Overheating early warning method for skull of vacuum suspension smelting system

Abstract

The invention relates to the technical field of metallurgical equipment, in particular to an overheating early warning method for a skull of a vacuum suspension smelting system. Comprising the following steps: establishing a power-time-temperature prediction model, and obtaining a predicted surface temperature data set; the surface temperature of the crucible / melt is collected in real time, an actual surface temperature data set is obtained, and an instantaneous surface temperature difference deviation data set is obtained through the data set; establishing a cooling liquid temperature rise prediction model, and obtaining a cooling loop temperature difference deviation data set; normalizing the instantaneous surface temperature difference deviation data set and the cooling loop temperature difference deviation data set to obtain a comprehensive instantaneous risk score; carrying out cumulative anomaly index and time weighting to obtain a cumulative anomaly index; and the vacuum suspension smelting system enters a corresponding early warning state according to a corresponding rule. In this way, the problems that an existing monitoring means for the vacuum suspension smelting process is single in information, lagged in response and lack of identification capacity for the upper skull and the lower overheating coupling state are solved.

Description

Technical Field

[0001] This invention relates to the field of metallurgical equipment technology, and more specifically, to a method for early warning of overheating of the solidified shell in a vacuum suspension melting system. Background Technology

[0002] Vacuum levitation melting furnaces are equipment that utilizes high-frequency electromagnetic induction to achieve non-contact levitation heating and melting of molten metal. This type of equipment offers advantages such as high purity, no crucible contamination, and rapid temperature control response, and is widely used in the preparation of high-temperature alloys, rare metals, and special functional materials. However, during vacuum levitation melting, due to the low-pressure environment exposed above the melt and the complex effects of electromagnetic force and thermal convection, a solidified shell easily forms on the melt surface. If this shell fails to melt in time, it may block the high-temperature melt below, causing the surface temperature to be significantly lower than the actual temperature distribution inside the furnace. Under continuous heating, this can even lead to localized overheating of the melt below and an abnormal increase in the furnace wall heat load, posing risks of equipment damage and safety.

[0003] In existing technologies, single-point infrared thermography or thermal imaging is commonly used to obtain the temperature distribution on the surface of the melt in order to determine the temperature state of the melting process. However, these methods have the following main drawbacks:

[0004] 1. Insufficient representativeness of surface temperature measurement: Due to the high reflectivity and low emissivity of the solidified shell surface, and the possibility of exposed high-temperature melt at the cracks in the solidified shell, single-point infrared temperature measurement is prone to misinterpreting the high temperature in the crack as the overall temperature, and cannot truly reflect the average temperature state of the solidified shell above.

[0005] 2. Lack of indirect ability to determine overheating of the melt below: Existing surface temperature measurement and thermal imaging monitoring can only observe the temperature change of the melt surface and cannot directly obtain the temperature information of the area covered by the solidified shell; when the upper solidified shell forms, the temperature below continues to rise but the surface temperature remains unchanged, making it difficult for the system to identify the anomaly in a timely manner.

[0006] 3. The monitoring system is not sensitive to changes in operating conditions: Current temperature control is mostly based on instantaneous values ​​or single-parameter closed-loop regulation, without considering the dynamic relationship between input energy (heating power) and temperature changes. When power is continuously input but the surface temperature does not rise as expected, the control system cannot identify the risk signals behind this "temperature rise lag" phenomenon.

[0007] 4. Abnormal cooling load cannot be correlated with smelting status: Although abnormal coolant temperature or flow can reflect changes in equipment heat load, there is a lack of a comprehensive judgment mechanism that integrates smelting power and temperature response, which often leads to the delayed detection of abnormal trends. Summary of the Invention

[0008] To address the problems of existing monitoring methods for vacuum levitation melting processes, such as limited information, delayed response, and lack of ability to identify the overheating coupling state of the upper solidified shell and the lower shell, this invention provides an overheating early warning method for the solidified shell in a vacuum levitation melting system.

[0009] In a first aspect, the present invention provides an overheating early warning method for the solidified shell of a vacuum suspension melting system, the overheating early warning method comprising:

[0010] S1: Establish a power-time-temperature prediction model to obtain a dataset of predicted surface temperatures for the crucible / melt;

[0011] S2: Real-time acquisition of crucible / melt surface temperature, obtaining actual surface temperature dataset of crucible / melt, and obtaining instantaneous surface temperature difference deviation dataset by predicting surface temperature dataset and actual surface temperature dataset;

[0012] S3: Establish a coolant temperature rise prediction model, collect heating power and coolant mass flow rate in real time, and obtain a cooling circuit temperature difference deviation dataset;

[0013] S4: Normalize the collected instantaneous surface temperature difference deviation dataset and the cooling circuit temperature difference deviation dataset to obtain a comprehensive instantaneous risk score; and perform cumulative anomaly index and time weighting to obtain cumulative anomaly index.

[0014] S5: The vacuum suspension melting system enters a Level 1 early warning state based on the comprehensive instantaneous risk score and cumulative abnormal indicators according to the first rule;

[0015] Alternatively, the vacuum suspension melting system may enter a level-two early warning state based on the comprehensive instantaneous risk score and cumulative abnormal indicators through the second rule;

[0016] Alternatively, the vacuum suspension melting system may enter a level three early warning state based on the comprehensive instantaneous risk score and cumulative abnormal indicators through the third rule.

[0017] In some embodiments, in step S1, establishing a power-time-temperature prediction model includes: collecting historical operating data of the vacuum suspension melting furnace under different charging amounts, different materials and different process parameters; synchronizing the historical operating data according to a unified time reference; removing, smoothing or filtering outliers; constructing an input sequence reflecting power-time behavior characteristics according to a preset time window; using the surface temperature of the crucible / melt at each time point as the model output target; establishing a power-time-temperature mapping model for time series prediction; and fitting the model parameters by minimizing the error between the predicted temperature and the actual measured temperature.

[0018] The historical operating data includes: heating power, cumulative energy, heating duration, coolant inlet temperature, coolant outlet temperature, and crucible / melt surface temperature as of time.

[0019] In some embodiments, in step S1, heating power and heating time are collected in real time to obtain a dataset of predicted surface temperatures of the crucible / melt.

[0020] In some embodiments, in step S1, establishing the power-time-temperature prediction model includes: treating the input coil power as the input heat source of the vacuum levitation melting system, and establishing a theoretical temperature rise relationship between the melt and surface temperatures by calculating the effective absorbed energy per unit time. In the formula, For real-time input power to the coil, For electromagnetic coupling efficiency, Where c is the melt mass and c is the specific heat capacity. To predict the rate of change of surface temperature, This refers to the heat dissipation component that changes with temperature.

[0021] In some embodiments, in step S1, the heating time is collected in real time to obtain a dataset of predicted surface temperatures of the crucible / melt.

[0022] In some embodiments, in step S3, establishing a coolant temperature rise prediction model includes: treating the input coil power as the input heat source of the vacuum suspension melting system, and establishing the temperature difference deviation relationship of the coolant's cooling circuit, i.e. In the formula, For real-time input power to the coil, Where cp is the coolant mass flow rate, and cp is the coolant's specific heat capacity at constant pressure. For the temperature difference deviation in the cooling circuit, This is the effective heat transfer ratio parameter.

[0023] In some embodiments, the normalization process in step S4 includes:

[0024] Define the normalized instantaneous score of the instantaneous surface temperature difference deviation dataset, i.e. In the formula, As the first reference standard, This refers to the instantaneous surface temperature difference deviation. This is the first normalized instantaneous score;

[0025] Define the normalized instantaneous score of the cooling loop temperature difference deviation dataset, i.e. In the formula, As the second reference scale, For the temperature difference deviation in the cooling circuit, This is the second normalized instantaneous score;

[0026] Define the comprehensive instantaneous risk score, i.e. In the formula, This is the first initial recommended value. This is the second initial recommended value. To calculate the comprehensive instantaneous risk score.

[0027] In some embodiments, in step S4, the cumulative anomaly index and time weighting include:

[0028] Weighted accumulation is used, i.e. In the formula, Here, W is the time weighting function, and W is the cumulative window length. For cumulative abnormal indicators;

[0029] To achieve discretization, i.e. In the formula, , The sampling interval is denoted as .

[0030] In some embodiments, the first rule is: and ;in, The first-level alarm threshold, Set a threshold for the first one;

[0031] The vacuum suspension melting system enters a first-level early warning state, which includes: interface prompts, event recording, and recommendations for manual inspection.

[0032] In some embodiments, the second rule is: ;or, and Exceeding the second set threshold; where, The first-level alarm threshold, This is the second-level alarm threshold. This refers to the outlet temperature of the coolant. This refers to the inlet temperature of the coolant.

[0033] The vacuum suspension melting system enters a level 2 early warning state by automatically reducing the heating power by a preset ratio or reducing it to the next power level, extending the steady-state observation time, and simultaneously triggering an audible or visual alarm and SMS or remote notification, and recording a snapshot.

[0034] In some embodiments, the third rule is: Or, instantaneously and Continued to rise; among them, This is the third-level alarm threshold. This is the critical threshold for the instantaneous change in temperature of the melt or crucible surface. This refers to the outlet temperature of the coolant. This refers to the inlet temperature of the coolant.

[0035] The vacuum suspension melting system enters a Level 3 early warning state, which means: immediately disconnecting the heating power supply, opening the emergency cooling circuit, recording all historical data, and notifying maintenance and on-duty management personnel.

[0036] To address the shortcomings of existing monitoring methods for vacuum suspension melting processes, such as limited information, delayed response, and lack of ability to identify the coupled state of the upper solidified shell and the lower overheating, this invention offers the following advantages:

[0037] The technical solution of this invention establishes a expected temperature rise curve for heating power and heating time during the smelting process. Combined with the real-time collected surface temperature and coolant temperature, the deviation value is calculated and the anomaly is accumulated. When the accumulated deviation exceeds the set threshold, an early warning is triggered, thereby realizing early monitoring and prevention of the formation of the upper shell and the overheating below. Attached Figure Description

[0038] Figure 1 A flowchart of an overheating early warning method for the solidified shell in a vacuum suspension melting system is shown. Detailed Implementation

[0039] The present disclosure will now be discussed with reference to several exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and thus implement the present disclosure, and are not intended to imply any limitation on the scope of the disclosure.

[0040] As used herein, the term "comprising" and its variations are to be interpreted as open-ended terms meaning "including but not limited to". The term "based on" is to be interpreted as "at least partially based on". The terms "one embodiment" and "an embodiment" are to be interpreted as "at least one embodiment". The term "another embodiment" is to be interpreted as "at least one other embodiment". The terms "upper", "lower", "left", "right", "front", "rear", "top", "bottom", "inner", "outer", "vertical", "horizontal", "lateral", "longitudinal", etc., indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings. These terms are primarily for the purpose of better describing this application and its embodiments and are not intended to limit the indicated devices, elements, or components to having a specific orientation or being constructed and operated in a specific orientation. Furthermore, some of the above terms may be used to indicate other meanings besides orientations or positional relationships; for example, the term "upper" may in some cases indicate a dependency or connection relationship. Those skilled in the art can understand the specific meaning of these terms in this application according to the specific circumstances. In addition, the terms "installed", "set up", "equipped with", "connected", and "linked" should be interpreted broadly. For example, it can be a fixed connection, a detachable connection, or an integral structure; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or an internal connection between two devices, elements, or components. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances. Furthermore, the terms "first," "second," etc., are mainly used to distinguish different devices, elements, or components (the specific types and structures may be the same or different), and are not used to indicate or imply the relative importance or quantity of the indicated devices, elements, or components. Unless otherwise stated, "a plurality of" means two or more.

[0041] Example 1

[0042] This embodiment discloses an overheating early warning method for the solidified shell in a vacuum suspension melting system, such as... Figure 1 As shown, the overheating early warning method includes the following steps:

[0043] S1: Establish a power-time-temperature prediction model to obtain a dataset of predicted surface temperatures for the crucible / melt;

[0044] S2: Real-time acquisition of crucible / melt surface temperature, obtaining actual surface temperature dataset of crucible / melt, and obtaining instantaneous surface temperature difference deviation dataset by predicting surface temperature dataset and actual surface temperature dataset;

[0045] S3: Establish a coolant temperature rise prediction model, collect heating power and coolant mass flow rate in real time, and obtain a cooling circuit temperature difference deviation dataset;

[0046] S4: Normalize the collected instantaneous surface temperature difference deviation dataset and the cooling circuit temperature difference deviation dataset to obtain a comprehensive instantaneous risk score; and perform cumulative anomaly index and time weighting to obtain cumulative anomaly index.

[0047] S5: The vacuum suspension melting system enters a Level 1 early warning state based on the comprehensive instantaneous risk score and cumulative abnormal indicators according to the first rule;

[0048] Alternatively, the vacuum suspension melting system may enter a level-two early warning state based on the comprehensive instantaneous risk score and cumulative abnormal indicators through the second rule;

[0049] Alternatively, the vacuum suspension melting system may enter a level three early warning state based on the comprehensive instantaneous risk score and cumulative abnormal indicators through the third rule.

[0050] In this embodiment, an overheating early warning method for the solidified shell of a vacuum suspension melting system is provided, which is applied to a vacuum suspension melting furnace. The overheating early warning method involves establishing a power-time-temperature prediction model and a coolant temperature rise prediction model.

[0051] Furthermore, a power-time-temperature prediction model is used to describe the dynamic relationship between coil heating power input and surface temperature response during the melting process in a vacuum levitation melting furnace. The establishment and application of this model includes offline modeling and online operation phases.

[0052] Specifically, during the offline modeling phase, historical operating data of the vacuum levitation melting furnace under different charging amounts, materials, and process parameters are collected. This historical operating data includes at least the heating power varying over time. Accumulated energy Heating time t, coolant inlet temperature Coolant outlet temperature Crucible / melt surface temperature The aforementioned data are synchronized according to a unified time benchmark, and outliers are removed, smoothed, or filtered. An input sequence reflecting power-time behavior characteristics is then constructed according to a preset time window. Using the surface temperature corresponding to each time point as the model output target, a power-time-temperature mapping model for time series prediction is established. The model parameters are fitted by minimizing the error between the predicted temperature and the actual measured temperature, thus obtaining the power-time-temperature prediction model. This power-time-temperature prediction model establishes a functional mapping relationship between the power input sequence and the surface temperature output; its calculation logic is as follows: based on the power, energy, and operating parameters within the current time and historical time windows, the surface temperature value at the corresponding time is predicted.

[0053] Specifically, during the online operation phase, the power-time-temperature prediction model collects heating power in real time (sampling frequency). Initial sampling frequency Using the current heating time and relevant operating parameters as input, the predicted surface temperature at the corresponding time point is calculated based on the trained model. It continuously outputs the predicted temperature rise curve over time, and obtains the predicted surface temperature dataset of the crucible / melt, which is used as a temperature reference under the current smelting conditions.

[0054] Simultaneously, the surface temperature of the crucible / melt is acquired in real time, which can be obtained from the thermal imaging ROI mean or area array averaging, with a sampling frequency of... (Initial sampling frequency) This process involves obtaining the actual surface temperature rise curve of the crucible / melt continuously outputting at that time, thus acquiring the actual surface temperature dataset of the crucible / melt. The instantaneous surface temperature difference deviation curve is then obtained by comparing the predicted surface temperature dataset with the actual surface temperature dataset, resulting in the instantaneous surface temperature difference deviation dataset. In this application, the instantaneous surface temperature difference deviation... (Unit: °C) A positive value indicates that the surface temperature of the crucible / melt is colder than expected.

[0055] To adapt to changes in charging volume, material, and furnace conditions, the aforementioned machine learning model can periodically update its parameters using incremental learning or sliding window retraining, allowing it to continuously absorb the latest operating data during online operation, thereby maintaining prediction accuracy and model stability. The prediction model established through this data-driven approach can automatically characterize the complex nonlinear relationship between heating power, time, and temperature, providing a reliable temperature reference curve for the instantaneous deviation calculation and early warning determination of this invention.

[0056] Furthermore, to obtain the expected temperature change of the cooling circuit, this invention establishes a coolant temperature rise prediction model in addition to the power-time model to improve the accuracy of overheating judgment. This model combines the simplified principle of energy conservation with an experimental calibration mechanism, and makes feasible predictions of coolant temperature changes by inputting power, coolant flow rate, and inlet temperature.

[0057] In actual smelting processes, not all the input power of the coil is transferred to the melt; some heat dissipates through the furnace wall, coil, and structural components into the coolant channels. To avoid complex calculations of heat dissipation, this invention employs an "effective heat transfer ratio parameter." To simplify, this parameter is calibrated using multiple sets of steady-state and quasi-steady-state melting experimental data, reflecting the average heat transfer efficiency under different furnace conditions, melt materials, loads, and coil conditions. (Calibrated...) It can be segmented according to power range, frequency or smelting stage, thus forming an adaptive parameter set.

[0058] Based on this, the coolant temperature rise prediction model is established by: treating the input coil power as the input heat source of the vacuum suspension melting system, and establishing the temperature difference deviation relationship of the coolant cooling circuit, that is, the expected temperature rise of the coolant can be expressed as: ;

[0059] in, For real-time input power to the coil, Let be the coolant mass flow rate, and cp be the coolant's specific heat capacity at constant pressure. This leads to the predicted coolant outlet temperature: In the formula, This refers to the inlet temperature of the coolant. This refers to the temperature difference deviation in the cooling circuit.

[0060] Therefore, as shown in the above formula, the real-time acquisition of heating power, coolant mass flow rate, and actual coolant outlet temperature is crucial. (sampling frequency) Initial sampling frequency ), to obtain the cooling circuit temperature difference deviation dataset.

[0061] To further improve the accuracy of the coolant temperature rise prediction model under dynamic operating conditions, this invention can also use a data-driven approach to analyze the above-mentioned... or Corrections were made. These were achieved by collecting data from a large number of historical smelting conditions. Data, Building A regression model with input variables. This regression model can employ multinomial fitting, random forest regression, gradient boosting regression, or other data training methods commonly used in the field, and the optimal model parameters are determined through cross-validation. This data-driven model can implicitly compensate for factors such as furnace wall heat dissipation, environmental changes, and coil aging, ensuring that the expected coolant temperature remains highly consistent under varying operating conditions.

[0062] Using the above method, the present invention can obtain a reliable expected temperature curve of the coolant outlet without complex calculations of electromagnetic coupling efficiency or structural heat dissipation, providing a stable reference for subsequent deviation calculation, condensation identification and overheating early warning.

[0063] Furthermore, the collected instantaneous surface temperature difference deviation dataset and cooling circuit temperature difference deviation dataset need to be normalized to obtain a comprehensive instantaneous risk score; and cumulative anomaly indexes are weighted by time to obtain cumulative anomaly indexes, which are used to reflect the comprehensive duration and magnitude of the deviation.

[0064] Specifically, the instantaneous surface temperature difference deviation is normalized to reduce the influence of dimensions and facilitate threshold setting: Here, the normalized instantaneous score of the instantaneous surface temperature difference deviation dataset is defined as... In the formula, The first reference scale indicates the order of magnitude of "significant deviation". This refers to the instantaneous surface temperature difference deviation. This is the first normalized instantaneous score.

[0065] Specifically, the temperature difference deviation of the cooling circuit is normalized to reduce the influence of dimensions and facilitate threshold setting: This involves defining the normalized instantaneous score of the cooling circuit temperature difference deviation dataset, i.e. In the formula, The second reference scale indicates the order of magnitude of "significant deviation". For the temperature difference deviation in the cooling circuit, This is the second normalized instantaneous score.

[0066] Specifically, The comprehensive risk score (dimensionless) is calculated from the normalized bias and weights; that is, the comprehensive instantaneous risk score can be expressed as: In the formula, The bias weight (dimensionless) Or set according to experience). This is the first initial recommended value. This is the second initial recommended value. To synthesize the instantaneous risk score, in this application, This serves as an initial recommendation (prioritizing surface deviations), but can be optimized based on historical data.

[0067] Specifically, cumulative abnormal indicators weighted by time:

[0068] To avoid false alarms caused by momentary jitter, weighted accumulation (sliding window integration) is used: ;in, The time-weighted function (which can be a constant of 1, or an exponential decay more biased towards the near end) (To emphasize recent bias); W is the cumulative window length, window length =120s is the initial value, which is adjustable, for example, 60–300s.

[0069] Discretization is achieved:

[0070] ;

[0071] in , The sampling interval is denoted as .

[0072] In this embodiment, the cumulative deviation index D cum(t) The calculation is performed using a discretization method. The system samples the surface temperature deviation at a preset sampling period and gradually accumulates the deviation at each discrete moment, thereby transforming the continuous-time deviation evolution process into a cumulative anomaly index in the form of a discrete-time series. This discretization method ensures that the calculation process of the cumulative deviation is consistent with the sampling period of the actual measurement data, facilitating real-time implementation in the control system. Simultaneously, it effectively suppresses instantaneous noise interference, highlights the continuous accumulation characteristics of abnormal deviations, and provides a stable and reliable basis for subsequent threshold determination and interlocking protection.

[0073] Furthermore, the vacuum suspension melting system of this vacuum suspension melting furnace defines three levels during the shell solidification process: Level 1 warning state, Level 2 warning state, and Level 3 warning state.

[0074] Specifically, the conditions for a Level 1 warning state are as follows: and ;in, The first-level alarm threshold, Set a threshold for the first one.

[0075] Its actions are: providing interface prompts, recording events, and recommending manual inspections; it does not intervene automatically.

[0076] Its purpose is to detect deviation trends early and manually check whether they are transient or interference.

[0077] Specifically, the conditions for a Level II alert are: ;or, and Exceeding the second set threshold; where, The first-level alarm threshold, This is the second-level alarm threshold. This refers to the outlet temperature of the coolant. The inlet temperature of the coolant is defined as the inlet temperature. In this application, the second set threshold is the acceptable rate of change during normal smelting. Specifically, when the rate of increase in the cooling water temperature difference exceeds the acceptable rate of change during normal smelting, it is determined to be an abnormal heat release trend.

[0078] Its operation is as follows: automatically reduce the heating power by a preset ratio (recommended reduction of 10%–30%, or reduce to the next power level), extend the steady-state observation time, and simultaneously trigger sound / light alarms and SMS / remote notifications, and record snapshots (thermal image, power, cooling temperature), etc.

[0079] Its purpose is to reduce heat input as quickly as possible to mitigate risks in the event of suspected internal heat accumulation.

[0080] Specifically, the conditions for a Level III alert are as follows: Or, instantaneously and Continued to rise; among them, This is the third-level alarm threshold. It is the critical threshold for the instantaneous change in surface temperature of the melt or crucible, that is, the risk boundary of abnormal increase in surface temperature per unit time.

[0081] The process involves: initiating the interlock protection procedure—immediately disconnecting the heating power supply, opening the emergency cooling circuit (if safety permits), recording all historical data, and notifying maintenance and on-duty management personnel. The system enters "pending verification" mode and requires manual reset to recover.

[0082] Its purpose is to ensure the safety of personnel and equipment and to prevent major accidents.

[0083] Each threshold needs to be optimized on pilot data according to the ROC curve (hit rate vs. false alarm rate).

[0084] In this embodiment, debouncing and hysteresis processing are also required: 1) Time debouncing of triggering: requiring that the triggering conditions be met continuously for at least 1) Alarms are only confirmed after the alarm is cleared to avoid short-term spikes; 2) Hysteresis is added to the alarm clearing condition: the clearing condition requires the score to drop below a certain level. And continue for a period of time To prevent repeated circuit breaker tripping due to alarms.

[0085] In this embodiment, the linkage control action is defined as:

[0086] Actions during Level 2 warning (automatic power reduction): Reduce power according to preset steps. in And record the execution reason and timestamp. If power is reduced... exist If the temperature drops to a safe level, normal operation will resume; otherwise, a Level 3 warning status will be activated.

[0087] The actions for a Level 3 alert (emergency) shall be executed in the following order:

[0088] 1) Automatically send a shutdown command to the inverter / power supply;

[0089] 2) Open the backup cooling circuit or increase the flow rate;

[0090] 3) Cut off any suspicious operations that may affect operator safety;

[0091] 4) Send alarms through multiple channels (local audio and visual alarms, SMS, email, SCADA alarms);

[0092] 5) Save at least the most recent 30 minutes of logs, raw high-frequency data snapshots, and generate event reports.

[0093] All actions must be consistent with the factory safety management / PLC interlock rules. The permissions and levels of automatic actions can be set in the system configuration (for example, during the trial operation period, automatic actions in the third-level warning state are turned off, and only actions in the prompt / second-level warning state are retained).

[0094] In this embodiment, log, evidence preservation, and tracing are handled as follows:

[0095] Each warning event must be recorded and saved with: trigger timestamp, raw time series of all sensors (covering at least 30 minutes before and after the event), thermal image frames (keyframes), and determining parameter values ​​( , , , The log should include records of actions taken, results taken, and manual interventions. Logs should be kept for the period stipulated by company regulations (recommended at least 3 years) and should be exportable.

[0096] In this embodiment, security and exception handling:

[0097] When a sensor malfunctions (disconnection, over-range measurement, excessive signal noise), the system immediately enters a "reduced trust" mode: automatic action privileges are reduced, manual intervention is prompted, and redundant sensors / estimation methods are used for compensation (e.g., using coolant and power as alternative signals). The system must provide a "manual reset" option, requiring confirmation from two people or a supervisor before the three-level interlock can be released to ensure safety and reliability.

[0098] Working Principle Explanation: The working principle of this invention is to establish a model of the correspondence between power input and heating time during vacuum suspension melting, predict the temperature rise trend of the crucible surface under normal operating conditions, and analyze the deviation by comparing the real-time measured surface temperature and coolant temperature, thereby indirectly inferring the thermal state of the upper solidified shell and the lower melt. When melting begins, the system calculates the theoretical change curve of the surface temperature according to the set power-time model; under normal circumstances, the surface temperature should show a stable upward trend with power accumulation and heating time. When the actual collected surface temperature is lower than the expected temperature rise curve, and the coolant outlet temperature continues to rise, it indicates that the input heat has not been effectively conducted to the surface layer, possibly due to the formation of a solidified shell above or overheating of the lower melt, leading to obstructed heat transfer. The system calculates the dynamic deviation between the surface temperature and the expected temperature, and between the coolant temperature and the theoretical temperature difference, and accumulates and integrates the deviation over time to reflect the persistence of the anomaly. This accumulated deviation index can effectively suppress the interference of instantaneous noise or short-term fluctuations, ensuring stable and reliable early warning judgments. When the cumulative deviation exceeds the set threshold, the system automatically triggers an early warning signal, reminding operators to take intervention measures such as reducing power, extending heating time, or checking the cooling system to prevent safety risks such as localized overheating of the lower part, crucible erosion, or explosion caused by the upper shell solidification. This method eliminates the need for direct measurement of the melt temperature inside the furnace, enabling non-contact judgment of internal thermal anomalies by utilizing the coupling relationship between power input and surface response. It boasts advantages such as simple structure, rapid response, and high reliability, making it particularly suitable for intelligent thermal safety early warning in high-temperature, enclosed environments such as vacuum levitation melting furnaces.

[0099] Example 2

[0100] This embodiment discloses an overheating early warning method for the solidified shell of a vacuum suspension melting system. The difference between this embodiment and Embodiment 1 is the power-time-temperature prediction model.

[0101] Specifically, the power-time-temperature prediction model can be established through mechanistic modeling. This method is based on the principle of energy conservation, treating the input electromagnetic power as the main heat source of the system. By calculating the effective absorbed energy per unit time, a theoretical temperature rise model for the melt and surface is established. Specifically, given the coil power... Electromagnetic coupling efficiency melt quality Under the condition of specific heat capacity c, the change of temperature with time can be described by the formula, i.e. ;

[0102] in, The heat dissipation term varies with temperature, including radiative heat loss, conduction to the crucible, and convective losses. The heat dissipation term can be established as a function based on the Stefan-Boltzmann radiation law, the thermal conductivity equation, and the coolant temperature, ensuring that the model satisfies thermodynamic laws while reflecting the loss characteristics of the actual melting structure.

[0103] By numerically solving the above differential equations, the theoretically expected curve between power input and melt temperature can be obtained. For different charge amounts, materials, and geometries, model parameters such as density, specific heat, emissivity, and electromagnetic absorption efficiency can be adjusted to obtain adaptive temperature rise prediction results. This mechanistic model can provide stable and reliable temperature predictions even without a large amount of historical data, and is another feasible implementation of the power-time model described in this invention.

[0104] Specifically, the heating time determines the overall temperature rise of the melt. This invention utilizes the actual heating time and power to construct the expected temperature rise curve. If the surface temperature does not rise accordingly within a sufficient heating time, it indicates an abnormality in the heat transfer path or that the solidified shell has not yet melted. This criterion avoids misjudgments caused by instantaneous power fluctuations and improves the stability of temperature rise assessment.

[0105] Specifically, surface temperature is a key signal for determining the state of condensation. This invention collects surface temperature in real time and compares it with model predictions. When the actual temperature is consistently lower than the expected temperature rise, it indicates surface obstruction or intensified internal temperature stratification. If the coolant temperature difference also increases simultaneously, it more reliably indicates overheating below. Surface temperature is the most direct basis for identifying temperature anomalies.

[0106] Specifically, melt height reflects the morphological changes of the melt under the influence of electromagnetic force. This invention identifies whether a solidified crust has formed by the trend of height changes: if the height fluctuation weakens, remains fixed in an abnormal position, or does not match the power change, it usually indicates surface solidification or uneven internal heating. The height signal can effectively supplement the temperature criterion and improve the ability to identify anomalies.

[0107] Specifically, thermal imaging provides a two-dimensional temperature distribution on the surface. This invention utilizes features such as temperature field uniformity, crack highlights, and localized cold areas to identify the thickness of the condensed crust and internal temperature shifts. When the overall temperature is relatively cool but localized cracks are bright, it often indicates that the internal high temperature is being obscured; if the temperature field non-uniformity worsens, it is often related to the growth of the condensed crust. Thermal imaging enhances the system's spatial identification capability of abnormal thermal states.

[0108] In summary, the above scheme constructs a expected temperature rise model based on the power-time relationship. The theoretical temperature rise trend of the melt surface under normal operating conditions is established by using a thermal equivalent model, energy accumulation model, or empirical fitting model. This enables the system to determine whether the "input energy - surface temperature response" matches, laying the foundation for subsequent anomaly identification.

[0109] The above scheme defines the surface temperature deviation. and coolant temperature difference deviation The dual-parameter judgment mechanism, by simultaneously monitoring abnormal surface temperature rise (surface is colder than expected) and abnormal cooling load (coolant temperature difference is higher), breaks through the limitations of single temperature monitoring and enables the identification of comprehensive anomalies such as thermal blockage, condensation sealing, and overheating below.

[0110] By introducing the weighted cumulative deviation index through the above scheme, As the core anomaly metric, the instantaneous deviation is accumulated over time using a sliding window integral and a weighting function to achieve a comprehensive evaluation of "deviation amplitude + deviation duration," avoiding noise interference and better reflecting the physical characteristics of the gradual accumulation of thermal anomalies.

[0111] The above scheme constructs a comprehensive risk scoring and early warning classification logic. By combining normalized deviation indicators, deviation weights, and change rates, a quantifiable risk score is constructed. Three alarm levels are set: Level 1 alert, Level 2 automatic intervention, and Level 3 interlock protection, effectively reducing the probability of escalation of danger.

[0112] By implementing the above scheme, a closed-loop early warning system is established, including power reduction, delayed observation, and emergency shutdown interlock. The system not only issues alarms but also automatically performs operations such as power reduction, enhanced cooling, or power cut-off, making the early warning system have a substantial protective effect and significantly improving the safety and reliability of vacuum melting equipment.

[0113] The above scheme employs robustness-enhancing strategies such as time jitter reduction, hysteresis detection, and data consistency verification to ensure stable operation of the system even under conditions of high-frequency electromagnetic environment, temperature measurement noise, and cooling system fluctuations, thereby improving the reliability of practical industrial applications.

[0114] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of equivalents of the claims be included within the present invention.

[0115] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style of the specification is merely for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. A method for early warning of overheating of the solidified shell in a vacuum suspension melting system, characterized in that, The overheat warning method includes: S1: Establish a power-time-temperature prediction model to obtain a dataset of predicted surface temperatures for the crucible / melt; S2: Real-time acquisition of crucible / melt surface temperature, obtaining actual surface temperature dataset of crucible / melt, and obtaining instantaneous surface temperature difference deviation dataset by predicting surface temperature dataset and actual surface temperature dataset; S3: Establish a coolant temperature rise prediction model, collect heating power and coolant mass flow rate in real time, and obtain a cooling circuit temperature difference deviation dataset; S4: Normalize the collected instantaneous surface temperature difference deviation dataset and the cooling circuit temperature difference deviation dataset to obtain a comprehensive instantaneous risk score; and perform cumulative anomaly index and time weighting to obtain cumulative anomaly index. S5: The vacuum suspension melting system enters a Level 1 early warning state based on the comprehensive instantaneous risk score and cumulative abnormal indicators according to the first rule; Alternatively, the vacuum suspension melting system may enter a level-two early warning state based on the comprehensive instantaneous risk score and cumulative abnormal indicators through the second rule; Alternatively, the vacuum suspension melting system may enter a level three early warning state based on the comprehensive instantaneous risk score and cumulative abnormal indicators through the third rule.

2. The overheating early warning method for the solidified shell in a vacuum suspension melting system as described in claim 1, characterized in that, In step S1, establishing the power-time-temperature prediction model includes: collecting historical operating data of the vacuum suspension melting furnace under different charging amounts, different materials, and different process parameters; synchronizing the historical operating data according to a unified time benchmark; removing, smoothing, or filtering outliers; constructing an input sequence reflecting the power-time behavior characteristics according to a preset time window; using the surface temperature of the crucible / melt at each time point as the model output target; establishing a power-time-temperature mapping model for time series prediction; and fitting the model parameters by minimizing the error between the predicted temperature and the actual measured temperature. The historical operating data includes: heating power, cumulative energy, heating duration, coolant inlet temperature, coolant outlet temperature, and crucible / melt surface temperature as of time.

3. The overheating early warning method for the solidified shell in a vacuum suspension melting system as described in claim 2, characterized in that, In step S1, heating power and heating time are collected in real time to obtain the predicted surface temperature dataset of the crucible / melt.

4. The overheating early warning method for the solidified shell in a vacuum suspension melting system as described in claim 1, characterized in that, In step S1, establishing the power-time-temperature prediction model includes: treating the input coil power as the input heat source of the vacuum levitation melting system, and establishing the theoretical temperature rise relationship between the melt and surface temperatures by calculating the effective absorbed energy per unit time. In the formula, For real-time input power to the coil, For electromagnetic coupling efficiency, Where c is the melt mass and c is the specific heat capacity. To predict the rate of change of surface temperature, This refers to the heat dissipation component that changes with temperature.

5. The overheating early warning method for the solidified shell in a vacuum suspension melting system as described in claim 4, characterized in that, In step S1, the heating time is collected in real time to obtain the predicted surface temperature dataset of the crucible / melt.

6. The overheating early warning method for the solidified shell in a vacuum suspension melting system as described in claim 1, characterized in that, In step S3, establishing the coolant temperature rise prediction model includes: treating the input coil power as the input heat source of the vacuum suspension melting system, and establishing the temperature difference deviation relationship of the coolant cooling circuit, i.e. In the formula, For real-time input power to the coil, Where cp is the coolant mass flow rate, and cp is the coolant's specific heat capacity at constant pressure. For the temperature difference deviation in the cooling circuit, This is the effective heat transfer ratio parameter.

7. The overheating early warning method for the solidified shell in a vacuum suspension melting system as described in claim 1, characterized in that, In step S4, the normalization process includes: Define the normalized instantaneous score of the instantaneous surface temperature difference deviation dataset, i.e. In the formula, As the first reference standard, This refers to the instantaneous surface temperature difference deviation. This is the first normalized instantaneous score; Define the normalized instantaneous score of the cooling loop temperature difference deviation dataset, i.e. In the formula, As the second reference scale, For the temperature difference deviation in the cooling circuit, This is the second normalized instantaneous score; Define the comprehensive instantaneous risk score, i.e. In the formula, This is the first initial recommended value. This is the second initial recommended value. To calculate the comprehensive instantaneous risk score.

8. The overheating early warning method for the solidified shell in a vacuum suspension melting system as described in claim 7, characterized in that, In step S4, the cumulative anomaly index and time weighting include: Weighted accumulation is used, i.e. In the formula, Here, W is the time weighting function, and W is the cumulative window length. For cumulative abnormal indicators; To achieve discretization, i.e. In the formula, , The sampling interval is denoted as .

9. The overheating early warning method for the solidified shell in a vacuum suspension melting system as described in claim 8, characterized in that, The first rule is: and ;in, The first-level alarm threshold, Set a threshold for the first one; The vacuum suspension melting system enters a first-level early warning state, which includes: interface prompts, event recording, and recommendations for manual inspection.

10. The overheating early warning method for the solidified shell in a vacuum suspension melting system as described in claim 8, characterized in that, The second rule is: ;or, and Exceeding the second set threshold; where, The first-level alarm threshold, This is the second-level alarm threshold. This refers to the outlet temperature of the coolant. This refers to the inlet temperature of the coolant. The vacuum suspension melting system enters a level 2 early warning state, which means that the heating power is automatically reduced to the next power level according to a preset ratio, the steady-state observation time is extended, and at the same time, sound or light alarms and SMS or remote notifications are triggered, and a snapshot is recorded.

11. The overheating early warning method for the solidified shell in a vacuum suspension melting system as described in claim 8, characterized in that, The third rule is: Or, instantaneously and Continued to rise; among them, This is the third-level alarm threshold. This is the critical threshold for the instantaneous change in temperature of the melt or crucible surface. This refers to the outlet temperature of the coolant. This refers to the inlet temperature of the coolant. The vacuum suspension melting system enters a Level 3 early warning state, which means: immediately disconnecting the heating power supply, opening the emergency cooling circuit, recording all historical data, and notifying maintenance and on-duty management personnel.

Images