Intermediate frequency furnace operation control system based on adaptive algorithm

Abstract

The present application relates to the technical field of industrial automation and intelligent control of metallurgical equipment, in particular to a medium frequency furnace operation control system based on an adaptive algorithm, comprising: a data acquisition module for generating real-time operation data; a benchmark reconstruction module for reconstructing an ideal operation benchmark based on a preset multi-physical field mechanism model and the real-time operation data; an injection simulation module for generating a theoretical abnormal trajectory; a residual extraction module for generating a real residual and a theoretical residual based on the theoretical abnormal trajectory and the ideal operation benchmark; a decision control module for determining a state evaluation result and generating an adaptive control instruction; and a control execution module for receiving the adaptive control instruction and controlling the output power, output frequency and / or working mode of the medium frequency power supply to adjust the operation state of the medium frequency furnace coupled load loop; the present application solves the blind power increasing behavior under the condition of incomplete information and realizes the closed-loop control with safety as the priority.

Description

Technical Field

[0001] This invention relates to the field of industrial automation and intelligent control technology for metallurgical equipment, specifically to a medium-frequency furnace operation control system based on an adaptive algorithm. Background Technology

[0002] In the current medium-frequency furnace smelting operation environment, the equipment will continuously generate electrical parameters and heat exchange status data of the cooling circuit during the complete cycle of charging, melting and heat preservation. As the furnace charge gradually collapses, softens and melts from a blocky solid, the equivalent magnetic conductivity and heat distribution continue to change. The above operating data are often accompanied by significant phase change disturbances and fluctuations.

[0003] To control the operating status of intermediate frequency furnaces, existing solutions generally rely on a single-point threshold at a certain moment or directly use historical averages as a reference for passive adjustment. Although this solution can achieve basic melting efficiency adjustment, it fails to construct a dynamic healthy operating benchmark by combining a multi-physics mechanism model and lacks comparison of the evolution time sequence of multi-dimensional characteristics. This makes it difficult to effectively distinguish normal phase transformation disturbances from dangerous fault signs such as furnace lining erosion and charge bridging. This passive approach, which only follows the surface changes of parameters, has a high probability of erroneous adjustment under incomplete information or complex operating conditions, which leads to safety risks such as furnace penetration, local overheating, and abnormal equipment operating status. It is difficult to support adaptive control that balances safety and stability.

[0004] Therefore, how to accurately identify the source of abnormal deviations based on multi-physics field mechanisms and improve the reliability of intermediate frequency furnace operation status assessment and directional adjustment has become an urgent technical problem to be solved. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a medium-frequency furnace operation control system based on an adaptive algorithm. Specifically, the technical solution of this invention includes:

[0006] The data acquisition module is used to collect the timing data of electrical parameters, including induction coil voltage, current, frequency and power, and heat exchange status data of the cooling circuit of the medium frequency furnace coupled load circuit, and generate real-time operating data.

[0007] The benchmark reconstruction module is used to reconstruct the ideal operating benchmark based on the preset multiphysics mechanism model and real-time operating data. The multiphysics mechanism model includes at least an electromagnetic coupling sub-model and a heat exchange sub-model. The injection simulation module is used to inject preset abnormal mechanism parameters into the ideal operating benchmark to generate theoretical abnormal trajectories.

[0008] The residual extraction module is used to generate real residuals based on real-time operating data and ideal operating benchmarks, and to generate theoretical residuals based on theoretical abnormal trajectories and ideal operating benchmarks; the decision control module is used to determine the state evaluation result based on the trajectory similarity between the real residuals and the theoretical residuals, and to generate adaptive control commands.

[0009] The control execution module, whose signal output terminal is connected to the intermediate frequency power supply controller of the intermediate frequency furnace, is used to receive adaptive control commands and control the output power, output frequency and / or operating mode of the intermediate frequency power supply to adjust the operating status of the intermediate frequency furnace coupled load circuit.

[0010] Optionally, the data acquisition module includes: an electrical parameter acquisition unit for acquiring induction coil voltage, induction coil current, operating frequency, active power, and reactive power; a thermal status acquisition unit for acquiring the temperature difference between the inlet and outlet mediums of the cooling circuit and the circulation flow rate; and a timing synchronization unit for time-aligning the induction coil voltage, induction coil current, operating frequency, active power, reactive power, temperature difference between the inlet and outlet mediums, and circulation flow rate to generate real-time operating data.

[0011] Optionally, the benchmark reconstruction module includes: an electromagnetic modeling unit for calculating the ideal impedance trajectory based on a preset multiphysics mechanism model and real-time operating data; a thermal modeling unit for calculating the ideal temperature rise trajectory based on a preset multiphysics mechanism model and real-time operating data; and a benchmark fusion unit for calculating the ideal frequency drift trajectory by coupling the ideal impedance trajectory and the ideal temperature rise trajectory, and generating an ideal operating benchmark based on the ideal impedance trajectory, the ideal temperature rise trajectory, and the ideal frequency drift trajectory.

[0012] Optionally, the injection simulation module includes: an erosion injection unit, used to inject a preset coupling attenuation parameter characterizing the degree of furnace lining thickness attenuation into an ideal operating reference to generate a theoretical trajectory of furnace lining erosion; a bridging injection unit, used to inject a preset local thermal resistance mutation parameter and a preset equivalent inductance step parameter into the ideal operating reference to generate a theoretical trajectory of furnace charge bridging; and a template generation unit, used to generate a theoretical abnormal trajectory based on the theoretical trajectory of furnace lining erosion and the theoretical trajectory of furnace charge bridging.

[0013] Optionally, the template generation unit is also used to: classify and store theoretical abnormal trajectories and / or theoretical residuals obtained therefrom based on preset disturbance types, so as to form a theoretical abnormal trajectory template library and / or a theoretical residual template library; wherein, the preset disturbance types include normal phase transformation disturbance, charging disturbance, furnace lining erosion disturbance and furnace charge bridging disturbance.

[0014] Optionally, the residual extraction module includes: a real-time difference unit, used to perform difference calculations between real-time operating data and an ideal operating benchmark to generate real-time residuals; a theoretical difference unit, used to perform difference calculations between theoretical abnormal trajectories and an ideal operating benchmark to generate theoretical residuals; and a normalization unit, used to normalize the magnitude and time scale of the real-time residuals and theoretical residuals; wherein the real-time residuals and theoretical residuals are multi-dimensional time-series residual sequences constructed according to predetermined feature dimensions within a preset time sliding window, and the predetermined feature dimensions include at least impedance deviation, temperature rise deviation, and frequency drift deviation.

[0015] Optionally, the decision control module includes: a similarity calculation unit, used to calculate the trajectory similarity between the actual residual and the theoretical residual based on the dynamic time warping algorithm; and a state decision unit, used to output anomaly labels and confidence levels based on the trajectory similarity. The state decision unit is configured with a first threshold and a second threshold. The first threshold is calibrated based on the equipment's safety limit deviation, and the second threshold is calibrated based on the disturbance standard deviation of historical normal operation data. The first threshold is greater than the second threshold. When the trajectory similarity is greater than or equal to the first threshold, an anomaly confirmation result is output; when the trajectory similarity is less than or equal to the second threshold, a normal operation result is output; and when the trajectory similarity is greater than the second threshold and less than the first threshold, a result to be verified is output.

[0016] Optionally, the decision control module further includes: a strategy switching unit, used to output a control command to reduce the output power and periodically adjust the operating frequency when an abnormal confirmation result is output and the abnormal label is furnace charge bridging; a conservative control unit, used to output a control command to maintain the current power and output a sampling control command to increase the sampling frequency and / or shorten the sampling period when a result to be verified is output; and an efficiency control unit, used to output a constant power adjustment command or a power factor optimization adjustment command when a normal operation result is output.

[0017] Optionally, the medium-frequency furnace coupled load circuit is a coupled circuit consisting of the medium-frequency furnace induction coil and the melt inside the furnace, the heat exchange status data is the status data of the medium-frequency furnace cooling water circulation system, and the status assessment results include the furnace lining erosion status and the charge bridging status.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] 1. This system dynamically reconstructs ideal impedance, temperature rise, and frequency drift trajectory based on a multiphysics mechanism model that includes electromagnetic coupling and heat exchange, and real-time operating data. This mechanism replaces the traditional single-point threshold or historical average reference, and can accurately reflect the evolution trend that the equipment should exhibit under the current process stage health condition. It provides a benchmark with clear physical meaning for anomaly identification and effectively avoids misjudging normal smelting fluctuations as faults.

[0020] 2. This system actively injects preset abnormal mechanism parameters characterizing furnace lining erosion or charge bridging into an ideal benchmark to generate theoretical abnormal trajectories, and calculates the trajectory similarity between the actual residuals and the theoretical residuals based on a dynamic time warping algorithm. This approach breaks through the limitations of passive response based solely on surface mutations of parameters, and realizes mechanism-driven evolutionary time-series comparison, which can accurately distinguish between conventional charging disturbances and real furnace penetration or local overheating hazards, greatly improving the accuracy of condition assessment.

[0021] 3. This system normalizes the amplitude and time scale of the multi-dimensional time-series real residuals and theoretical residuals, which include impedance deviation, temperature rise deviation and frequency drift deviation. This design transforms the original multi-source absolute values ​​into comparable deviation feature sequences across furnace batches and process stages, effectively eliminating the interference caused by differences in absolute power levels or melting progress, highlighting the core structural deviation morphology of various abnormal states, and enhancing the stability of complex abnormal feature extraction.

[0022] 4. This system relies on trajectory similarity to output abnormal confirmation, pending verification, or normal operation results, and accurately switches control commands accordingly. When bridging is confirmed, the power is reduced and the frequency is adjusted to curb the risk of local overheating. During the pending verification stage, the current power is maintained and the sampling period is shortened to prevent erroneous adjustment. During the normal stage, efficiency optimization is performed. This hierarchical response mechanism solves the problem of blindly increasing power under incomplete information conditions and realizes closed-loop control with safety as the priority.

[0023] 5. This system uses a timing synchronization unit to strictly align the instantaneous electrical parameters of the induction coil circuit with the heat exchange status data of the cooling circuit. This effectively overcomes the problem of data time misalignment caused by the lag of thermodynamic responses such as cooling water temperature difference behind electromagnetic responses such as voltage and current. It ensures accurate mapping of multi-dimensional operating data at the same physical time node, providing reliable basic data input for subsequent comparison of similar residuals and accurate comprehensive judgment of furnace conditions. Attached Figure Description

[0024] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0025] Figure 1 This is a schematic diagram of a module of an adaptive algorithm-based medium-frequency furnace operation control system provided in an embodiment of this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0027] An adaptive algorithm-based operation control system for a medium-frequency furnace includes:

[0028] The data acquisition module is used to collect the timing data of electrical parameters, including induction coil voltage, current, frequency and power, and heat exchange status data of the cooling circuit of the medium frequency furnace coupled load circuit, and generate real-time operating data.

[0029] The benchmark reconstruction module is used to reconstruct an ideal operating benchmark based on a preset multiphysics mechanism model and real-time operating data. The multiphysics mechanism model includes at least an electromagnetic coupling sub-model and a heat exchange sub-model.

[0030] The injection simulation module is used to inject preset anomaly mechanism parameters into the ideal operating benchmark to generate theoretical anomaly trajectories;

[0031] The residual extraction module is used to generate realistic residuals based on real-time operating data and ideal operating benchmarks, and to generate theoretical residuals based on theoretical abnormal trajectories and ideal operating benchmarks.

[0032] The decision control module, whose signal output terminal is connected to the intermediate frequency power controller of the intermediate frequency furnace, is used to determine the state evaluation result based on the trajectory similarity between the actual residual and the theoretical residual, and to generate adaptive control commands.

[0033] The control execution module is used to receive adaptive control commands and control the output power, output frequency and / or operating mode of the intermediate frequency power supply to adjust the operating status of the intermediate frequency furnace coupled load circuit.

[0034] This embodiment provides an adaptive algorithm-based medium-frequency furnace operation control mechanism. Specifically, taking a medium-frequency furnace in a foundry workshop used for melting scrap steel and preparing molten steel as the main scenario, the entire system operates around the complete life cycle of the same furnace: first, it goes through charging and melting, then through the solid-liquid coexistence stage, and then enters the stable melting and heat preservation stage. During this period, it continuously judges whether the current impedance change belongs to the normal melting evolution, or whether dangerous conditions such as furnace lining erosion or furnace charge bridging have occurred, and switches the control strategy accordingly.

[0035] Specifically, the data acquisition module synchronously acquires furnace condition information from the intermediate frequency power supply side and the cooling circuit side; the electrical parameters reflect the electromagnetic coupling state between the induction coil and the metal inside the furnace, and the heat exchange state reflects the heat load borne by the furnace wall, coil and cooling system.

[0036] For medium-frequency furnaces, fluctuations in current, frequency, power, and cooling water temperature difference exceeding preset conventional thresholds may occur during normal smelting. This is because the furnace charge gradually collapses, contacts, softens, and melts from a blocky solid state, and the equivalent magnetic conductivity and heat distribution continuously change. Therefore, it is difficult to distinguish between normal phase transition disturbances and real fault symptoms based solely on a single-point threshold at a certain moment.

[0037] The benchmark reconstruction module does not directly use historical averages as a reference. Instead, it establishes a state that the system should exhibit at the current time point if the furnace condition is healthy, the charge distribution is uniform, and the furnace lining is intact, based on the actual input conditions of the current furnace. This ideal operating benchmark includes at least two mechanisms: electromagnetic coupling and heat exchange. The former focuses on the coupling strength, equivalent impedance change, and frequency response between the coil and the melt, while the latter focuses on the reasonable temperature rise process when heat is transferred from inside the furnace to the furnace body and then to the cooling circuit. The benchmark obtained in this way is not an abstract mathematical curve, but a physical mapping of a healthy melting trajectory.

[0038] The injection simulation module further extrapolates from this, not responding after a fault occurs, but actively superimposing the mechanism parameters of typical anomalies onto the ideal operating baseline; for example, thinning of the furnace lining will change the coupling relationship between the coil and the melt, causing the impedance trajectory, which should change smoothly, to show a continuous deviation; bridging of the furnace charge will cause the upper part of the charge to be suspended and the lower part to be locally overheated, accompanied by abrupt changes in the local electromagnetic path, which manifests as sudden anomalies in frequency and impedance trajectory; the system forms multiple sets of theoretical anomaly trajectories based on this, which are equivalent to pre-prepared references for anomaly evolution;

[0039] The residual extraction module constructs two comparison paths: First, it compares real-time operating data with the ideal operating benchmark to obtain the actual residual, which contains real furnace condition changes, charging disturbances, power grid fluctuations, and sensor noise. Second, it compares theoretical anomaly trajectories with the ideal operating benchmark to obtain the theoretical residual, which is closer to the anomaly characteristics extracted by the mechanism model. The system compares the similarity of the forms of these two types of residuals to determine whether the changes in reality have a consistent evolutionary trend with a certain type of anomaly.

[0040] The decision control module does not regard all deviations as dangerous, but outputs the state evaluation result based on the similarity of the residual trajectory. If the actual residual is highly consistent with a certain theoretical abnormal residual, it indicates that the abnormality is not a normal melting fluctuation, but a dangerous state with a clear mechanism. The control execution module adjusts the output power, output frequency or working mode of the intermediate frequency power supply accordingly, so that the system changes from a simple adjustment method that pursues melting efficiency to an adaptive control method that takes into account safety, stability and operational reliability.

[0041] As a fault-tolerant mechanism, if sampling is missing, temperature sensor is distorted for a short time, or external power grid disturbance exceeds the preset allowable range during a certain period, the system may temporarily not output a forced abnormal conclusion, but maintain the current control state and mark that period as a state to be verified, waiting for more data to be supplemented before re-comparing; if key data is missing for multiple consecutive cycles, the control execution module will first enter a conservative operation mode to limit the power increase, so as to avoid erroneous adjustment under incomplete information conditions;

[0042] For example, in the middle of the same heat, the scrap steel has entered a semi-melting state, and the coil current and frequency fluctuate beyond the set range, while the temperature difference between the inlet and outlet of the cooling water increases simultaneously. If the system relies solely on traditional adaptive control, it is easy to regard such fluctuations as a load reduction and continue to increase power. However, in this embodiment, the reference reconstruction module first generates the ideal impedance and temperature rise trajectory under healthy melting, and then the injection simulation module generates the theoretical abnormal trajectory when the furnace charge is bridged. When the actual residual is close to the theoretical residual, the decision control module determines that the current situation is not a normal phase transition, but is closer to the local abnormal heating caused by bridging, and immediately issues power reduction and frequency adjustment commands.

[0043] The purpose of this step is to transform the control of the medium-frequency furnace from a simple feedback-following control method based on real-time parameter changes to a mechanism-driven method that first establishes a healthy reference, then identifies the source of deviation, and makes targeted adjustments. This will enable early identification of furnace lining erosion and charge bridging, and reduce the risk of furnace penetration, overheating, or abnormal equipment operation caused by misjudgment.

[0044] In a preferred embodiment of the present invention, the data acquisition module includes: an electrical parameter acquisition unit for acquiring induction coil voltage, induction coil current, operating frequency, active power, and reactive power; a thermal state acquisition unit for acquiring the temperature difference between the inlet and outlet mediums of the cooling circuit and the circulation flow rate; and a timing synchronization unit for time-aligning the induction coil voltage, induction coil current, operating frequency, active power, reactive power, temperature difference between the inlet and outlet mediums, and circulation flow rate to generate real-time operating data.

[0045] This embodiment provides a real-time operation data acquisition and synchronization mechanism. Specifically, continuing the same furnace scenario, the rate of change of each sensor quantity is not consistent in the initial charging stage, the middle melting stage, and the end of the holding stage: the rate of change of voltage, current, and operating frequency is higher than the preset value, while the changes in cooling water temperature difference and circulation flow rate have a time delay. If these quantities are directly involved in subsequent judgments without unified time alignment, it is easy to incorrectly superimpose physical phenomena that originally belong to different time points, resulting in state recognition deviation.

[0046] Specifically, the electrical parameter acquisition unit can be set on the output side of the medium frequency power supply or at the corresponding detection position of the induction coil circuit to continuously acquire the induction coil voltage, induction coil current, operating frequency, active power, and reactive power. This set of data reflects the energized state of the coupled load. Among them, voltage and current reflect the instantaneous load characteristics of the circuit, operating frequency reflects the inverter control's tracking of the load resonance point, active power reflects the actual melting work level, and reactive power reflects the electromagnetic energy storage and coupling deviation.

[0047] The thermal state acquisition unit is installed in the cooling circuit to collect the temperature difference between the inlet and outlet media and the circulation flow rate. The temperature difference between the inlet and outlet media corresponds to the intensity of heat released by the furnace body and coils to the cooling water, while the circulation flow rate determines the response mode of the temperature difference under the same heat load. For problems such as furnace lining erosion, relying solely on electrical parameters may only show a drift below the judgment threshold. However, if the heat load of the cooling circuit is observed to be higher than the preset normal value at the same time, it will be more helpful to identify the abnormal wall heat transfer behind it.

[0048] The timing synchronization unit is used to align data from different sources on a unified time axis. It can use methods such as unified sampling clock, timestamp resampling, or window alignment to ensure that the same set of real-time running data corresponds to the furnace condition status near the same physical moment. For example, within a short time window, electrical parameters may have changed first, while the cooling water temperature difference will appear later. The timing synchronization unit can map them to the same furnace condition segment according to preset synchronization rules for subsequent modules to perform same-source comparison.

[0049] As a fault-tolerance mechanism, if a certain electrical parameter experiences a short-term interruption, the effective value of the most recent stable interval can be temporarily stored and marked as reduced confidence data. In subsequent trajectory similarity calculations, the time window containing reduced confidence data will be assigned a lower calculation weight than normal data. If the cooling circuit flow signal changes abnormally and is inconsistent with the on-site pump unit operating conditions, the direct contribution of that channel to the thermal state can be suspended to prevent misjudgment caused by a single sensor failure. If both critical electrical and thermal parameters are missing, the system should freeze the abnormal judgment of that time window and only retain the basic power limit protection.

[0050] For example, in the first ten minutes after charging in the same heat, the contact between scrap steel blocks is unstable, the coil current fluctuates frequently, and the temperature difference of the cooling water rises relatively slowly. After timing synchronization, the system can attribute the temperature difference change that occurs slightly later after the current change to the same charging disturbance event, instead of mistakenly identifying them as two unrelated anomalies. The real-time operating data formed in this way is closer to a complete expression of the actual furnace condition.

[0051] The purpose of this step is to provide unified, comparable, and physically relevant input data for subsequent ideal benchmark reconstruction and anomaly identification, thereby achieving synchronous perception of the electromagnetic state and heat exchange state of the medium-frequency furnace.

[0052] In a preferred embodiment of the present invention, the reference reconstruction module includes: an electromagnetic modeling unit for calculating an ideal impedance trajectory based on a preset multiphysics mechanism model and real-time operating data; a thermal modeling unit for calculating an ideal temperature rise trajectory based on a preset multiphysics mechanism model and real-time operating data; and a reference fusion unit for calculating an ideal frequency drift trajectory by coupling the ideal impedance trajectory and the ideal temperature rise trajectory, and generating an ideal operating reference based on the ideal impedance trajectory, the ideal temperature rise trajectory, and the ideal frequency drift trajectory.

[0053] This embodiment provides an ideal operating baseline reconstruction mechanism. Specifically, although the raw data collected above reflects the actual furnace condition, it contains normal process fluctuations and potential abnormal signs. If real-time data is used directly as the control basis, the system may still be affected by instantaneous fluctuations and perform ineffective compensation control. Therefore, it is necessary to first construct an ideal operating baseline under the current process stage and when the equipment is in a healthy state.

[0054] Specifically, the electromagnetic modeling unit is used to reconstruct the ideal impedance trajectory. In an intermediate frequency furnace, the induction coil and the molten material inside the furnace form a coupled circuit. The impedance is not a fixed value, but varies with the shape of the furnace charge, the conductive path, the coupling distance, and the filling state. Under healthy conditions, although this variation is complex, it has interpretable physical boundaries: for example, as the metal gradually melts, the current path tends to be continuous, and the coupling relationship should change from discrete contact to a more uniform liquid coupling. Based on this, the electromagnetic modeling unit provides the ideal impedance variation trend that should occur in this process stage.

[0055] Specifically, the electromagnetic modeling unit adopts an equivalent circuit network model, which equates the induction coil and the furnace charge under healthy conditions to the primary and secondary windings of a transformer, respectively. The voltage and operating frequency in the real-time operating data are used as boundary input conditions. Combined with the preset equivalent conductivity, permeability and physical property parameter curves of healthy furnace charge as well as the curves of temperature-dependent physical properties, the ideal impedance trajectory that should be exhibited in this process stage is calculated.

[0056] The thermal modeling unit is used to reconstruct the ideal temperature rise trajectory. The heat absorbed in the furnace will be transferred outward through the furnace lining, coils and furnace structure, and then carried away by the cooling circuit. If the furnace lining is intact, the material is evenly distributed and there is no local overheating, the heat load borne by the cooling system should match the input power and the melting stage. The thermal modeling unit provides the ideal temperature rise trajectory under healthy melting conditions through heat exchange mechanism constraints, which describes how the heat should be released.

[0057] Specifically, the thermal modeling unit uses the active power in the real-time operating data as the effective heat source input. Based on the preset thermal resistance, heat capacity parameters and heat conduction equations of each layer of the healthy furnace lining structure, it calculates the ideal heat flow rate from the inside of the furnace to the cooling circuit interface, and then combines the real-time measured circulation flow rate to derive the ideal temperature rise trajectory of the inlet and outlet medium.

[0058] As a preferred embodiment of the present invention, in order to avoid the difficulty of real-time calculation caused by directly solving the complex three-dimensional partial differential heat conduction equation, the thermal modeling unit adopts the lumped parameter method to discretize the furnace lining, induction coil and furnace body into several one-dimensional equivalent thermal resistance nodes along the radial direction. The heat flow transfer between adjacent nodes depends only on the ratio of node temperature difference to equivalent thermal resistance, thereby transforming the abstract heat conduction equation into a clearly defined data flow node balance operation.

[0059] The reference fusion unit further couples the ideal impedance trajectory with the ideal temperature rise trajectory to generate the ideal frequency drift trajectory; the operating frequency of the intermediate frequency power supply is not an isolated quantity, it is affected by the changes in the electromagnetic parameters of the circuit, as well as the changes in material properties caused by temperature changes; therefore, the ideal frequency drift trajectory can be regarded as a healthy response under the combined action of electromagnetic and thermal states.

[0060] Ultimately, the ideal impedance trajectory, the ideal temperature rise trajectory, and the ideal frequency drift trajectory together constitute the ideal operating reference.

[0061] As a fault-tolerance mechanism, if the current furnace is in an extreme uneven charging phase or a transition phase after a temporary power outage, the model input conditions may deviate from the normal process window. In this case, the reference fusion unit can reduce its sensitivity to short-term frequency drift and retain only the more stable impedance and temperature rise references. The complete reference reconstruction will be restored after the system re-enters the continuous melting state. If the thermal state data quality is insufficient, the ideal temperature rise trajectory can be conservatively estimated to avoid false accuracy affecting the overall judgment.

[0062] For example, in the middle of the same heat, the scrap steel has transitioned from a stockpile to a continuous collapse state; the electromagnetic modeling unit can determine that the change in healthy impedance should be a gradual convergence rather than a sudden jump; the thermal modeling unit can determine that the cooling load should rise steadily with the progress of melting rather than a local spike; the ideal operating benchmark formed after benchmark fusion represents the expected trajectory of this heat under conditions of no bridging and no erosion; any subsequent phenomenon that significantly deviates from this trajectory is more likely to point to a real anomaly.

[0063] The purpose of this step is to provide a health reference that is dynamically changing with each furnace stage and has physical meaning for anomaly identification, thereby achieving effective separation of normal phase transition disturbances and dangerous anomalies.

[0064] In a preferred embodiment of the present invention, the injection simulation module includes: an erosion injection unit, used to inject a preset coupling attenuation parameter characterizing the degree of furnace lining thickness attenuation into an ideal operating reference to generate a theoretical trajectory of furnace lining erosion; a bridging injection unit, used to inject a preset local thermal resistance mutation parameter and a preset equivalent inductance step parameter into the ideal operating reference to generate a theoretical trajectory of furnace charge bridging; and a template generation unit, used to generate a theoretical abnormal trajectory based on the theoretical trajectory of furnace lining erosion and the theoretical trajectory of furnace charge bridging.

[0065] This embodiment provides an anomaly mechanism injection simulation mechanism; specifically, an ideal operating benchmark alone is insufficient to support fine discrimination, because although the system can know that the actual data deviates from the healthy state, it cannot clearly determine whether the deviation feature belongs to furnace lining erosion or furnace charge bridging; therefore, it is necessary to inject typical anomalies into the healthy benchmark in the form of mechanism parameters to form a comparable theoretical anomaly trajectory.

[0066] Specifically, the erosion injection unit is used to simulate the effect of furnace lining thickness decay on the coupling loop. After the furnace lining becomes thinner, the geometric and thermal boundaries between the coil and the melt change, which may change the electromagnetic coupling strength and the path of heat transfer to the outside. The result is usually a continuous gradual change rather than a one-time step, showing an evolutionary trend of continuous offset and increasing deviation. Therefore, after injecting the parameters characterizing coupling decay into the healthy baseline, the theoretical trajectory of furnace lining erosion can be obtained, which focuses more on long-term, slow-changing, and cumulative anomaly characteristics.

[0067] The bridging injection unit is used to simulate the bridging condition of the furnace charge. Bridging refers to the formation of a suspended structure in the upper part of the furnace charge, with local areas in the lower part overheating first, while the overall melt distribution is not uniform. This will lead to two significant consequences: first, the local thermal resistance will change abruptly, and the heat cannot be evenly diffused in a healthy state; second, the equivalent inductance response will change abruptly because the actual coupling path is reorganized. After injecting these two types of parameters into the ideal operating benchmark, the resulting theoretical trajectory of furnace charge bridging is usually more abrupt and phased.

[0068] The template generation unit forms a theoretical set of anomalous trajectories based on the two types of trajectories mentioned above. This set can cover light, moderate and heavy erosion, as well as different evolutionary segments such as the early stage of bridging formation, the stable suspension period and the local collapse period. The significance of doing so is that anomalies in reality do not always appear completely in the standard form. The system needs to prepare multiple comparable anomaly templates in order to improve the robustness of identification.

[0069] As a fault-tolerance mechanism, if the process formula of a certain furnace deviates from the preset standard range, such as abnormally large charge particle size or the use of special alloy furnace charge, the bridging trajectory and erosion trajectory may be subject to additional interference. At this time, the template generation range can be limited, and only the theoretical trajectory consistent with the current furnace type, furnace volume and process stage can be retained to avoid introducing incompatible templates. If the abnormal mechanism parameters lack the latest calibration, the system can call the parameter set that has been confirmed to be valid after the last maintenance and reduce the current judgment result to an auxiliary reference.

[0070] For example, in the same furnace cycle, if the health benchmark indicates that the impedance should transition smoothly, but in reality there is a short-term frequency spike, a local increase in cooling load, and an incomplete recovery, the system will call the erosion injection trajectory and the bridging injection trajectory for comparison. If the actual trajectory is closer to the abrupt segment after the formation of a local suspension in the bridging template, the subsequent judgment will prioritize convergence in the bridging direction, rather than simply classifying it as ordinary power grid fluctuations. The purpose of this mechanism is to transform the abnormal furnace conditions that are common in industry experience but difficult to quantify directly into traceable, comparable, and interpretable theoretical evolution trajectories, thereby achieving targeted identification of different dangerous states.

[0071] In a preferred embodiment of the present invention, the template generation unit is further configured to: classify and store theoretical abnormal trajectories and / or theoretical residuals obtained therefrom based on preset disturbance types, so as to form a theoretical abnormal trajectory template library and / or a theoretical residual template library; wherein, the preset disturbance types include normal phase transformation disturbance, charging disturbance, furnace lining erosion disturbance and furnace charge bridging disturbance.

[0072] This embodiment provides a theoretical template classification and storage mechanism. Specifically, generating only a few abnormal trajectories still has a problem: in actual furnace runs, normal phase change disturbances and feeding disturbances can also cause significant fluctuations. If the system only stores abnormal templates and lacks normal disturbance templates, it may still misjudge normal phase change events as equipment failures under high fluctuation conditions. Therefore, this embodiment further classifies and stores the theoretical trajectories and their corresponding residuals to form a template library.

[0073] Specifically, the template library includes at least four types: normal phase transition disturbance, charging disturbance, furnace lining erosion disturbance, and furnace charge bridging disturbance. The normal phase transition disturbance template is used to express the fluctuation boundary that should exist when the lump material gradually softens, the molten pool expands, and the conductive path becomes continuous. The charging disturbance template is used to express the transition state in which the coupling relationship is temporarily disrupted after the addition of new material.

[0074] The furnace lining erosion disturbance template is used to express the offset characteristics caused by cumulative degradation; the furnace charge bridging disturbance template is used to express the sudden characteristics caused by local suspension and bottom overheating; by incorporating both abnormal templates and non-abnormal but strong disturbance templates, the system can avoid attributing all deviations to faults.

[0075] In terms of implementation, the template library can store both the theoretical anomaly trajectory itself and the theoretical residuals obtained from it; the former retains a more complete physical evolution background and is suitable for full-process comparison; the latter highlights the deviation pattern relative to the healthy benchmark and is suitable for rapid classification; for different furnace types, different capacities and different typical material types, a hierarchical template library can also be established to match the scope of use with the field equipment.

[0076] As a fault-tolerance mechanism, if a new type of disturbance is not yet included in the existing classification, such as unconventional fluctuations caused by special feeding methods, the system can first classify it into the undetermined disturbance buffer area and not directly merge it into the mature template library. After the type of working condition is verified on-site, it will be transferred to the formal classification. If a template no longer matches the actual data for a long time, such as the original template becoming invalid after equipment modification, it should be suspended from participating in the main decision and only retained as a historical reference.

[0077] For example, after charging again in the same heat, the system detects that the fluctuations in current and frequency exceed the preset fluctuation thresholds, but the heat load of the cooling circuit does not show abnormal spikes simultaneously. At this time, if there is a charging disturbance template in the template library, the system can prioritize comparing the actual residual with this type of template, rather than directly judging it in the bridging direction. Only when it is significantly different from the charging disturbance template but closer to the bridging disturbance template will it be further upgraded to an anomaly judgment. The purpose of this mechanism is to improve the differentiation ability in complex smelting scenarios by establishing a classified template library that covers normal disturbances and abnormal disturbances, thereby reducing false alarms and enhancing the interpretability of the judgment results.

[0078] In a preferred embodiment of the present invention, the residual extraction module includes: a real-time differential unit, used to perform differential calculation between real-time operating data and an ideal operating benchmark to generate a real-time residual; a theoretical differential unit, used to perform differential calculation between a theoretical abnormal trajectory and an ideal operating benchmark to generate a theoretical residual; and a normalization unit, used to normalize the magnitude and time scale of the real-time residual and the theoretical residual; wherein the real-time residual and the theoretical residual are multi-dimensional time-series residual sequences constructed according to predetermined feature dimensions within a preset time sliding window, and the predetermined feature dimensions include at least impedance deviation, temperature rise deviation, and frequency drift deviation.

[0079] This embodiment provides a multi-dimensional time-series residual extraction mechanism. Specifically, after obtaining the health baseline and various theoretical trajectories, the system cannot directly use the original curves for judgment because there are significant differences in the distribution range of absolute values ​​for different furnace batches and different process stages. The more suitable feature indicators for anomaly identification are: the deviation magnitude of the relative health state, the duration of the deviation, and the spatiotemporal combination relationship of each deviation dimension. The residual extraction module is used to extract the above feature indicators.

[0080] Specifically, the real-time differential unit compares real-time operating data with the ideal operating benchmark to form a real residual. The residual here is not a single value, but a deviation sequence that unfolds continuously over time. The impedance deviation mainly reflects the degree of electromagnetic deviation of the coupling circuit relative to the healthy state. The temperature rise deviation reflects whether the heat exchange state exceeds the range of healthy furnace conditions. The frequency drift deviation reflects whether the adjustment made by the power supply to track the load is abnormal. The combination of the three can better depict the real furnace condition than looking at a single dimension.

[0081] The theoretical difference unit processes the theoretical anomaly trajectory in the same way to form theoretical residuals; this can transform different types of anomalies from the original state description into a sequence of deviation characteristics relative to the healthy state; for example, the theoretical residual of furnace lining erosion is more likely to show a continuous deviation in the medium to long term, while the theoretical residual of furnace charge bridging is more likely to show a sharp increase in residual amplitude and accompanied by thermal deviation anomalies in certain time periods.

[0082] The normalization unit is used to normalize the amplitude and time scale of the actual residuals and theoretical residuals. The significance of amplitude normalization is that the absolute power levels of different furnaces may be different, but the deviation direction and structural relationship of the same anomaly are still comparable. The significance of time scale normalization is that the evolution rate of the same anomaly may be different in different furnaces, but its evolution order and form can still be similar.

[0083] In its specific implementation, amplitude normalization adopts per-unit processing based on the rated value of the current process stage, dividing the absolute difference of impedance deviation, temperature rise deviation and other quantities with different physical dimensions by the corresponding reference nominal value, and mapping it to the dimensionless numerical range.

[0084] In a preferred embodiment of the present invention, to prevent the normalization result from diverging due to an excessively small reference nominal value when the system is in a special stage such as initial melting, the normalization unit adds a preset positive smoothing micro constant to the reference nominal value corresponding to the denominator when performing division operations. This micro constant is an empirical real number that is greater than 0 and much smaller than the corresponding reference nominal value, so as to avoid triggering a division-by-zero error. The time-scale normalization is based on the relative progress of key nodes such as initial melting and full melting in the smelting process, and resamples the original residual sequences of different durations into a standard sequence of fixed length through linear interpolation. After this normalization process, the comparison basis of the system is converted to the morphological similarity of the normalized residual sequences, rather than the consistency of the original absolute values.

[0085] In terms of fault tolerance, if a certain dimension is missing for a short period of time, such as when the cooling water temperature difference signal is interrupted by maintenance operations, a temporary judgment can be made by constructing a dimension-reduced residual based on the existing dimensions, but the confidence level of the corresponding result should be lowered; if a certain dimension is distorted for a long period of time, normalization based on that dimension should be suspended to prevent errors from amplifying the residual; if the residual is too flat and close to the noise level, the system can directly mark it as insufficient information and not enter the high-risk judgment process.

[0086] For example, in the same heat, the actual data showed the following trends relative to the healthy baseline: the impedance deviation first surged and then remained stable, the temperature rise deviation increased, and the frequency drift deviation showed a follow-up change; the system combined these three deviation sequences into actual residuals, and then compared them with the bridging theory residuals and the charging theory residuals respectively; even if the total power of this heat is higher than that of the previous heat, as long as the normalized three-dimensional deviation shape is closer to the bridging template, the system can still identify its essence and not be misled by the absolute value;

[0087] The purpose of this mechanism is to transform raw multi-source data into multi-dimensional deviation features that can be compared across furnace batches and stages, thereby achieving stable extraction of complex anomaly patterns.

[0088] In a preferred embodiment of the present invention, the decision control module includes: a similarity calculation unit, used to calculate the trajectory similarity between the actual residual and the theoretical residual based on a dynamic time warping algorithm; and a state decision unit, used to output anomaly labels and confidence levels based on the trajectory similarity. The state decision unit is configured with a first threshold and a second threshold. The first threshold is calibrated based on the equipment safety limit deviation, and the second threshold is calibrated based on the disturbance standard deviation of historical normal operation data. The first threshold is greater than the second threshold. When the trajectory similarity is greater than or equal to the first threshold, an anomaly confirmation result is output; when the trajectory similarity is less than or equal to the second threshold, a normal operation result is output; and when the trajectory similarity is greater than the second threshold and less than the first threshold, a result to be verified is output.

[0089] This embodiment provides a state determination mechanism based on trajectory similarity. Specifically, if the state determination is based solely on the deviation amplitude at a single moment, it is easy to misjudge high-frequency fluctuations in the normal phase transition process as abnormal states. The more reliable determination criterion adopted by the system is whether the evolution sequence and change trend of the actual residual sequence and the theoretical residual sequence are consistent within a predetermined time window. Therefore, this embodiment uses trajectory similarity for determination.

[0090] Specifically, the similarity calculation unit compares the actual residual with the theoretical residual based on dynamic time warping. This algorithm can effectively quantify the structural consistency of the evolution characteristics of the same physical event in different furnaces even if the same physical event occurs at different evolution rates through a nonlinear time series alignment mechanism.

[0091] In practice, the similarity calculation unit constructs a distance matrix between the normalized real residual sequence and the theoretical residual sequence at each alignment time point, and uses Euclidean distance to quantify the point-to-point differences of the multidimensional feature sequences.

[0092] Under the preset global window bandwidth constraint, the dynamic programming algorithm is used to search for a regular path with the minimum cumulative distance from the starting point to the ending point in the distance matrix;

[0093] Finally, the minimum cumulative distance D is mapped to a trajectory similarity score S between 0 and 1 through an exponential decay function controlled by a proportional adjustment constant λ. Here, the proportional adjustment constant λ is a system adjustment coefficient normalized based on the dynamic time warping path length L, and it is inversely proportional to the path length L. For example, a value of λ... Where k is a constant fitted based on historical samples, used to control the decay rate of similarity score with increasing cumulative distance, and exp is an exponential function with the natural constant e as the base, and its calculation form is:

[0094]

[0095] Therefore, the smaller the minimum cumulative distance, the closer the trajectory similarity score is to 1; and as the distance increases, the score decreases smoothly according to the preset decay rate; for example, for the same bridging, some furnaces form faster and some furnaces form slower, but they may all show a sequential structure of impedance deviation first abnormal, temperature rise deviation increase, and frequency drift followed by change; through dynamic time warping path optimization and quantitative comparison, the system pays more attention to the mechanism outline of the abnormality, rather than a one-to-one correspondence at a fixed time.

[0096] The state decision unit outputs anomaly labels and confidence levels based on similarity. If the actual residual is highly similar to a certain type of theoretical residual, the anomaly label for that type is output and a high confidence level is assigned. If the similarity to all anomaly templates is low, the normal operation result is output. If it is in the middle range, it indicates that the actual data is not sufficient to support a clear conclusion. In this case, outputting the result to be verified is more in line with industrial safety requirements.

[0097] In terms of threshold settings, the first threshold is higher than the second threshold. The former is used to confirm anomalies, while the latter is used to eliminate anomalies. The numerical range formed between the two is used as the area to be verified to prevent the system from outputting erroneous adjustment commands under conditions of insufficient feature information.

[0098] In practical applications, the first threshold can be calibrated based on the maximum permissible residual offset corresponding to the preset equipment safety level. For example, the lower quartile of multiple test residuals under the critical safety limit state can be taken. The second threshold can be adaptively updated based on the statistical average of daily disturbance intensity of historical normal furnaces under the same process stage. For example, the average of historical normal residuals plus three times the standard deviation can be taken as the second threshold to balance sensitivity and false alarm rate.

[0099] In the fault-tolerant processing of anomaly judgment, if the actual residual is similar to both types of templates, such as the boundary between feeding disturbance and early bridging is not clear, the system can compare the persistence, thermal deviation degree and historical window consistency of the two. If they still cannot be distinguished, the result to be verified will be output first rather than the anomaly confirmation will be forced. If the input window length is insufficient during the similarity calculation, no clear anomaly label will be output in this period, and only the trend cache will be updated.

[0100] For example, in the middle of the same furnace cycle, the system compares the most recent real residual with the normal phase change template, the charging template, the furnace lining erosion template, and the furnace charge bridging template respectively. If the similarity between the residual and the bridging template reaches a preset first threshold and is significantly different from other templates, the system outputs the furnace charge bridging label and high confidence. If the similarity between each template is lower than the preset second threshold, it indicates that the current disturbance is closer to normal noise, and the system outputs the normal operation result. If the residual is in the middle range, it is marked as pending verification.

[0101] The purpose of this mechanism is to enable the control system to balance recognition sensitivity and judgment prudence in high-risk industrial scenarios by retaining three levels: an anomaly confirmation area, a normal exclusion area, and a verification area.

[0102] In a preferred embodiment of the present invention, the decision control module further includes: a strategy switching unit, used to output a control command to reduce the output power and periodically adjust the operating frequency when an abnormal confirmation result is output and the abnormal label is furnace charge bridging; a conservative control unit, used to output a control command to maintain the current power and output a sampling control command to increase the sampling frequency and / or shorten the sampling period when a result to be verified is output; and an efficiency control unit, used to output a constant power adjustment command or a power factor optimization adjustment command when a normal operation result is output.

[0103] This embodiment provides a hierarchical control strategy switching mechanism. Specifically, if the control system only performs state recognition without outputting corresponding control commands, it cannot form a complete closed-loop control mechanism. At the same time, if the system adopts a single threshold shutdown strategy, it will reduce the continuity of system operation. Therefore, this embodiment corresponds to different control actions according to three results: anomaly confirmation, pending verification, and normal operation.

[0104] Specifically, when the status is confirmed and the abnormal label is furnace charge bridging, the strategy switching unit outputs control commands to reduce the output power and periodically adjust the operating frequency. The physical basis is that if the power is increased in the bridging state, the lower local area will overheat faster and the risk will be drastically amplified. Appropriate power reduction helps to reduce local thermal shock, and periodically adjusting the operating frequency helps to change the distribution of electromagnetic effects, causing the suspended structure to become unstable, collapse or recouple, thereby reducing the danger caused by the continued existence of bridging.

[0105] When the result is pending verification, the conservative control unit does not immediately execute large-scale adjustment control, but maintains the current power while increasing the sampling frequency and / or shortening the sampling period. The technical basis for this is that the system has detected a possible anomaly, but the evidence is not yet sufficient to support a forced power reduction. At this time, the system does not output adjustment commands, but increases the sampling frequency for encrypted monitoring, so that subsequent sampling data can more quickly reflect the evolution trend of the anomaly. This mechanism avoids erroneous adjustments to normal smelting conditions and also provides higher resolution data support for the status judgment of the next cycle.

[0106] When the result is normal operation, the efficiency control unit can output a constant power adjustment command or a power factor optimization adjustment command; that is, after confirming that the current deviation belongs to normal phase change or normal disturbance, the system can return to the control mode based on melting efficiency and power supply efficiency, maintain the established system output and optimize the electromagnetic coupling state.

[0107] Regarding the continuous state response mechanism, if the system is in a state of pending verification for several consecutive cycles and the similarity increases slowly, progressive conservative control can be triggered, such as limiting the power increase slope, without waiting for complete matching to the abnormal confirmation interval before outputting instructions; if the actual furnace condition does not improve after control execution, such as the frequency cycle adjustment not causing the residual to fall back after bridging confirmation, it can be further upgraded to a more stringent safety mode, including continuing to reduce power, prompting manual inspection, or suspending feeding; if the result fluctuates frequently between normal and pending verification, time lag should be introduced to avoid frequent switching of control instructions.

[0108] For example, in the same furnace cycle, the system first identifies a certain fluctuation as needing verification, so it maintains the current power unchanged and shortens the sampling period, so that the impedance and temperature rise deviations in the next time window are recorded more intensively. If subsequent data shows that its proximity to the bridging template further increases and reaches the confirmation zone, the system switches to a safe power reduction mode and adjusts the frequency according to a preset rhythm to reduce local overheating at the furnace bottom. Conversely, if intensive sampling finds that it is more consistent with normal feeding disturbances, the system resumes efficiency optimization control.

[0109] The purpose of this mechanism is to enable the identification results to directly drive differentiated control actions, thereby achieving a closed-loop operation that prioritizes safety, provides sufficient evidence, and balances efficiency.

[0110] As a preferred embodiment of the present invention, the medium-frequency furnace coupling load circuit is a coupling circuit composed of the medium-frequency furnace induction coil and the melt inside the furnace, the heat exchange status data is the status data of the medium-frequency furnace cooling water circulation system, and the status evaluation results include the furnace lining erosion status and the furnace charge bridging status.

[0111] This embodiment provides an application limitation mechanism for actual medium-frequency furnaces; specifically, the aforementioned system explicitly targets the coupled load circuit formed by the induction coil of the medium-frequency furnace and the melt inside the furnace, as well as the corresponding cooling water circulation system; such limitation gives the entire identification and control chain a clear industrial object basis, rather than an abstract general algorithm framework.

[0112] Specifically, the coupled load circuit is composed of an induction coil and the molten material inside the furnace. The coil is the energy input terminal, and the molten material is the main body for heating and conduction. The coupling state between the two determines the impedance, frequency, power absorption, and heat distribution. One of the most critical hidden dangers in the operation of a medium-frequency furnace is that this coupling state, while seemingly still having power input, has actually deviated from a healthy operating condition due to charge bridging or furnace lining degradation. From the perspective of power supply control alone, it is often only possible to monitor load changes, but not to identify the specific cause of these changes.

[0113] The heat exchange status data is limited to the status data of the cooling water circulation system, which has practical engineering significance. For medium-frequency furnaces, cooling water not only serves the function of heat dissipation, but also acts as an external mapping window of the furnace body's thermal state. When the furnace lining is eroded, the local heat transfer path will change, and the combination of cooling water temperature difference and flow rate may become abnormal. When the furnace charge bridging causes local overheating, the cooling system may also reflect the heat load characteristics under unhealthy operating conditions. Therefore, incorporating the cooling water circulation status into the evaluation can compensate for the limitations of relying solely on electrical parameters.

[0114] The condition assessment results include at least the furnace lining erosion condition and the furnace charge bridging condition; the former corresponds more to equipment life and structural safety issues, while the latter corresponds more to the immediate safety risks of the current furnace; the two can have different priorities in control: furnace charge bridging is more inclined to immediate intervention, while furnace lining erosion can trigger both conservative operation and maintenance early warning and equipment life cycle management.

[0115] As a fault-tolerant mechanism, if a complete cooling water circulation monitoring device is not installed on site, the system can first use electrical parameters to achieve basic identification, but the relevant status assessment should be marked as simplified mode, especially the confidence level of the judgment on furnace lining erosion should be reduced; if a stable melt has not yet formed in the furnace, for example, there are only scattered conductive contacts in the initial melting stage, the state of the coupling circuit fluctuates greatly, and the system can delay the output of structural anomaly conclusions until sustainable coupling is formed before entering the complete assessment.

[0116] For example, in the same heat cycle, once a stable molten pool has formed in the furnace, the system continuously observes the coupling state between the coil and the melt, as well as the cooling water circulation state. If the impedance trajectory is found to deviate from the healthy baseline for a long period of time and the cooling heat load gradually increases, the evaluation result may be biased towards furnace lining erosion. If a sudden abnormality is found in the impedance and frequency, and the cooling state shows local overheating characteristics, the evaluation result may be biased towards charge bridging. The two types of results are used for safety control and maintenance management, respectively.

[0117] The purpose of this mechanism is to firmly anchor the entire identification and control process to the actual electromagnetic coupling and cooling objects of the medium-frequency furnace, thereby achieving targeted assessment of the furnace lining erosion state and the charge bridging state.

[0118] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A medium-frequency furnace operation control system based on an adaptive algorithm, characterized in that, include: The data acquisition module is used to collect the timing data of electrical parameters, including induction coil voltage, current, frequency and power, and heat exchange status data of the cooling circuit of the medium frequency furnace coupled load circuit, and generate real-time operating data. The benchmark reconstruction module is used to reconstruct an ideal operating benchmark based on a preset multiphysics mechanism model and the real-time operating data. The multiphysics mechanism model includes at least an electromagnetic coupling sub-model and a heat exchange sub-model. The injection simulation module is used to inject preset abnormal mechanism parameters into the ideal operating benchmark to generate theoretical abnormal trajectories; The residual extraction module is used to generate a real residual based on the real-time running data and the ideal running benchmark, and to generate a theoretical residual based on the theoretical abnormal trajectory and the ideal running benchmark. The decision control module is used to determine the state evaluation result based on the trajectory similarity between the actual residual and the theoretical residual, and to generate adaptive control instructions; The control execution module has its signal output terminal connected to the intermediate frequency power controller of the intermediate frequency furnace. It is used to receive the adaptive control command and control the output power, output frequency and / or working mode of the intermediate frequency power supply to adjust the operating state of the intermediate frequency furnace coupled load circuit. The injection simulation module includes: an erosion injection unit, used to inject a preset coupling attenuation parameter characterizing the degree of furnace lining thickness attenuation into the ideal operating benchmark to generate a theoretical trajectory of furnace lining erosion; a bridging injection unit, used to inject a preset local thermal resistance mutation parameter and a preset equivalent inductance step parameter into the ideal operating benchmark to generate a theoretical trajectory of furnace charge bridging; and a template generation unit, used to generate the theoretical anomaly trajectory based on the theoretical trajectory of furnace lining erosion and the theoretical trajectory of furnace charge bridging. The template generation unit is further configured to: classify and store the theoretical abnormal trajectory and / or the theoretical residual obtained therefrom based on a preset disturbance type, so as to form a theoretical abnormal trajectory template library and / or a theoretical residual template library; wherein, the preset disturbance type includes normal phase transformation disturbance, charging disturbance, furnace lining erosion disturbance and furnace charge bridging disturbance.

2. The medium-frequency furnace operation control system based on an adaptive algorithm according to claim 1, characterized in that, The data acquisition module includes: an electrical parameter acquisition unit for acquiring induction coil voltage, induction coil current, operating frequency, active power, and reactive power; a thermal state acquisition unit for acquiring the temperature difference between the inlet and outlet mediums of the cooling circuit and the circulation flow rate; and a timing synchronization unit for time-aligning the induction coil voltage, induction coil current, operating frequency, active power, reactive power, temperature difference between the inlet and outlet mediums, and circulation flow rate to generate the real-time operating data.

3. The medium-frequency furnace operation control system based on an adaptive algorithm according to claim 1, characterized in that, The reference reconstruction module includes: an electromagnetic modeling unit for calculating an ideal impedance trajectory based on the preset multiphysics mechanism model and the real-time operating data; a thermal modeling unit for calculating an ideal temperature rise trajectory based on the preset multiphysics mechanism model and the real-time operating data; and a reference fusion unit for calculating an ideal frequency drift trajectory by coupling the ideal impedance trajectory and the ideal temperature rise trajectory, and generating the ideal operating reference based on the ideal impedance trajectory, the ideal temperature rise trajectory, and the ideal frequency drift trajectory.

4. The medium-frequency furnace operation control system based on an adaptive algorithm according to claim 1, characterized in that, The residual extraction module includes: a real-time difference unit, used to perform difference calculation between the real-time operating data and the ideal operating benchmark to generate the real-time residual; a theoretical difference unit, used to perform difference calculation between the theoretical abnormal trajectory and the ideal operating benchmark to generate the theoretical residual; and a normalization unit, used to normalize the amplitude and time scale of the real-time residual and the theoretical residual; wherein the real-time residual and the theoretical residual are multi-dimensional time-series residual sequences constructed according to predetermined feature dimensions within a preset time sliding window, and the predetermined feature dimensions include at least impedance deviation, temperature rise deviation, and frequency drift deviation.

5. The medium-frequency furnace operation control system based on an adaptive algorithm according to claim 1, characterized in that, The decision control module includes: The similarity calculation unit is used to calculate the trajectory similarity between the actual residual and the theoretical residual based on the dynamic time warping algorithm; A status decision unit is used to output an anomaly label and confidence level based on the trajectory similarity. The status decision unit has a first threshold and a second threshold. The first threshold is calibrated based on the equipment's safety limit deviation, and the second threshold is calibrated based on the disturbance standard deviation of historical normal operation data. The first threshold is greater than the second threshold. When the trajectory similarity is greater than or equal to the first threshold, an anomaly confirmation result is output; when the trajectory similarity is less than or equal to the second threshold, a normal operation result is output; when the trajectory similarity is greater than the second threshold and less than the first threshold, a result to be verified is output.

6. The medium-frequency furnace operation control system based on an adaptive algorithm according to claim 5, characterized in that, The decision control module also includes: The strategy switching unit is used to output a control command to reduce the output power and periodically adjust the operating frequency when the abnormality confirmation result is output and the abnormality label is furnace charge bridging. A conservative control unit is used to output a control command to maintain the current power when outputting the result to be verified, and to output a sampling control command to increase the sampling frequency and / or shorten the sampling period; An efficiency control unit is used to output a constant power adjustment command or a power factor optimization adjustment command when outputting the normal operation results.

7. The medium-frequency furnace operation control system based on an adaptive algorithm according to claim 1, characterized in that, The medium-frequency furnace coupling load circuit is a coupling circuit formed by the medium-frequency furnace induction coil and the melt inside the furnace. The heat exchange status data is the status data of the medium-frequency furnace cooling water circulation system. The status evaluation results include the furnace lining erosion status and the furnace charge bridging status.

Images