control increment where the link intersects are written into the same link control slot to perform contention reduction. Based on the connection weight, offset weight, and residual change status, the reverse control increment is offset, truncated, and redistributed, and the remaining reverse control increment continues to be passed along the offset chain. For the link control slots corresponding to electromagnetic intensity, molten metal flow rate, cooling intensity and casting speed in the same area, control cycle alignment is performed, and in-direction control increments are merged and local competition is performed on in-reverse control increments to obtain regional candidate control vectors. By performing cross-validation on candidate control vectors in adjacent regions according to the main control chain sequence through distributed control, the control components that enable downstream residual convergence are retained, and the control components that enhance the residual of the offset chain are withdrawn, thus obtaining multi-region control vectors.

[0013] An electromagnetic casting collaborative control system based on multi-zone parameter linkage according to an embodiment of the present invention includes: The data acquisition module is used to collect temperature, liquid level, flow rate, electromagnetic intensity and cooling intensity of each casting area through a sensor network, and to perform same-period sampling alignment and frame loss recursive compensation through narrowband Internet of Things to construct a multi-zone parameter cube; The state mapping module is used to establish digital twin state nodes based on the location of the casting area and the corresponding relationship of the equipment. It performs staggered slicing of the multi-zone parameter cube according to the region axis and parameter axis, and calculates the phase shift, energy offset and state residual of adjacent slices. The dual-axis linkage module is used to construct a region parameter coupling token based on the phase shift, energy offset and state residual, alternately execute the region axis forward propagation and parameter axis reverse migration, and update the token connection weight through residual competition to obtain the dual-axis linkage tensor; The link filtering module is used to perform placeholder ablation backtracking on the dual-axis linkage tensor, freeze the region-coupled tokens one by one and recalculate the corresponding digital twin state nodes, and filter the master chain and the offset chain based on the sign migration and amplitude transition of the node residuals. The collaborative decision-making module is used to construct a positive and negative dual-strategy field based on the main control chain and the offset chain, recursively allocate the same-direction control increment along the main control chain, competitively reduce the opposite-direction control increment along the offset chain, and calculate the multi-zone control vector through distributed control. The feedback control module is used to adjust the electromagnetic intensity, molten metal flow rate, cooling intensity and casting speed according to the multi-zone control vector of the industrial control system, write the execution response back to the digital twin state node, and correct the connection weight and dual-strategy field parameters according to the updated state residual.

[0014] The beneficial effects of this invention are: First, this invention synchronously collects temperature, liquid level, flow rate, electromagnetic intensity, and cooling intensity of each casting area through a sensor network and narrowband Internet of Things. It also achieves temporal unification of multi-source parameters by using same-period sampling alignment, frame loss recursion compensation, and multi-zone parameter cubes. By extracting the dynamic change relationship between different regions and different parameters through digital twin state nodes, misaligned slices, phase shift, energy offset, and state residuals, the completeness and accuracy of multi-zone casting state characterization are improved.

[0015] Secondly, this invention establishes a dual-axis linkage tensor through zone parameter coupling tokens, zone axis forward propagation, parameter axis reverse migration, and residual competition. It then freezes and recalculates the state of different zone parameter coupling tokens through placeholder ablation back-pushing. Based on residual sign migration and amplitude transition, it filters the main control chain and the offset chain, thereby distinguishing the parameter propagation relationship that plays a dominant role in the casting state from the control relationship that produces mutual offsetting effects. This reduces the misadjustment, repeated adjustment, and control conflict caused by traditional single-zone, single-parameter independent control.

[0016] Finally, this invention constructs a positive and negative dual-strategy field based on the main control chain and the offset chain. Through incremental recursive allocation of control in the same direction, competitive reduction of incremental control in the opposite direction, and distributed control, multi-zone control vectors are generated. The industrial control system coordinates and adjusts the electromagnetic intensity, molten metal flow rate, cooling intensity, and casting speed. At the same time, the state residual, connection weight, and dual-strategy field parameters are continuously corrected according to the execution response, so as to realize multi-zone, multi-parameter closed-loop collaborative regulation and improve the control accuracy, operational stability, and forming quality consistency of the electromagnetic casting process. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating the overall process of a collaborative control method for electromagnetic casting based on multi-zone parameter linkage proposed in this invention. Figure 2 This is a flowchart of multi-zone parameter acquisition and cube construction for an electromagnetic casting collaborative control method based on multi-zone parameter linkage proposed in this invention. Figure 3 This is a flowchart of digital twin state mapping and misaligned slicing for an electromagnetic casting collaborative control method based on multi-zone parameter linkage proposed in this invention. Figure 4 This is a flowchart illustrating the dual-axis linkage construction process of an electromagnetic casting collaborative control method based on multi-zone parameter linkage proposed in this invention. Figure 5 This is a flowchart of the occupational ablation backtracking and link screening process for an electromagnetic casting collaborative control method based on multi-zone parameter linkage proposed in this invention. Figure 6 This is a flowchart illustrating the collaborative decision-making and closed-loop update process of an electromagnetic casting collaborative control method based on multi-zone parameter linkage proposed in this invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0019] refer to Figures 1-6 A collaborative control method for electromagnetic casting based on multi-zone parameter linkage includes the following steps: S1. Collect temperature, liquid level, flow rate, electromagnetic intensity and cooling intensity of each casting area through sensor network, and perform same-period sampling alignment and frame loss recursive compensation through narrowband Internet of Things to construct a multi-zone parameter cube; S2. Based on the location of the casting area and the corresponding relationship of the equipment, establish digital twin state nodes, perform misaligned slicing of the multi-zone parameter cube according to the region axis and parameter axis, and calculate the phase shift, energy offset and state residual of adjacent slices. S3. Construct a region parameter coupling token based on the phase shift, energy offset and state residual, alternately perform forward propagation of the region axis and reverse migration of the parameter axis, update the token connection weight through residual competition, and obtain the dual-axis linkage tensor. S4. Perform placeholder ablation backtracking on the dual-axis linkage tensor, freeze the region-coupled tokens one by one and recalculate the corresponding digital twin state nodes, and select the master chain and the cancellation chain based on the sign shift and amplitude jump of the node residuals. S5. Construct a positive and negative dual-strategy field based on the main control chain and the offset chain. The control increment in the same direction is recursively allocated along the main control chain, and the control increment in the opposite direction is competitively reduced along the offset chain. Multi-zone control vectors are calculated through distributed control. S6. The industrial control system adjusts the electromagnetic intensity, molten metal flow rate, cooling intensity and casting speed according to the multi-zone control vector, and writes the execution response back to the digital twin state node, and corrects the connection weight and dual-strategy field parameters according to the updated state residual.

[0020] In this embodiment, S1 specifically includes: The temperature, liquid level, flow rate, electromagnetic intensity and cooling intensity of each casting area are collected by a sensor network. The area number, parameter category, sampling sequence number and local timestamp are written to each sampling channel. A global sampling benchmark is established based on the control cycle of the industrial control system. After receiving data from each sampling channel, the narrowband IoT retrieves the timestamp offset within adjacent control cycles according to the sampling sequence number, and performs a progressive translation of the sampling position for channels with the same offset direction for three consecutive sampling cycles. For channels where the offset direction is reversed, the reversal point is retained and the sampling position is repositioned according to the local change slope before and after the reversal point. This process is defined as same-period sampling alignment. Same-period sampling alignment refers to rearranging the sampled values ​​corresponding to different regions and different parameters to the same control period based on the global sampling benchmark. For channels with discontinuous sampling sequence numbers, establish forward change trajectories and backward constraint trajectories; The forward change trajectory is obtained by recursively calculating the change direction, change magnitude and change speed of the continuous sampled values ​​before the missing position, while the backward constraint trajectory is expanded in the opposite direction from the first valid sampled value after the missing position to the missing position. The two trajectories compete point by point at the missing position, retaining the trajectory segments with the same direction of change. For positions with conflicting directions, the compensation value is corrected according to the synchronous change direction of the same parameters in adjacent areas. This process is defined as frame missing recursive compensation. Frame loss recursive compensation refers to reconstructing missing sample values ​​by utilizing the parameter change trajectory before and after the missing position and the synchronous response of similar parameters in adjacent areas; After completing the same-cycle sampling alignment and missing frame recursive compensation, a region axis is established according to the arrangement order of the casting areas, a parameter axis is established according to temperature, liquid level, flow rate, electromagnetic intensity and cooling intensity, and a time axis is established according to the continuous control cycle. Each sampled value is written into the corresponding region position, parameter position and time position, and the original sampling mark, alignment displacement mark and compensation source mark are recorded simultaneously to construct a multi-region parameter cube. A multi-zone parameter cube is a three-dimensional data structure that is defined by a zone axis, a parameter axis, and a time axis, and preserves the continuous parameter status and data processing markers of each casting zone.

[0021] In this embodiment, S2 specifically includes: Based on the arrangement of each casting zone along the direction of molten metal flow, the installation position of the sensors, and the correspondence between the electromagnetic coil, the liquid supply actuator, and the cooling actuator, a digital twin state node is established. Each digital twin state node is bound to a casting zone and the corresponding sampling positions of temperature, liquid level, flow rate, electromagnetic intensity, and cooling intensity. The parameter values, alignment displacement marks, and compensation source marks are written into the multi-zone parameter cube according to the continuous control cycle. A digital twin state node refers to a data node that maintains a fixed mapping relationship with the actual casting area and corresponding equipment, and stores the state of the area parameters according to the control cycle. Extract the continuous parameter window of the current casting region along the region axis of the multi-zone parameter cube, and at the same time shift the parameter windows of adjacent casting regions forward by one control cycle and backward by one control cycle along the time axis to obtain a region misalignment slice; Cross-swapping is performed along the parameter axis with electromagnetic intensity as the action parameter, temperature and liquid level as the state parameters, and flow rate and cooling intensity as the adjustment parameters. Different parameter windows within the same control cycle are recombined according to the corresponding order of action parameter-state parameter and state parameter-adjustment parameter to obtain misaligned slices. Misaligned slicing refers to a parameter subsequence obtained by offsetting the region position, time position, or parameter position while retaining the original control cycle index; Search for locations of continuous rise to fall, continuous fall to rise, and local amplitude abrupt change in the region misalignment slice and parameter misalignment slice respectively, register the corresponding locations as migration anchor points, and perform forward pairing and reverse verification on the migration anchor points in different slices according to the original control cycle index, and delete migration anchor points whose forward pairing and reverse verification results are inconsistent. The periodic displacement is obtained by subtracting the control period index of the retained migration anchor point in the adjacent misaligned slice. The periodic displacement is then assigned a positive or negative label according to the direction of change, and normalized in combination with the duration period of the corresponding migration anchor point to obtain the phase migration amount.

[0022] Phase shift refers to the position and direction of control cycle shift that occurs between different misaligned slices or parameter misaligned slices of the same parameter change characteristic. The local fluctuation range is defined by two adjacent migration anchor points. The continuous change amplitude of the parameter within the range relative to the starting value of the range is accumulated, and the fluctuation contribution is recorded according to the rising segment, falling segment and reversal segment respectively. The contribution difference of the corresponding fluctuation range of adjacent slices is recorded as the energy offset. Energy offset refers to the increase or decrease in the contribution of parameter fluctuations within the same local fluctuation range of adjacent slices; For each digital twin state node, the node reference state is obtained by rearranging the state of the previous control cycle in the same area, the synchronous state of the adjacent area, and the misaligned slice of the associated parameters. The current node state is compared with the node reference state parameter by parameter, and the difference direction, the number of consecutive deviations, and the deviation magnitude are retained as state residuals. State residuals refer to the continuous deviations of the current parameter state of a digital twin state node from its reference state. According to the digital twin state node number, parameter category and control cycle order, the phase shift, energy offset and state residual are written back to the corresponding digital twin state node.

[0023] In this embodiment, S3 specifically includes: According to the digital twin state node number, parameter category and control cycle, read the phase migration amount, energy offset and state residual. For the same digital twin state node in the same control cycle, encode the phase migration direction, control cycle displacement, energy increase / decrease direction, energy offset amplitude, residual deviation direction, number of consecutive deviations and residual amplitude in a fixed field order, and write them into the regional location index, parameter category index and control cycle index to construct the regional parameter coupling token. The zone parameter coupling token refers to a feature unit that uses a single parameter in a single casting region as the object, jointly encodes the phase shift, energy offset and state residual, and retains the correspondence between the region location, parameter type and control cycle. Establish a region axis adjacency chain according to the arrangement order of the casting regions along the direction of molten metal flow, and connect the region parameter coupling tokens corresponding to the same parameter category sequentially from upstream region to downstream region, including: When the phase migration direction of the coupled tokens in adjacent regions is consistent and the control period displacement is continuous, the region connection is preserved; When the phase migration direction is opposite, the region connection is severed; When a cross-level jump occurs in the control cycle displacement, the connection will be crossed to the downstream digital twin state node corresponding to the jump cycle; Using the pre-level region parameter coupling token of each preserved region connection as the propagation starting point, the phase migration direction, energy offset amplitude and state residual are transmitted node by node along the region axis. The received characteristics of the current node and the original region parameter coupling token of the current node are rearranged according to the consistency of the change direction, the continuity of energy increase and decrease and the number of residual continuations, thus completing the forward propagation of the region axis. Establish parameter axis connection chains according to parameter categories such as electromagnetic intensity, flow rate, cooling intensity, liquid level, and temperature; Taking the downstream state residual after forward propagation along the region axis as the migration starting point, the region parameter coupling tokens in the same control cycle and the previous control cycle are retrieved along the parameter axis in reverse order from state parameter to adjustment parameter and from adjustment parameter to action parameter. The residual deviation direction is matched with the phase migration direction of the corresponding parameter step by step, and the matching result is written back along the parameter axis in reverse order to complete the reverse migration of the parameter axis. Parameter axis reverse migration refers to the token migration process that starts from the state residual of the digital twin state node and traces the residual change back to the corresponding adjustment parameter and action parameter in the reverse order of the parameter action relationship. After completing one forward propagation along the region axis and one reverse migration along the parameter axis, residual competition is performed on the region propagation component and the parameter migration component connecting the same target region parameter coupling token. The residual deviation direction, number of consecutive deviations, and residual magnitude of the two components are then aligned with the target region parameter coupling token, respectively. The component with a consistent residual direction and an increasing number of consecutive deviations is used to obtain the connection weight increment. The components whose residual direction reverses and whose residual magnitude decreases are recorded as offset weights; Components with reversed residual direction and increased residual magnitude have reduced connection weights; Residual competition refers to the process by which the region propagation component and the parameter migration component compete in direction matching, continuity comparison and amplitude around the same target state residual, and update the region parameter coupling token connection weight accordingly. The region axis adjacency chain and parameter axis connection chain are rearranged according to the updated connection weights. Connections whose weights decrease for two consecutive control cycles and whose residual contributions do not continue are deleted. Connections whose weights increase continuously are retained. Two-way binding is established for connections that simultaneously obtain region propagation increments and parameter migration increments. The two-way binding results are carried into the next control cycle to continue executing region axis forward propagation, parameter axis reverse migration and residual competition. The first dimension is established according to the regional location index, the second dimension is established according to the parameter category index, and the third dimension is established according to the control cycle index. The phase migration direction, energy offset state, state residual, regional propagation weight, parameter migration weight and bidirectional binding mark of each region parameter coupling token are written into the corresponding index position to obtain the dual-axis linkage tensor. The dual-axis linkage tensor refers to a multi-dimensional relational data structure that simultaneously stores the propagation relationship of the region parameter coupling tokens along the region axis, the migration relationship of the parameter axis, the connection weights, and the residual competition results.

[0024] In this embodiment, S4 specifically includes: Read the region parameter coupling token, region propagation weight, parameter migration weight, cancellation weight and bidirectional binding flag in the dual-axis linkage tensor, and build the original residual snapshot according to the control cycle. The original residual snapshot is obtained by arranging the current state residual of each digital twin state node according to the region location index and parameter category index. Placement ablation backtracking is performed on the dual-axis linkage tensor. The ablation order is determined from largest to smallest according to the number of bidirectional bindings of the region parameter coupling tokens, the region propagation weight, and the parameter migration weight. Placement freezing is performed on the current region parameter coupling token. The corresponding region location index, parameter category index, control cycle index, and connection topology are retained. The phase migration amount, energy offset, and state residual are replaced with the placement state obtained by rearranging the previous control cycle state of the same digital twin state node and the synchronization state of the adjacent node. The placeholder ablation pushback method sequentially performs placeholder freezing, dual-axis propagation recalculation, residual comparison and original state restoration on the regional parameter coupling token, and tracks the processing of changes in the response of the digital twin state node along the connection direction of the regional parameter coupling token. Placeholder freezing refers to the process of keeping the index position and connection relationship of the zone parameter coupling token unchanged, and replacing the current propagation feature with a placeholder state, so that the current zone parameter coupling token stops transmitting the change amount of this period to the associated connection. Starting from the region parameter coupling token that is occupied and frozen, the region propagation is replayed node by node downstream along the region axis adjacency chain. Then, the parameter transmission is replayed along the parameter axis connection chain in the order of action parameter, adjustment parameter to state parameter. The state residual of the associated digital twin state node is recalculated based on the original connection weight to obtain the ablation residual snapshot. Align the ablation residual snapshot with the original residual snapshot item by item according to the digital twin state nodes, and record the nodes where the state residual changes from positive deviation to negative deviation, from negative deviation to positive deviation, and the deviation state returns to the baseline position to obtain the residual sign migration record; The residual sign migration record refers to the continuous record of the change in the direction of the state residual of the same digital twin state node before and after the region parameter coupling token is occupied and frozen; Compare the residual amplitudes before and after the freeze, and write the nodes with continuously decreasing, continuously increasing, and first decreasing and then increasing residual amplitudes into the decay sequence, gain sequence and transition sequence respectively. Register the amplitude transition point according to the position where the direction of change of residual amplitude of adjacent nodes changes. Amplitude transition refers to the change in the amplitude of the residual state before and after the occupancy freeze between adjacent propagation nodes, from increasing to decreasing or from decreasing to increasing. After completing the residual sign migration record and amplitude jump point registration of the current region parameter coupling token, restore the original features of the current region parameter coupling token and freeze the region parameter coupling token corresponding to the next ablation sequence. Then, traverse the region parameter coupling tokens in the dual-axis linkage tensor in turn to obtain the ablation response trajectory corresponding to each region parameter coupling token. Cross-matching of each ablation response trajectory is performed. The region parameter coupling tokens that show a continuous decrease in the residual amplitude of downstream nodes after the occupancy freeze and whose residual signs migrate towards the reference position are marked as master control tokens. The master control chain is then screened by connecting consecutive master control tokens according to the order of region propagation connection, parameter migration connection, and control cycle. The master chain refers to the domain parameter coupled token connection path consisting of consecutive master tokens, and whose residuals of associated digital twin state nodes continue to decay after being frozen. After the placeholder is frozen, the residual magnitude of the associated node increases continuously, the original residual sign is maintained, or the region parameter coupling token is marked as the offset token. Then, the continuous offset tokens are connected according to the offset weight and the two-way binding relationship to filter the offset chain. An offset chain refers to a region-parameter coupled token connection path consisting of consecutive offset tokens, which causes the residual of the associated digital twin state node to continuously increase after the placeholder is frozen. Write the corresponding regional parameter coupling token number, connection direction, control cycle position, residual symbol migration record, and amplitude transition point of the main control chain and the offset chain into the dual-axis linkage tensor.

[0025] In this embodiment, S5 specifically includes: Read the master control chain, cancellation chain, regional parameter coupling token, connection weight, cancellation weight, residual sign migration record and amplitude jump point in the dual-axis linkage tensor, establish link control slots according to regional location, parameter category and control period, and write the link control slots corresponding to the master control chain into the forward policy layer and the link control slots corresponding to the cancellation chain into the reverse policy layer to construct a positive and negative dual policy field. A link control slot refers to a data location that corresponds to a single area, a single control parameter, and a single control cycle, and is used to record the control direction, control increment, link weight, and transmission order. A dual-strategy field refers to a control relationship structure formed by superimposing the forward strategy layer and the reverse strategy layer according to the same region location index and parameter category index; For the main control chain, the first-level control direction is determined by the residual deviation direction corresponding to the coupling token in the starting area of ​​the main control chain. The first-level control increment is determined according to the connection weight, residual amplitude and amplitude transition position, and is sequentially passed to the next link control slot along the region propagation connection and parameter migration connection of the main control chain. During the step-by-step transmission of control increments, the control direction received by the current link control slot is matched with the residual deviation direction of the corresponding area parameter coupling token; When the directions are the same and the residual magnitude increases continuously along the link, the control direction is retained and the current control increment is increased; When the directions are the same and the residual magnitude decreases continuously along the link, the control direction is retained and the current control increment is compressed. When encountering an amplitude transition point, freeze the previous stage increment and redetermine the control increment of the next link control slot based on the residual change amplitude after the transition point, and complete the recursive allocation of the same-direction control increment along the main control chain. In unidirectional control increment refers to the control change that is consistent with the direction of the correction of the state residual in the main control chain and is redistributed step by step according to the propagation order of the link. For the offset chain, a reverse control increment is generated based on the residual deviation direction and offset weight corresponding to the offset token. The main control chain link control slot that intersects with the same region, the same parameter category, or the same control cycle is retrieved according to the two-way binding relationship of the offset chain. The reverse control increment and the corresponding same-direction control increment are written to the same competitive position. The competition order is determined by the connection weight of the main control chain and the offset weight of the offset chain; First, reduce the reverse control increment that is opposite in direction to the same-direction control increment and has a larger offsetting weight; Further reduce the inverse control increment that causes an increase in residual amplitude after cross-regional propagation; Truncate the reverse control increment that conflicts with the control direction of the main control chain for two consecutive control cycles. The remaining reverse control increments after reduction are redistributed according to the subsequent links of the offset chain; Complete the reverse control increment along the offset chain competitive reduction; Competition reduction refers to the process of offsetting, truncating, and redistributing the directional conflict control increments generated by the master control chain and the offset chain within the same link control slot, according to the connection weight, offset weight, and residual change status. After completing the main control chain recursive allocation and the elimination chain competition reduction, the control cycle of the link control slots belonging to electromagnetic strength, molten metal flow, cooling intensity and casting speed in the same area is aligned. The remaining control increments with the same direction are merged according to the link propagation order. The remaining control increments with opposite directions are subjected to local competition again. The control increments that exceed the continuous range of adjacent control cycle changes are migrated to the next control cycle to obtain the regional candidate control vector. By using distributed control, candidate control vectors from each region are cross-validated between adjacent regions according to the connection order of the main control chain. Control components that cause the downstream main control chain residuals to continue to converge after the current region's control increment are retained, while control components that cause the downstream offset chain residuals to increase are pulled back. The pullback amount is then redistributed to the control slots of the links with the second highest connection weight in the same main control chain. After completing the cross-validation of each region, the final control increments corresponding to electromagnetic intensity, molten metal flow rate, cooling intensity and casting speed are collected in sequence according to the region location index. The control direction, execution order, main control chain source and offset chain reduction mark are recorded simultaneously to calculate the multi-region control vector. Multi-zone control vectors refer to a joint control data structure that organizes and stores the final control increment, control direction, execution order, and link source of each zone according to the order of casting zones.

[0026] In this embodiment, S6 specifically includes: The industrial control system adjusts the electromagnetic intensity, molten metal flow rate, cooling intensity, and casting speed of each casting zone according to the zone position, control direction, and execution sequence in the multi-zone control vector, and collects the adjusted parameter responses through a sensor network. The parameter difference before and after adjustment, response direction and response duration are written back according to the digital twin state node number. Response writing back refers to updating the actual control results to the corresponding digital twin state node according to the area location and parameter category. Based on the updated digital twin state nodes, the state residuals are recalculated. The connection weights corresponding to the continuous decay of the state residuals are increased, and the connection weights corresponding to the continuous enhancement of the state residuals are decreased. The connection record strategy is reversed for the connection with the reversed residual direction. Based on the updated connection weights and policy reversal flags, the link control slots in the positive and negative dual-policy fields are redistributed, the same-direction control increments and reverse control increments are corrected, and the correction results are used for the calculation of the multi-zone control vector in the next control cycle.

[0027] An electromagnetic casting collaborative control system based on multi-zone parameter linkage includes: The data acquisition module is used to collect temperature, liquid level, flow rate, electromagnetic intensity and cooling intensity of each casting area through a sensor network, and to perform same-period sampling alignment and frame loss recursive compensation through narrowband Internet of Things to construct a multi-zone parameter cube; The state mapping module is used to establish digital twin state nodes based on the location of the casting area and the corresponding relationship of the equipment. It performs staggered slicing of the multi-zone parameter cube according to the region axis and parameter axis, and calculates the phase shift, energy offset and state residual of adjacent slices. The dual-axis linkage module is used to construct a region parameter coupling token based on the phase shift, energy offset and state residual, alternately execute the region axis forward propagation and parameter axis reverse migration, and update the token connection weight through residual competition to obtain the dual-axis linkage tensor; The link filtering module is used to perform placeholder ablation backtracking on the dual-axis linkage tensor, freeze the region-coupled tokens one by one and recalculate the corresponding digital twin state nodes, and filter the master chain and the offset chain based on the sign migration and amplitude transition of the node residuals. The collaborative decision-making module is used to construct a positive and negative dual-strategy field based on the main control chain and the offset chain, recursively allocate the same-direction control increment along the main control chain, competitively reduce the opposite-direction control increment along the offset chain, and calculate the multi-zone control vector through distributed control. The feedback control module is used to adjust the electromagnetic intensity, molten metal flow rate, cooling intensity and casting speed according to the multi-zone control vector of the industrial control system, write the execution response back to the digital twin state node, and correct the connection weight and dual-strategy field parameters according to the updated state residual.

[0028] Example 1: To verify the feasibility of this invention in practice, it was applied to a continuous electromagnetic casting production scenario. This scenario employs a multi-zone continuous casting structure, divided along the direction of molten metal flow and solidification into a supply zone, an electromagnetic flow stabilization zone, a solidification transition zone, and a cooling and forming zone. Each zone is equipped with temperature sensors, level sensors, flow sensors, electromagnetic intensity detection channels, and cooling intensity detection channels. A data transmission link is established with the industrial control system via a narrowband Internet of Things (IoT). During implementation, the molten metal temperature mainly ranges from 683.6℃ to 714.8℃, the level fluctuation range is ±6.8mm, the molten metal flow rate is maintained at 18.2–21.6 L / min, the electromagnetic intensity adjustment range is 23.5–31.8 mT, the cooling intensity equivalent to water flow rate is 34.6–43.2 L / min, and the casting speed is controlled at 72–86 mm / min. The original on-site control method mainly adjusts the electromagnetic intensity, molten metal flow rate, and cooling intensity based on the local parameter deviations of each area. When the flow rate in the supply area fluctuates, the liquid level in the electromagnetic stabilization zone usually responds after 2 to 4 control cycles. Subsequently, the temperature in the solidification transition zone and the cooling forming zone continue to change with a delay. This easily leads to control conflicts where the flow rate in one area is decreasing while the electromagnetic intensity in the next area is increasing. As a result, the peak liquid level remains in the range of 8.4 to 11.7 mm for a long time, accompanied by local temperature oscillations and frequent corrections to the casting speed.

[0029] In implementing this invention, parameters of each region are continuously collected via a sensor network, with each control cycle lasting 0.5 seconds. Each sampling channel is written with a region number, parameter category, sampling sequence number, and local timestamp. During continuous operation, a total of 21,600 sets of valid control cycle data are obtained, of which 426 sets of sampling offsets caused by wireless transmission delays and 137 sets of single-channel missing frames are detected. The narrowband IoT performs same-cycle sampling alignment according to the unified control cycle of the industrial control system. For sampling channels with consistent offset directions, the sampling position is progressively shifted. For channels where offset reversal occurs, the sampling position is redefined based on the local change slope before and after the reversal point. For the 137 sets of missing frame data, forward change trajectories and backward constraint trajectories are established based on the change direction, amplitude, and speed before and after the missing position, respectively. Missing frame recursive compensation is then completed by combining the synchronous change direction of similar parameters in adjacent regions. After compensation, each region, parameter, and continuous control cycle are mapped to the region axis, parameter axis, and time axis, respectively, constructing a multi-region parameter cube. After data verification, the average time offset of effective data in different regions after sampling and alignment in the same period was reduced from 148ms to 21ms, and the average relative deviation of parameters after frame loss compensation was controlled at 1.63%, providing a continuous data foundation for subsequent regional linkage calculations.

[0030] Based on a multi-zone parameter cube, digital twin state nodes are established according to the location of each casting zone and the correspondence between sensors, electromagnetic coils, liquid supply actuators, and cooling actuators. Using the liquid supply zone, electromagnetic stabilization zone, solidification transition zone, and cooling forming zone as zone location indices, misaligned slicing of the execution zone and parameter misalignment is performed on continuous parameter windows. During a typical flow disturbance, the flow rate in the liquid supply zone decreases from 20.4 L / min to 18.9 L / min over 3.0 s. The algorithm first detects a downward migration anchor point in the liquid supply zone flow rate parameter, then detects a corresponding downward migration anchor point in the electromagnetic stabilization zone level parameter 1.0 s later, and finally a downward migration anchor point appears in the solidification transition zone temperature parameter 2.0 s later. After calculating the phase migration amount through the control cycle index difference, change direction, and duration period in adjacent misaligned slices, the migration span from the liquid supply flow rate to the stabilization zone level is found to be 2 control cycles, and the migration span from the stabilization zone level to the solidification zone temperature is also found to be 2 control cycles. After simultaneously calculating the energy offset of the local fluctuation range, it was found that the fluctuation contribution increased by about 17.6% when the flow fluctuation was transmitted to the steady flow zone, and decreased to 11.2% when it was transmitted to the solidification transition zone. This indicates that the disturbance does not act on the entire region at the same time, but migrates gradually along the regional direction.

[0031] Subsequently, the phase shift, energy offset, and state residual are encoded into region-parameter coupling tokens according to the region location index, parameter category index, and control cycle index. The algorithm performs forward propagation along the region axis, passing region-parameter coupling tokens of the same parameter category in the order from the liquid supply zone to the cooling forming zone; simultaneously, starting from the downstream state residual, it performs reverse propagation along the parameter axis according to the inverse relationship between state parameters, adjustment parameters, and action parameters. After residual competition, the connection weight of the link from liquid supply flow rate to steady flow zone level to solidification zone temperature increases for three consecutive control cycles, from the initial 0.48 to 0.73, while the connection weight of cooling intensity directly affecting the steady flow zone level decreases from 0.41 to 0.19 and is pruned. The dual-axis linkage tensor thus retains parameter links with continuous residual contributions, reducing control associations that are locally correlated but lack continuous propagation relationships.

[0032] The dual-axis linkage tensor is further subjected to placeholder ablation pushback. While maintaining the token index and connection topology unchanged, the region-coupling tokens are frozen sequentially, and placeholder states are generated based on the node's previous control cycle state and the synchronization state of adjacent nodes, subsequently propagating again along the region axis and parameter axis. After freezing the region-coupling tokens corresponding to the liquid supply flow rate, the residual state of the downstream steady flow zone liquid level decreases from 4.7 mm to 1.9 mm, and the residual state of the solidification transition zone temperature decreases from 6.3℃ to 2.4℃; after freezing the region-coupling tokens corresponding to the electromagnetic intensity of the steady flow zone, the downstream liquid level residual increases from 3.8 mm to 5.1 mm. Based on the residual sign shift and amplitude transition results, the continuous tokens corresponding to the liquid supply flow rate—steady flow zone liquid level—solidification zone temperature are selected as the main control chain, and the electromagnetic intensity branch that produces opposite adjustments to the main control chain and causes an increase in downstream residuals is selected as the cancellation chain.

[0033] After determining the main control chain and the offset chain, a link control slot is established based on the region location, parameter category, and control cycle, and a dual-strategy field is constructed. During a level rise disturbance, the liquid level in the steady flow zone increases by 5.6 mm from the baseline value. The forward strategy layer first allocates a control increment of -0.72 L / min to the liquid supply flow rate and allocates electromagnetic intensity to -1.6 mT and cooling intensity to +1.9 L / min along the main control chain. The reverse strategy layer detects that the original local control logic generates an electromagnetic intensity increment of +1.1 mT, which conflicts with the control direction of the main control chain. It then performs competitive reduction, reducing the reverse increment to +0.2 mT and canceling it in the next control cycle. Distributed control further performs cross-validation on candidate control vectors in adjacent regions, retaining control components that converge the downstream state residuals and withdrawing control components that enhance the offset chain residuals, thus obtaining the final multi-region control vector.

[0034] The industrial control system synchronously adjusts electromagnetic intensity, molten metal flow rate, cooling intensity, and casting speed based on multi-zone control vectors, and writes the actual parameter responses back to the digital twin state node. Over 21,600 consecutive control cycles, the state residual is updated once per control cycle, connection weights are increased or decreased according to the residual convergence direction, and the positive and negative dual-strategy fields are synchronously corrected according to changes in the main control chain and the offset chain. After reaching a stable state, the standard deviation of the flow rate in the supply zone decreased from 0.74 L / min to 0.31 L / min, the peak-to-peak value of the liquid level in the steady flow zone decreased from 10.6 mm to 4.3 mm, the standard deviation of the temperature in the solidification transition zone decreased from 4.8℃ to 2.1℃, and the average number of electromagnetic intensity adjustments per cycle decreased from 16.4 times per 100 control cycles to 8.7 times, indicating a significant reduction in unnecessary repeated adjustments. The average casting speed increased from 77.6 mm / min to 81.2 mm / min, and while the speed increased, the temperature fluctuation in the cooling and forming zone did not increase synchronously.

[0035] To further verify the control effect, the control method of this invention was compared with the conventional control method using single-region independent feedback under the same parameter range, as shown in Table 1. Thirty groups of continuously stable casting batches were statistically analyzed for both control methods, with 720 control cycles recorded for each group. The conventional control method required an average of 7.8 control cycles to restore the liquid level to within ±2.0 mm after multi-region disturbances, while the present invention required an average of 4.1 control cycles. The regional control conflict rate of the conventional control method was 12.6%, which decreased to 3.4% for the present invention. The peak-to-peak temperature of the solidification transition zone decreased from 13.2℃ to 6.7℃. The number of repeated adjustments caused by control direction reversal per unit time decreased from an average of 11.3 times to 4.6 times. During quality sampling of the formed billet, the surface ripple abnormality rate decreased from 4.8% to 1.9%, and the local edge dimension deviation rate decreased from 3.6% to 1.4%, indicating that multi-region parameter linkage control not only improved the control process itself but also significantly improved the consistency of casting.

[0036] Table 1 Comparison of the Implementation Effects of Multi-Zone Parameter Linkage Control As can be seen from the above implementation data, the present invention improves the consistency of multi-region data through same-period sampling alignment and missing frame recursive compensation. It identifies the temporal propagation relationship between different casting regions through misaligned slicing, phase shift, energy offset and state residual. It filters the main control chain and the cancellation chain through region parameter coupling token, dual-axis linkage tensor and occupancy ablation back push. Then, it coordinates the control increment of multiple regions through positive and negative dual strategy fields and distributed control, so that the electromagnetic intensity, molten metal flow rate, cooling intensity and casting speed are transformed from independent adjustment to coordinated adjustment based on regional propagation relationship, thereby reducing control conflict and parameter oscillation, and improving the stability of casting process, control accuracy and the consistency of forming quality.

[0037] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A collaborative control method for electromagnetic casting based on multi-zone parameter linkage, characterized in that, Includes the following steps: S1. Collect temperature, liquid level, flow rate, electromagnetic intensity and cooling intensity of each casting area through sensor network, and perform same-period sampling alignment and frame loss recursive compensation through narrowband Internet of Things to construct a multi-zone parameter cube; S2. Based on the location of the casting area and the corresponding relationship of the equipment, establish digital twin state nodes, perform misaligned slicing of the multi-zone parameter cube according to the region axis and parameter axis, and calculate the phase shift, energy offset and state residual of adjacent slices. S3. Construct a region parameter coupling token based on the phase shift, energy offset and state residual, alternately perform forward propagation of the region axis and reverse migration of the parameter axis, update the token connection weight through residual competition, and obtain the dual-axis linkage tensor. S4. Perform placeholder ablation backtracking on the dual-axis linkage tensor, freeze the region-coupled tokens one by one and recalculate the corresponding digital twin state nodes, and select the master chain and the cancellation chain based on the sign shift and amplitude jump of the node residuals. S5. Construct a positive and negative dual-strategy field based on the main control chain and the offset chain. The control increment in the same direction is recursively allocated along the main control chain, and the control increment in the opposite direction is competitively reduced along the offset chain. Multi-zone control vectors are calculated through distributed control. S6. The industrial control system adjusts the electromagnetic intensity, molten metal flow rate, cooling intensity and casting speed according to the multi-zone control vector, and writes the execution response back to the digital twin state node, and corrects the connection weight and dual-strategy field parameters according to the updated state residual.

2. The electromagnetic casting collaborative control method based on multi-zone parameter linkage according to claim 1, characterized in that, S1 specifically includes: The sensor network collects the temperature, liquid level, flow rate, electromagnetic intensity and cooling intensity of each casting area, writes the area number, parameter category, sampling sequence number and local timestamp to each sampling channel, and establishes a global sampling benchmark based on the industrial control system control cycle. Narrowband IoT calculates the timestamp offset of adjacent control cycles according to the sampling sequence number. For channels with the same offset direction for three consecutive sampling cycles, the sampling position is progressively shifted. For channels with reversed offset direction, the sampling position is repositioned according to the local change slope before and after the reversal point to complete the sampling alignment in the same cycle. For channels with discontinuous sampling numbers, forward change trajectories and backward constraint trajectories are established respectively, and bidirectional recursion is performed based on the change direction, change amplitude and change speed before and after the missing position; At the missing position, the forward change trajectory and the backward constraint trajectory compete point by point, retain the trajectory segments with the same direction, and correct the directional conflict position according to the synchronous change direction of the same parameters in the adjacent area to complete the frame missing recursive compensation. Establish region axis, parameter axis and time axis according to casting area, parameter type and continuous control cycle respectively. Write the sampled values ​​that have completed the same period sampling alignment and missing frame recursive compensation into the corresponding positions, and record the alignment displacement mark and compensation source mark to construct a multi-region parameter cube.

3. The electromagnetic casting collaborative control method based on multi-zone parameter linkage according to claim 1, characterized in that, S2 specifically includes: A digital twin state node is established based on the arrangement of the casting area, the installation position of the sensor, and the corresponding relationship of the actuator. The parameter values ​​and data processing markers are written into the multi-zone parameter cube according to the continuous control cycle. Digital twin status nodes refer to data nodes that establish a fixed mapping with the actual casting area and corresponding equipment, and are used to record the status of area parameters; Extract continuous parameter windows along the region axis, and shift the parameter windows of adjacent casting regions forward or backward by one control cycle along the time axis to obtain a region misalignment slice; The parameter window is cross-transformed along the parameter axis according to the interaction between electromagnetic intensity, temperature, liquid level, flow rate and cooling intensity to obtain parameter misalignment slices. Region misalignment slices and parameter misalignment slices are collectively referred to as misalignment slices. In the misaligned slice, the locations of reversed change direction and abrupt change in local amplitude are extracted as migration anchors. Forward pairing and reverse verification are performed based on the original control cycle index, and migration anchors with consistent bidirectional matching are retained. The phase migration amount is obtained based on the control period displacement, change direction and duration period of the migration anchor point in the adjacent misaligned slice, and the energy offset is obtained based on the difference in local fluctuation contribution between adjacent migration anchor points. The node reference state is obtained by rearranging the previous control cycle state of the same area, the synchronous state of adjacent areas, and the misaligned slices of associated parameters. The current node state is compared with the node reference state parameter by parameter to obtain the state residual including the deviation direction, the number of consecutive deviations, and the deviation magnitude. According to the digital twin state node number, parameter category and control cycle, the phase shift, energy offset and state residual are written back to the corresponding digital twin state node.

4. The electromagnetic casting collaborative control method based on multi-zone parameter linkage according to claim 1, characterized in that, S3 specifically includes: Based on the digital twin state node number, parameter category, and control cycle, the phase shift, energy offset, and state residual are jointly encoded and written into the regional location index, parameter category index, and control cycle index to construct a regional parameter coupling token. The zone parameter coupling token refers to a feature unit that uses a single parameter in a single casting region as the object, jointly encodes the phase shift, energy offset and state residual, and retains the correspondence between the region location, parameter type and control cycle. Establish a regional axis adjacency chain along the direction of molten metal flow, connect regional parameter coupling tokens of the same parameter category in sequence, and perform connection retention, disconnection and cross-level jump according to the phase migration direction and control cycle displacement; Starting with the pre-connected pre-connected region parameter coupling token, the phase migration direction, energy offset amplitude, and state residual are transmitted node by node along the region axis, and the forward propagation of the region axis is completed based on the consistency of direction, energy continuity, and number of residual continuations. Establish a parameter axis connection chain based on state parameters, adjustment parameters, and action parameters. Starting from the state residual after the forward propagation of the region axis, match the region parameter coupling tokens in the reverse order from state parameters to adjustment parameters and from adjustment parameters to action parameters to complete the reverse migration of the parameter axis. Perform residual competition on the region propagation component and parameter migration component of the coupling tokens that connect to the same target area, update the connection weights according to the residual direction, number of consecutive deviations and residual magnitude, and record the cancellation weights; The region axis adjacency chain and parameter axis connection chain are rearranged according to the updated connection weights. Continuously decaying connections are deleted, continuously gaining connections are retained, and bidirectional binding is established for connections that gain simultaneously during region propagation and parameter migration. Based on the region location index, parameter category index, and control cycle index, write the region parameter coupling token, connection weight, cancellation weight, and bidirectional binding flag to the corresponding positions to obtain the dual-axis linkage tensor.

5. The electromagnetic casting collaborative control method based on multi-zone parameter linkage according to claim 1, characterized in that, S4 specifically includes: Read the region parameter coupling token, connection weight and bidirectional binding mark in the dual-axis linkage tensor, and build the original residual snapshot according to the region location index and parameter category index of the digital twin state node; The placeholder ablation pushback is performed on the dual-axis linkage tensor. The ablation order is determined according to the number of bidirectional bindings and the connection weight. Placeholder freezing is performed on the regional parameter coupling tokens in sequence, the token index and connection topology are retained, and the placeholder state is obtained by rearranging the state of the corresponding node in the previous control cycle and the synchronization state of the adjacent nodes. The placeholder ablation pushback method sequentially performs placeholder freezing, dual-axis propagation recalculation, residual comparison and original state restoration on the regional parameter coupling token, and tracks the processing of changes in the response of the digital twin state node along the connection direction of the regional parameter coupling token. Starting from the placeholder frozen region parameter coupling token, the process is re-propagated along the region axis adjacency chain and parameter axis connection chain. The state residuals of the associated digital twin state nodes are recalculated based on the original connection weights to obtain the ablation residual snapshot. Align the ablation residual snapshot with the original residual snapshot node by node, record the changes in the direction of the state residual and the increase or decrease of the residual amplitude, and register the position where the direction of the residual amplitude change of adjacent nodes switches as the amplitude transition point. Restore the current region parameter coupling token and continue freezing the next token according to the ablation order. After traversing the dual-axis linkage tensor, obtain the ablation response trajectory corresponding to each region parameter coupling token. Cross-matching is performed on the ablation response trajectory, and the region parameter coupling tokens that have continuously decayed residuals of associated nodes after freezing and whose residual direction migrates towards the reference position are connected to the main control chain. Connect the region parameter coupling tokens that continuously enhance the residuals of the associated nodes after freezing and whose residual directions continuously deviate from the reference position into an offset chain, and write the main control chain, the offset chain and the corresponding amplitude transition point back to the dual-axis linkage tensor.

6. The electromagnetic casting collaborative control method based on multi-zone parameter linkage according to claim 1, characterized in that, S5 specifically includes: Read the main control chain, cancellation chain, connection weight, cancellation weight and amplitude transition point in the dual-axis linkage tensor, establish link control slots according to region location, parameter type and control period, and write them into the forward policy layer and reverse policy layer respectively to construct a positive and negative dual policy field; A link control slot refers to a data location that corresponds to a single area, a single control parameter, and a single control cycle, and is used to record the control direction, control increment, link weight, and transmission order. A dual-strategy field refers to a control relationship structure formed by superimposing the forward strategy layer and the reverse strategy layer according to the same region location index and parameter category index; The control direction is determined by the residual deviation direction of the coupled token in the starting area of ​​the main control chain. The first-level control increment is determined according to the connection weight, residual amplitude and amplitude transition point, and then passed down level by level along the main control chain. During the transmission of the main control chain, the control increment is adjusted according to the continuous increase or decrease of the residual amplitude. At the amplitude transition point, the previous increment is frozen and the subsequent control increment is redefined to complete the recursive allocation of the same-direction control increment. The reverse control increment is generated using the residual deviation direction and offset weight corresponding to the offset chain. The reverse control increment and the same-direction control increment where the link intersects are written into the same link control slot to perform contention reduction. Based on the connection weight, offset weight, and residual change status, the reverse control increment is offset, truncated, and redistributed, and the remaining reverse control increment continues to be passed along the offset chain. For the link control slots corresponding to electromagnetic intensity, molten metal flow rate, cooling intensity and casting speed in the same area, control cycle alignment is performed, and in-direction control increments are merged and local competition is performed on in-reverse control increments to obtain regional candidate control vectors. By performing cross-validation on candidate control vectors in adjacent regions according to the main control chain sequence through distributed control, the control components that enable downstream residual convergence are retained, and the control components that enhance the residual of the offset chain are withdrawn, thus obtaining multi-region control vectors.

7. A multi-zone parameter linkage-based electromagnetic casting collaborative control system, executing the multi-zone parameter linkage-based electromagnetic casting collaborative control method according to any one of claims 1 to 6, characterized in that, include: The data acquisition module is used to collect temperature, liquid level, flow rate, electromagnetic intensity and cooling intensity of each casting area through a sensor network, and to perform same-period sampling alignment and frame loss recursive compensation through narrowband Internet of Things to construct a multi-zone parameter cube; The state mapping module is used to establish digital twin state nodes based on the location of the casting area and the corresponding relationship of the equipment. It performs staggered slicing of the multi-zone parameter cube according to the region axis and parameter axis, and calculates the phase shift, energy offset and state residual of adjacent slices. The dual-axis linkage module is used to construct a region parameter coupling token based on the phase shift, energy offset and state residual, alternately execute the region axis forward propagation and parameter axis reverse migration, and update the token connection weight through residual competition to obtain the dual-axis linkage tensor; The link filtering module is used to perform placeholder ablation backtracking on the dual-axis linkage tensor, freeze the region-coupled tokens one by one and recalculate the corresponding digital twin state nodes, and filter the master chain and the offset chain based on the sign migration and amplitude transition of the node residuals. The collaborative decision-making module is used to construct a positive and negative dual-strategy field based on the main control chain and the offset chain, recursively allocate the same-direction control increment along the main control chain, competitively reduce the opposite-direction control increment along the offset chain, and calculate the multi-zone control vector through distributed control. The feedback control module is used to adjust the electromagnetic intensity, molten metal flow rate, cooling intensity and casting speed according to the multi-zone control vector of the industrial control system, write the execution response back to the digital twin state node, and correct the connection weight and dual-strategy field parameters according to the updated state residual.

Images