Copper melting furnace operation monitoring system based on industrial internet

Abstract

The present application relates to the technical field of industrial internet and industrial process monitoring, in particular to a copper melting furnace operation monitoring system based on industrial internet, comprising field sensors, edge devices and cloud platforms; through a multi-source data acquisition module, the hearth temperature, the furnace wall temperature, the exhaust gas temperature, the heat input parameter, the heat loss parameter and the process operation action and network transmission information are collected; through a process action recording module, the feeding, the furnace door state change, the electric power change, the gas flow change and the copper output action are converted into a process action sequence; then through data time synchronization, abnormal correction, furnace state calculation, working condition and abnormality determination, energy consumption evaluation and operation and maintenance output, and by the division of labor of the edge and the platform, the furnace condition monitoring, abnormal early warning and energy consumption evaluation are realized.

Description

Technical Field

[0001] This invention relates to the field of industrial internet and industrial process monitoring technology, specifically to a copper smelting furnace operation monitoring system based on industrial internet. Background Technology

[0002] During the daily operation of a copper melting furnace, the furnace temperature, furnace wall heating, flue gas status, and processes such as charging and copper tapping will continuously change. The stability of the smelting process, the timeliness of anomaly identification, and the unit energy consumption level depend on the ability to continuously monitor and comprehensively analyze furnace operation information. In traditional methods, monitoring the operation of copper melting furnaces often relies on a single temperature measuring point, manual inspection records, or simple data uploads. There is a lack of unified handling for sampling times, network transmission delays, and thermal response lags between different measuring points. When sensor drift, smoke and dust obstruction, data loss, or changes in operating conditions occur, the accuracy of furnace condition judgment remains low. Summary of the Invention

[0003] The purpose of this invention is to provide a copper melting furnace operation monitoring system based on the Industrial Internet, which avoids the problem of low furnace condition judgment when sensor drift, smoke and dust blockage occurs due to the lack of unified processing of sampling time, network transmission delay and thermal response lag of different measuring points in traditional methods. It also makes it easier to realize physical time sequence alignment of multi-source heterogeneous data and dynamic repair of distorted data, thereby accurately identifying the operating stage and anomaly type and completing energy consumption assessment.

[0004] The objective of this invention can be achieved through the following technical solutions: The copper melting furnace operation monitoring system based on the Industrial Internet includes field sensors, edge devices, and a cloud platform: The multi-source data acquisition module is deployed on edge devices to collect information transmitted through the network, process operation actions, and measurement data including at least furnace temperature, furnace wall temperature, flue gas temperature, heat input and heat loss parameters through field sensors and external input. The process action recording module generates a process action sequence based on the process operation actions; The data time synchronization module corrects the sampling time and response delay of the measuring points according to the network transmission information and process action sequence, and maps the measurement data to the same preset monitoring cycle; The data anomaly correction module performs validity calculations and deviation corrections on the synchronized measurement data. The furnace state calculation module calculates the equivalent temperature of the furnace, the equivalent temperature of the molten copper, the furnace lining heat storage index, the energy balance index, and the molten pool disturbance index based on the preset relationship between the above temperature and heat parameters. The operating condition and anomaly determination module determines the operating condition stage and anomaly type based on the furnace equivalent temperature, copper liquid equivalent temperature, furnace lining heat storage index, energy balance index and molten pool disturbance index calculated above. The energy consumption assessment and operation and maintenance output module outputs operation and maintenance information including furnace status, abnormal warnings, energy consumption assessment results and maintenance suggestions, measurement point reliability, abnormal heat loss or stage switching prompts; The edge and platform division of labor execution modules will assign computational tasks from data collection to anomaly correction to edge devices, and computational tasks from status calculation to operation and maintenance output to the cloud platform.

[0005] Furthermore, the multi-source data acquisition module collects data through field sensors including furnace infrared temperature, furnace wall contact temperature, copper molten contact temperature, flue gas temperature, ambient temperature, furnace pressure, flue gas oxygen content, gas flow rate, electric power, furnace door opening and closing status, feeding action, copper tapping action, and molten pool level data, among which gas flow rate and electric power are included at least one. It also receives externally input network latency, packet loss rate, jitter information, low calorific value of gas, initial furnace material quality, feeding quality, copper output quality, and loss quality through edge devices; For furnace types without gas heating, the gas flow rate is set to zero; for furnace types without electric heating, the electric power is set to zero. Each data item is accompanied by a sampling timestamp, upload timestamp, source device identifier, measurement point identifier, process location identifier, unit identifier, and data validity identifier.

[0006] Furthermore, the process action recording module, according to a unified preset monitoring cycle, converts feeding actions, furnace door opening and closing status, electrical power, gas flow rate, and copper tapping actions into a process action sequence using binary events, hierarchical coded values, or normalized changes. The process action sequence contains the first... The sequence value for each monitoring cycle is equal to the sum of the products of the feeding action, furnace door opening / closing status, electrical power, gas flow rate, and copper tapping action within that cycle, respectively, and the corresponding first to fifth preset weighting coefficients. It is a positive integer representing the monitoring cycle number.

[0007] Furthermore, the data time synchronization module determines the network latency based on the difference between the upload timestamp and the sampling timestamp, and calculates the adjacent period difference of each measurement point data and the response degree of the process action sequence under different candidate delays within a preset delay lower limit to a preset delay upper limit and within a preset length sliding window. The response degree is equal to the sum of the absolute values ​​of the products of the adjacent period difference at each time point within the sliding window and the sequence value of the process action sequence. The candidate delay that maximizes the response level is determined as the measurement point response delay. The measurement point data is then mapped to the time corresponding to the current physical time minus the measurement point response delay using an interpolation operator. Linear interpolation is used for continuous measurement signals, while nearest neighbor mapping is used for discrete event signals.

[0008] Furthermore, the data anomaly correction module calculates the validity score of each measuring point. The validity score is equal to the sum of the first to fourth score components multiplied by their corresponding first to fourth evaluation weight coefficients. The first score component is determined by missing points, saturation points, and jump points exceeding the preset unit time change limit within the sliding window. The second score component is determined by upload jitter and packet loss rate. The third score component is determined by the deviation between the synchronized measurement value and the verification reference value calculated based on other related measuring points and the preset heat balance relationship. The fourth score component is determined by the degree of fluctuation after removing the change trend. The sum of the four weight coefficients is one. The data anomaly correction module updates the measurement point deviation. The measurement point deviation in the current period is equal to the measurement point deviation in the previous period multiplied by the forgetting factor, plus the difference between the synchronized measurement value and the verification reference value multiplied by one minus the forgetting factor. The forgetting factor is greater than or equal to zero and less than one. The corrected value after deviation correction is obtained by subtracting the measurement point deviation of the current cycle from the synchronized measurement value; if the validity score is less than the set lower limit, the measurement value of the corresponding measurement point is replaced by the verification reference value; otherwise, the corrected value is used.

[0009] Furthermore, the furnace state calculation module calculates the equivalent temperature of the furnace, the equivalent temperature of the molten copper, the furnace lining heat storage index, the energy balance index, and the molten pool disturbance index based on the correction values ​​of each measuring point and the corresponding validity score.

[0010] Furthermore, the equivalent furnace temperature is calculated by weighting and summing the corrected furnace infrared temperature, furnace wall contact temperature, flue gas temperature, and heat input equivalent temperature according to their respective effectiveness scores. Heat input equivalent temperature equals gas flow rate, gas lower heating value, and current monitoring cycle duration. Multiply the product of these three factors by the first conversion factor, and then add the power and the current monitoring cycle duration. The product is then multiplied by the second conversion factor; The compensation amount from flue gas temperature to furnace temperature is equal to the third conversion base, plus the product of the energy balance index and the fourth conversion coefficient, plus the product of the furnace door opening / closing status and the fifth conversion coefficient. The equivalent temperature of molten copper is equal to the corrected contact temperature of molten copper multiplied by the first normalization weight, plus the difference between the equivalent temperature of the furnace and the compensation amount multiplied by the second normalization weight, plus the product of the furnace lining heat storage index and the corrected contact temperature of the furnace wall multiplied by the third normalization weight, and then minus the product of the molten pool disturbance index and the preset liquid surface disturbance temperature drop constant multiplied by the fourth normalization weight. The sum of the first to fourth normalization weights is one, and they are determined by normalization based on the effectiveness scores of the copper liquid contact measurement points, furnace measurement points, furnace wall measurement points, and molten pool disturbance information, respectively.

[0011] Furthermore, the furnace lining heat storage index is derived from the furnace lining heat storage index of the previous monitoring period, the temperature difference between the furnace and the furnace wall, the heat input per unit time, and the heat loss per unit time. The heat loss per unit time is determined based on the difference between the flue gas temperature and the ambient temperature, the furnace door opening and closing status, and the deviation of the furnace pressure from the preset staged reference furnace pressure. The energy balance index is calculated based on the deviation of flue gas oxygen content from the preset staged reference values, the deviation of furnace pressure from the staged reference furnace pressure, and the deviation of flue gas temperature from the preset staged reference values. The molten pool disturbance index is calculated by normalizing and weighting the scores based on the variance of furnace infrared temperature within the sliding window, the change in molten pool surface data, and the number of feeding actions.

[0012] Furthermore, the operating condition and anomaly judgment module divides the copper melting furnace operating condition into the charging disturbance stage, rapid heating stage, melting and homogenization stage, holding and waiting for copper tapping stage, copper tapping stage, or abnormal fluctuation stage based on the process action sequence, furnace equivalent temperature change rate, copper liquid equivalent temperature change rate, furnace lining heat storage index, energy balance index, and molten pool disturbance index. If the feeding action component in the process action sequence is greater than or equal to the first threshold and the molten pool disturbance index is greater than or equal to the second threshold, then the current working condition stage is determined to be the feeding disturbance stage. If the decrease in furnace infrared temperature is greater than the lower limit of the occlusion judgment set based on the historical statistical values ​​of each working condition stage, and the change in furnace equivalent temperature and flue gas temperature is not greater than their respective upper limit of fluctuation, and the effectiveness score of the furnace infrared temperature corresponding to the measuring point is less than the set lower limit of the score, then an infrared temperature measurement distortion abnormality prompt message will be generated. The operating condition and anomaly judgment module is also used to set the stage upper limit threshold and duration based on the historical statistical values ​​of each operating condition stage. If the corresponding upper limit threshold is exceeded and the duration is longer than the set time, an anomaly prompt message is generated for overheating, energy supply imbalance, excessive heat storage in the furnace lining, or instability of the molten pool.

[0013] Furthermore, the energy consumption assessment and operation and maintenance output module calculates the input energy within the statistical period based on the gas flow rate, the lower heating value of the gas, the electric power, and the preset gas utilization coefficient and electric energy utilization coefficient. It calculates the effective smelting quality based on the initial furnace material quality, the charging quality, the copper output quality, and the loss quality at the beginning of the period, and uses the ratio of input energy to effective smelting quality as the unit smelting energy consumption. The edge device and platform division of labor execution module complete data access, timestamp processing, unified periodic resampling, process action sequence generation, and validity calculation at the edge device, and output a rapid warning when the validity score of each measuring point is less than the set lower limit, the furnace pressure exceeds the limit, or the temperature drops suddenly; The cloud platform completes the search range for measurement point response delay, state calculation coefficient, staged reference values ​​for flue gas temperature and furnace pressure, staged reference values ​​and threshold rolling correction, as well as energy consumption assessment and operation and maintenance information output. The cloud platform completes the search range for measurement point response delay, state calculation coefficient, staged reference values ​​for flue gas temperature and furnace pressure for each operating condition, rolling correction of thresholds, energy consumption assessment, and operation and maintenance information output.

[0014] The beneficial effects of this invention are: 1. This invention uses a data time synchronization module to correct the sampling time and response delay of measurement points based on network transmission information and process action sequence, and maps the measurement data to the same preset monitoring cycle. This solution eliminates the misalignment of heterogeneous data in network transmission and thermal response, and transforms scattered data that is easily confused by network fluctuations into a time-series stream with clear traceability conditions, thus solving the problem of joint analysis failure caused by the lack of unified time-series processing in the existing technology. 2. This invention uses a data anomaly correction module to calculate an effectiveness score by combining missing measurement points, jump points, network transmission jitter, deviation of verification reference values, and the degree of data fluctuation. It also uses a forgetting factor to progressively update the measurement point deviation for correction or replacement. This mechanism can identify and compensate for local data distortion caused by smoke and dust obstruction or high-temperature thermal drift, avoiding frequent false alarms caused by abnormal data directly dominating the overall judgment under the traditional single temperature measurement method. 3. This invention integrates multi-dimensional physical parameters through an in-furnace state calculation module to reconstruct multiple state indices, such as the equivalent temperature of the furnace chamber, the equivalent temperature of the molten copper, and the heat storage index of the furnace lining. The operating condition and anomaly determination module then uses these indices in conjunction with the process action sequence to determine the current operating condition stage and anomaly type. This design extracts and quantifies the operation excitation signal, and based on physical consistency and multi-dimensional state characteristics, it solves the problems of overgeneralization and insensitive identification of operating condition switching caused by traditional monitoring that relies solely on instantaneous temperature assessment. 4. This invention relies on the division of labor between the edge device and the platform to execute modules. High-time-sensitive calculation tasks such as data acquisition, action coding, and initial calculation of sudden temperature drops are performed by edge devices, while complex calculations such as long-cycle state reconstruction, energy consumption assessment, and rolling correction of reference thresholds are performed by the cloud platform. This architecture ensures that early warnings can be issued quickly when critical indicators appear in industrial equipment on site, while continuously updating model parameters adaptively using historical patterns, forming a complete closed loop for industrial Internet monitoring. Attached Figure Description

[0015] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a structural block diagram of the copper melting furnace operation monitoring system based on the Industrial Internet in this application embodiment. Detailed Implementation

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

[0017] Please see Figure 1 An industrial internet-based copper melting furnace operation monitoring system includes field sensors, edge devices, and a cloud platform: The multi-source data acquisition module is deployed on edge devices to collect information transmitted through the network, process operation actions, and measurement data including at least furnace temperature, furnace wall temperature, flue gas temperature, heat input and heat loss parameters through field sensors and external input. The process action recording module generates a process action sequence based on the process operation actions; The data time synchronization module corrects the sampling time and response delay of the measuring points according to the network transmission information and process action sequence, and maps the measurement data to the same preset monitoring cycle; The data anomaly correction module performs validity calculations and deviation corrections on the synchronized measurement data. The furnace state calculation module calculates the equivalent temperature of the furnace, the equivalent temperature of the molten copper, the furnace lining heat storage index, the energy balance index, and the molten pool disturbance index based on the preset relationship between the above temperature and heat parameters. The operating condition and anomaly determination module determines the operating condition stage and anomaly type based on the furnace equivalent temperature, copper liquid equivalent temperature, furnace lining heat storage index, energy balance index and molten pool disturbance index calculated above. The energy consumption assessment and operation and maintenance output module outputs operation and maintenance information including furnace status, abnormal warnings, energy consumption assessment results and maintenance suggestions, measurement point reliability, abnormal heat loss or stage switching prompts; The edge and platform division of labor execution modules will assign computational tasks from data collection to anomaly correction to edge devices, and computational tasks from status calculation to operation and maintenance output to the cloud platform. As a preferred embodiment, the multi-source data acquisition module collects data from field sensors including furnace infrared temperature, furnace wall contact temperature, copper molten contact temperature, flue gas temperature, ambient temperature, furnace pressure, flue gas oxygen content, gas flow rate, electric power, furnace door opening and closing status, feeding action, copper tapping action, and molten pool level data, wherein at least one of gas flow rate and electric power is included. It also receives externally input network latency, packet loss rate, jitter information, low calorific value of gas, initial furnace material quality, feeding quality, copper output quality, and loss quality through edge devices; For furnace types without gas heating, the gas flow rate is set to zero; for furnace types without electric heating, the electric power is set to zero. Each data item includes a sampling timestamp, upload timestamp, source device identifier, measuring point identifier, process location identifier, unit identifier, and data validity identifier. The process action recording module converts feeding actions, furnace door opening and closing status, electrical power, gas flow rate, and copper tapping actions into process action sequences using binary events, hierarchical coded values, or normalized changes, according to a unified preset monitoring cycle. The process action sequence contains the first... The sequence value for each monitoring cycle is equal to the sum of the products of the feeding action, furnace door opening / closing status, electrical power, gas flow rate, and copper tapping action within that cycle, respectively, and the corresponding first to fifth preset weighting coefficients. It is a positive integer representing the monitoring cycle number.

[0018] In this embodiment, a progressive processing link is constructed in the overall system architecture, from bottom-level data acquisition to high-level state assessment; the system acquires raw measurement and externally entered data with complete metadata through a multi-source data acquisition module; The discrete operation events are extracted using the process action recording module as the reference anchor points for time alignment; the misalignment of heterogeneous data in network transmission and thermal response is eliminated through the data time synchronization module and the data anomaly correction module, and distorted data is dynamically repaired. Based on this, the furnace condition calculation module integrates multi-dimensional parameters to reconstruct various indices that reflect the actual furnace condition; finally, the working condition and anomaly judgment module and the energy consumption assessment and operation and maintenance output module realize the identification of the current production stage, anomaly prompts, and unit smelting energy consumption analysis. The system relies on edge devices and platform-based modules to perform high-frequency, time-sensitive tasks on edge devices and complex fusion computing on cloud platforms, thereby achieving a corresponding allocation of computing tasks and device processing capabilities. Furthermore, based on the above system architecture, the data source acquisition and preprocessing stages are described in detail; continuous physical quantities and discrete state quantities collected by field sensors are acquired, network and process configuration parameters are received according to the external input interface of the edge device, and metadata including timestamps and location identifiers are added to each data to generate a standardized dataset. The data collected by the on-site sensors include furnace infrared temperature, furnace wall contact temperature, copper molten contact temperature, flue gas temperature, ambient temperature, furnace pressure, flue gas oxygen content, gas flow rate, electric power, furnace door opening and closing status, feeding action, copper tapping action, and molten pool level data. The data recorded externally by the edge device includes network latency, packet loss rate, jitter information, low calorific value of gas, initial furnace material quality, feeding quality, copper output quality, and loss quality. After entering the unified data queue, both types of data retain the sampling timestamp, upload timestamp, source device identifier, measurement point identifier, process location identifier, unit identifier, and data validity identifier to avoid mixing up the real-time sensor data with the data entered manually or by the system. To address the concurrent processing requirements of heterogeneous data, the unified data queue adopts a ring buffer architecture with time-to-live tags. When underlying data is enqueued, a double-key binding between the spatial position pointer and the time tag is automatically triggered. When instantaneous data concurrency causes the cache level to exceed a set safety threshold, the system activates a priority-based backpressure current limiting mechanism. This ensures that high-priority discrete process events, such as copper ejection, are queued first, while low-priority environmental parameters are smoothly downsampled. Specifically, this is achieved by reducing the sampling frequency of low-priority data by a preset ratio and averaging and compressing redundant data within the downsampled range, thereby ensuring the determinism of the data source. This embodiment binds sampling timestamps, upload timestamps, and source device identifiers to retain complete transmission and spatial mapping attributes at the data source head, transforming scattered data that is easily confused due to network fluctuations into a time-series stream with clear traceability conditions. As a result, a standardized access mechanism for heterogeneous data can be achieved based on the fusion acquisition mechanism of multidimensional metadata, avoiding spatial misalignment and temporal confusion in subsequent calculations, and solving the problem that the existing technology has a weak foundation for fusion analysis due to the lack of unified metadata management. After data acquisition is completed, discrete process operation actions are obtained. Based on the preset event impact priority, various actions are quantified and encoded to generate a process action sequence. Among them, the first to fifth preset weight coefficients refer to the quantitative characterization of the disturbance intensity of various process operations on the thermal state inside the furnace. The calibration basis for the first to fifth preset weight coefficients is as follows: during the equipment no-load or historical calibration test phase, a single process action is applied independently, and the maximum temperature change slope caused by the action is recorded. The maximum temperature change slope corresponding to each action is divided by the highest slope value among all test actions for normalization. This is used as the weight coefficient of each process action, thereby objectively reflecting the disturbance intensity of each action on the furnace condition. For example, the weighting coefficient for the feeding action can be set to 1.0, the weighting coefficient for the change of furnace door status can be set to 0.7, and the weighting coefficient for the fine adjustment of gas flow can be set to 0.5. Meanwhile, binary events refer to open and closed state operations, such as state triggering being recorded as 1 for opening and 0 for closing; hierarchical coding values ​​are used to characterize the physical level of action intensity, such as the material feeding quality being divided into three levels: light, medium and heavy, recorded as 0.3, 0.6 and 1.0 respectively; The normalized change is used for dimensionless transformation in continuous control operations and is expressed as the ratio of the absolute value of the current control to the maximum rated change limit. The above-mentioned coded values ​​are used to represent the intensity or magnitude of operational disturbances, rather than directly representing the direction of heating or cooling. For actions with opposite physical directions, such as furnace door closing or gas flow rate reduction, the system retains the corresponding state direction or change direction field while retaining the disturbance intensity value. Subsequently, the thermal effect is determined by the difference between adjacent periods of the measuring point, the equivalent temperature change rate of the furnace, and the equivalent temperature change rate of the copper liquid. The judgment condition is as follows: The system compares the sign of the operation status direction field with the equivalent temperature change rate of the measuring point in real time. If the process action includes a negative state direction, such as the cooling excitation of gas reduction, and the furnace equivalent temperature change rate is greater than zero in the synchronous cycle, it is determined that the thermal effect of the negative action is delayed or masked by other heat sources. At this time, the negative action is not simply linearly superimposed with the positive heating effect. Instead, the thermal effect weight of the action in the current cycle is reduced proportionally or reset to zero until the sign of the temperature change rate is consistent with the action direction. Then the response association is restored to avoid simply superimposing process actions with opposite directions into the same thermal effect. When a proportional reduction is determined, the system executes an exponential decay function in the control flow based on the absolute value of the deviation of the equivalent temperature change rate: using the natural constant... Based on the base, the attenuation ratio is dynamically generated by multiplying the preset attenuation constant by the negative exponent of the absolute value of the deviation and applied to the thermal effect weight of the current cycle. The preset attenuation constant is set based on its reciprocal, which reflects the furnace body's tolerance for the thermal effect in the opposite direction of the operation. The reciprocal is calibrated by extracting the average temperature change deviation value when such an operation is completely covered in the previous working condition records. This mechanism ensures that all heterogeneous discrete operations and continuous adjustment variables are calculated within a unified dimensionless interval. This embodiment extracts high-priority operation events and performs weighted integration to capture the operation excitation signals that drive furnace condition switching, transforming the originally difficult-to-quantify manual operation variables into calculable process anchor points; Therefore, the serialized representation of process events can be achieved based on the weighted encoding of operational actions, avoiding the time lag caused by simply relying on temperature response, and solving the problem that existing technologies are not sensitive to changes in operating conditions due to the lack of prior information on operational stimuli.

[0019] In a preferred embodiment, the data time synchronization module determines the network latency based on the difference between the upload timestamp and the sampling timestamp, and calculates the adjacent period difference of each measurement point data and the response degree of the process action sequence under different candidate delays within a preset delay lower limit to a preset delay upper limit and within a preset length sliding window. The response degree is equal to the sum of the absolute values ​​of the products of the adjacent period difference at each time point within the sliding window and the sequence value of the process action sequence. The candidate delay that maximizes the response level is determined as the measurement point response delay, and the measurement point data is mapped to the time corresponding to the current physical time minus the measurement point response delay through an interpolation operator. Linear interpolation is used for continuous measurement signals, and nearest neighbor mapping is used for discrete event signals. The data anomaly correction module calculates the validity score of each measuring point. The validity score is equal to the sum of the first to fourth score components multiplied by their corresponding first to fourth evaluation weight coefficients. The first score component is determined by missing points, saturation points, and jump points exceeding the preset unit time change limit within the sliding window. The second score component is determined by upload jitter and packet loss rate. The third score component is determined by the deviation between the synchronized measurement value and the verification reference value calculated based on other related measuring points and the preset heat balance relationship. The fourth score component is determined by the degree of fluctuation after removing the change trend. The sum of the four weight coefficients is one. The data anomaly correction module updates the measurement point deviation. The measurement point deviation in the current period is equal to the measurement point deviation in the previous period multiplied by the forgetting factor, plus the difference between the synchronized measurement value and the verification reference value multiplied by one minus the forgetting factor. The forgetting factor is greater than or equal to zero and less than one. The corrected value after deviation correction is obtained by subtracting the measurement point deviation of the current period from the synchronized measurement value; if the validity score is less than the set lower limit, the measurement value of the corresponding measurement point is replaced by the verification reference value; otherwise, the corrected value is used.

[0020] In this embodiment, based on the generated process action sequence, the timestamp information of the original data is obtained. According to the network latency and the response degree of the data changes within the sliding window and the process action, the data of each measuring point is dynamically time-off adjusted to generate synchronized data mapped to a unified physical time. During the dynamic time offset adjustment process, if there are consecutive missing values ​​exceeding the preset limit within the sliding window, resulting in a low confidence level in the relevance response search, the system's internal time state machine will trigger the search degradation rule, directly calling and locking the response delay of the previous steady-state cycle as the current physical compensation item, until the data integrity within the sliding window is restored. If the process action recording module detects that the feeding action or the change in the weight of the action sequence exceeds the set working condition switching threshold during the locking period, the delay lock will be forcibly released, and the missing interval will be marked as a strong interference area and will not participate in the dynamic synchronization correction. Furthermore, during the execution of the interpolation operator, the system synchronously checks the time span of adjacent valid sampling points. If the time difference between two valid points exceeds the preset physical thermal inertia safety span, which is set according to the thermal response time constant calibrated by the target measuring point sensor and the preset safety factor, the linear interpolation calculation is terminated, the data in the missing interval is routed to the zero-level hold degradation channel to maintain the previous valid value, and a dense interpolation failure label is attached to the synchronous data to prevent false signal injection caused by unbounded linear extension. Among them, the response degree is used to characterize the correlation between the fluctuation of sensor measurement values ​​and the timing of process actions; the preset delay upper limit refers to the maximum response lag time allowed by the thermal inertia of different types of sensors, for example, the upper limit for infrared temperature measurement is 25 seconds, and the upper limit for furnace wall temperature measurement is 120 seconds. Externally input network latency, packet loss rate, and jitter information are used as a priori criteria for link quality or anomaly scoring, and the difference between the upload timestamp and the sampling timestamp is used as the transmission latency of a single data item. When searching for candidate response delays, the transmission delay is first deducted from the reception time so that the candidate delays only correspond to the thermal response lag of the measurement point, and the communication link delay and the thermal inertia of the furnace body are not confused as the same physical quantity. This embodiment adaptively captures the actual thermal response hysteresis of different measuring points under the current operating conditions through correlation search within a sliding window, transforming the timing mismatch caused by thermal inertia into a synchronous result with traceable physical processes. This enables physical temporal alignment of multi-source data based on event-driven response delay search, avoiding misjudgment of furnace conditions caused by differences in sensor thermal hysteresis, and solving the problem of joint analysis failure caused by temporal misalignment in existing technologies. After completing the timing alignment, the synchronized measurement data and associated network information are obtained. Based on the integrity of the data itself, the quality of network transmission, physical constraints and signal stability, the validity of each measurement point data is quantitatively scored and long-term drift is corrected to determine the correction or replacement value used for subsequent calculations. Among them, the jump point in the first scoring component refers to an abnormal sudden change that exceeds the reasonable physical change limit of the sensor, such as an infrared temperature change of more than 40°C per second. The second scoring component is related to transmission reliability. When the packet loss rate or upload jitter exceeds the limit benchmark, it decreases linearly according to the extent of the exceedance. The third scoring component reflects the consistency of physical mechanisms. The smaller the absolute deviation between the verification reference value calculated based on heat balance and the measured value, the higher the score. The calculation process for the verification reference value is as follows: Taking the infrared temperature measurement point in the furnace as an example, the verification reference value is not the direct addition of heat and temperature, but follows the thermodynamic law of conservation of energy to perform dimensional conversion. Specifically, it is equal to the base ambient temperature plus the input power represented by the product of the gas flow rate and the lower heating value of the gas in the current cycle, multiplied by the current monitoring cycle duration, and then divided by the comprehensive equivalent heat capacity of the current furnace material and furnace lining. It is equal to the sum of the mass of the furnace material multiplied by the specific heat capacity of copper material and the mass of the furnace lining multiplied by the specific heat capacity of the furnace lining. The theoretical heating temperature rise derived from this is then subtracted from the heat loss energy calculated based on the flue gas temperature and divided by the comprehensive equivalent heat capacity to obtain the cooling conversion value. When heat loss is expressed as heat loss power per unit time, it is first multiplied by the current monitoring cycle duration to convert it into heat dissipation energy for that cycle, and then the temperature rise is calculated. In this way, the physical expected boundary of the target measurement point is reconstructed through the parameters associated with the heating and heat dissipation ends, ensuring that the dimensions of power, energy, heat capacity and temperature increment are strictly aligned. For electrical power input, it is converted into electrical energy for the current cycle using the same cumulative method for cycle duration and then used in the calculation of the verification reference value; For furnace types that are not configured with the corresponding heating method, the corresponding input item is zero; the verification reference value is used to characterize the expected temperature change within the current monitoring period, rather than using instantaneous power directly as a temperature quantity; In the specific calculation of the score, each component is normalized and restricted using the corresponding structured logic. Within the interval: For example, the first scoring component is obtained by subtracting 1 from the ratio of the number of outliers (missing points, saturation points, and jump points) within the sliding window to the total number of points in the window; When the packet loss rate or upload jitter exceeds the limit benchmark, the second scoring component is linearly reduced by subtracting the ratio of the exceedance range to the maximum allowable exceedance threshold from 1, with a minimum limit of 0. The third scoring component first divides the absolute deviation by the allowable deviation or rated range of the corresponding measuring point to obtain the dimensionless deviation, and then uses the following reciprocal penalty logic for mapping: in, As the third scoring component, To preset the penalty coefficient, The deviation is dimensionless, so that the larger the deviation, the smoother the score approaches 0, and avoids directly adding or taking the reciprocal of deviations of different dimensions such as temperature, pressure or power. The preset penalty coefficient characterizes the sensitivity of the measuring point to the expected thermodynamic boundary, and is obtained by inverse proportional mapping of the allowable nominal deviation of each measuring point. The fourth scoring component uses the moving average of the measurement sequence calculated within a window of a set length as a baseline. Low-frequency long-wave trends are removed to obtain a high-frequency deviation sequence. The variance of this deviation sequence within the window is calculated and expressed as a natural constant. The operation, which maps the variance to a base and the variance multiplied by a negative decay coefficient as the exponent, is then performed. The interval, where the attenuation coefficient is a preset constant used to calibrate the noise sensitivity of the corresponding measuring point, specifically calibrated as the reciprocal of the maximum variance of the deviation sequence of the measuring point under historical steady-state conditions; The first to fourth weight coefficients are given by a preset configuration file based on the process environment and historical reliability of each sensor. The sum of the four weight coefficients is limited to one. For example, if an infrared temperature measuring device is easily interfered with by dust in the furnace, the third weight coefficient, which represents the physical mechanism consistency verification, can have a first preset weight, such as 0.5, while the fourth weight coefficient, which represents trend stability, has a lower weight, such as 0.1. The method for determining the lower limit is based on the lower quantile of the corresponding measurement point validity score sample dataset accumulated during the historical normal operation cycle, such as the 5th percentile, and statistical calibration. The forgetting factor is used to control the proportion of historical deviation information retained at the current moment. For example, it is set to 0.95. Its value is equal to the ratio of the current monitoring cycle duration to the long-term thermal drift time constant of the sensor, so as to smoothly separate the slow thermal drift of the sensor. This embodiment, through multi-dimensional dynamic validity evaluation and progressive bias compensation based on forgetting factor, can identify local distortions caused by smoke and dust obstruction or high-temperature thermal drift, and transform single-point measurements that are susceptible to noise into reliable data after multi-dimensional verification. Therefore, the adaptive correction of heterogeneous data can be achieved based on the dynamic credibility assessment mechanism, avoiding abnormal data from directly participating in the core state calculation, and solving the problem of frequent false alarms caused by the distortion of single-point data dominating the global judgment in the existing technology.

[0021] As a preferred embodiment, the furnace state calculation module calculates the equivalent temperature of the furnace, the equivalent temperature of the molten copper, the heat storage index of the furnace lining, the energy balance index, and the molten pool disturbance index based on the correction values ​​of each measuring point and the corresponding effectiveness scores. The equivalent temperature of the furnace is calculated by normalizing the corrected furnace infrared temperature, furnace wall contact temperature, flue gas temperature, and heat input equivalent temperature according to their respective effectiveness scores, determining their weights, and then weighted summing them. Heat input equivalent temperature equals gas flow rate, gas lower heating value, and current monitoring cycle duration. Multiply the product of these three factors by the first conversion factor, and then add the power and the current monitoring cycle duration. The product is then multiplied by the second conversion factor; The compensation amount from flue gas temperature to furnace temperature is equal to the third conversion base, plus the product of the energy balance index and the fourth conversion coefficient, plus the product of the furnace door opening / closing status and the fifth conversion coefficient. The equivalent temperature of molten copper is equal to the corrected contact temperature of molten copper multiplied by the first normalization weight, plus the difference between the equivalent temperature of the furnace and the compensation amount multiplied by the second normalization weight, plus the product of the furnace lining heat storage index and the corrected contact temperature of the furnace wall multiplied by the third normalization weight, and then minus the product of the molten pool disturbance index and the preset liquid surface disturbance temperature drop constant multiplied by the fourth normalization weight. Among them, the sum of the first to fourth normalization weights is one, and they are determined by normalization based on the effectiveness scores of copper liquid contact measurement points, furnace measurement points, furnace wall measurement points and molten pool disturbance information, respectively. The furnace lining heat storage index is derived from the furnace lining heat storage index of the previous monitoring period, the temperature difference between the furnace and the furnace wall, the heat input per unit time, and the heat loss per unit time. The heat loss per unit time is determined based on the difference between the flue gas temperature and the ambient temperature, the furnace door opening and closing status, and the deviation of the furnace pressure from the preset staged reference furnace pressure. The energy balance index is calculated based on the deviation of flue gas oxygen content from the preset staged reference values, the deviation of furnace pressure from the staged reference furnace pressure, and the deviation of flue gas temperature from the preset staged reference values. The molten pool disturbance index is calculated by normalizing the score based on the variance of furnace infrared temperature within the sliding window, the change in molten pool surface data, and the number of feeding actions, and then weighting and summing the results. The operating condition and anomaly judgment module divides the copper melting furnace operating condition into the charging disturbance stage, rapid heating stage, melting and homogenization stage, holding and waiting for copper tapping stage, copper tapping stage or abnormal fluctuation stage based on the process action sequence, furnace equivalent temperature change rate, copper liquid equivalent temperature change rate, furnace lining heat storage index, energy balance index and molten pool disturbance index. If the feeding action component in the process action sequence is greater than or equal to the first threshold and the molten pool disturbance index is greater than or equal to the second threshold, then the current working condition stage is determined to be the feeding disturbance stage. If the decrease in furnace infrared temperature is greater than the lower limit of the occlusion judgment set based on the historical statistical values ​​of each working condition stage, and the change in furnace equivalent temperature and flue gas temperature is not greater than their respective upper limit of fluctuation, and the effectiveness score of the furnace infrared temperature corresponding to the measuring point is less than the set lower limit of the score, then an infrared temperature measurement distortion abnormality prompt message will be generated. The operating condition and anomaly judgment module is also used to: based on the historical statistical values ​​of each operating condition stage, set the stage upper limit threshold and duration, if the corresponding upper limit threshold is exceeded and the duration is longer than the set time, generate abnormal prompt information such as overheating, energy supply imbalance, excessive furnace lining heat storage or molten pool instability. The energy consumption assessment and operation and maintenance output module calculates the input energy within the statistical period based on the gas flow rate, the lower heating value of the gas, the electric power, and the preset gas utilization coefficient and electric energy utilization coefficient. It calculates the effective smelting quality based on the initial furnace material quality, the charging quality, the copper output quality, and the loss quality at the beginning of the period, and uses the ratio of input energy to effective smelting quality as the unit smelting energy consumption. The edge device and platform division of labor execution module complete data access, timestamp processing, unified periodic resampling, process action sequence generation, and validity calculation at the edge device, and output a rapid warning when the validity score of each measuring point is less than the set lower limit, the furnace pressure exceeds the limit, or the temperature drops suddenly; The cloud platform completes the search range for measurement point response delay, state calculation coefficient, staged reference values ​​for flue gas temperature and furnace pressure for each operating condition, rolling correction of thresholds, energy consumption assessment, and operation and maintenance information output.

[0022] In this embodiment, after obtaining highly reliable correction data, the correction values ​​and corresponding validity scores of each measuring point are obtained. Based on the preset physical heat transfer constraints, cross-channel fusion calculations are performed on various parameters to generate five state indices that characterize the true physical state of the copper melting furnace. Among them, the first to second conversion coefficients refer to the conversion ratio of mapping the energy physical quantity input from the energy supply side to the furnace temperature support quantity. The energy supply side input includes the gas heating capacity corresponding to the lower heating value of the gas and the electric heating capacity corresponding to the electric power. The product of gas flow rate and gas lower heating value, as well as electrical power, are first accumulated according to the duration of the current monitoring cycle, and then converted into gas input energy and electrical input energy for that cycle, respectively. They are then converted into equivalent temperature increments by a conversion factor that includes time scale, equivalent heat capacity of the furnace body, and heat exchange efficiency. Therefore, the first and second conversion factors each have units for converting the corresponding input energy into temperature, and instantaneous power that has not been accumulated over time cannot be directly added to the temperature value; The third conversion base represents the fixed temperature compensation base for the conversion of the basic heat loss of flue gas. The fourth and fifth conversion coefficients are used to quantify the dynamic heat dissipation conversion compensation caused by energy balance deviation and changes in furnace door status, respectively. The conversion factors mentioned above are all given by the offline experimental calibration configuration and include the corresponding unit conversion meanings. The specific calibration process for the conversion factors is as follows: Under offline no-load and typical load conditions, while keeping other heat exchange boundaries stable, a unit amount of gas or electrical energy is independently applied, and the temperature rise increment in the furnace after reaching steady state is measured. The ratio of this temperature rise increment to the energy input increment is used as the first and second conversion factors. Similarly, the third conversion base and the fourth to fifth conversion factors are determined by recording the stable temperature drop caused by the single-boundary fixed exhaust heat dissipation and the furnace door opening heat dissipation. This ensures that the furnace infrared temperature, furnace wall contact temperature, exhaust temperature and heat input conversion temperature participating in the weighting are all under the temperature dimension or equivalent temperature dimension. The first to fourth normalized weights are used to adaptively increase the contribution of other heat transfer correlation terms, such as furnace radiation or furnace lining thermal conductivity, to the copper liquid temperature reconstruction when the effectiveness of a certain type of sensor, such as a copper liquid thermocouple, decreases. The conversion formulas for the first to fourth normalized weights are based on: The validity score of the current item participating in the reconstruction calculation, namely the copper liquid contact measurement point score, furnace measurement point score, furnace wall measurement point score and molten pool disturbance information score, is divided by the sum of these four validity scores in turn. To avoid division by zero anomalies caused by simultaneous interference from multiple sources, a preset non-zero constant is uniformly superimposed as a smoothing penalty term in the denominator of the normalized calculation. The resulting quotients are the first to fourth normalized weights of the corresponding items. When a local source of information, such as a damaged copper liquid thermocouple, causes the score to approach 0, the weight of that item will automatically shrink or even be eliminated. Meanwhile, the weights of the remaining valid items will automatically increase proportionally by reducing the normalized denominator, ensuring that the total weight always approaches one. If the system detects that the sum of the four validity scores is lower than the safety setting value, the system fault tolerance safety mechanism is triggered, and the historical weight distribution of the previous valid monitoring cycle is forcibly called to achieve mathematical robustness and dynamic self-healing of the calculated weights. The system maintains a fixed-length safety weight snapshot buffer stack in memory, and pushes the current normalized weight onto the top of the stack only when the sum of the periodic validity scores is greater than a preset lower limit. When the system fault tolerance exception branch is triggered, the system pointer is directly popped from the top of the stack to read the unbiased weight allocation of the most recent period. If the number of consecutive failed pops reaches the bottom stack level, the system will intercept the calculation exception and force a switch to safe mode, and forcibly distribute the weights of each incompletely failed information source equally. To ensure the consistency of dimensions among the components of the equivalent temperature of molten copper, the furnace lining heat storage index and the molten pool disturbance index are both defined as dimensionless state indices. The product of the furnace lining heat storage index and the furnace wall contact temperature represents the equivalent temperature term after the heat storage state correction, and the product of the molten pool disturbance index and the liquid surface disturbance temperature drop constant represents the disturbance temperature drop term in terms of temperature dimensions. Among them, the liquid surface disturbance temperature drop constant is determined based on the extreme value of transient temperature drop when the maximum intensity of copper tapping or charging causes the molten pool liquid surface to exceed the limit in historical records; the first to fourth normalization weights are all dimensionless and only serve to distribute the proportions between the temperature terms. This avoids directly adding or multiplying dimensionless exponents, temperature values, and energy values ​​without dimension basis. This embodiment introduces heat input conversion and multi-source normalization weighting mechanism to explore the physical coupling relationship between heat flow in the furnace, furnace wall and copper liquid, and transforms the measurement gap caused by local sensor failure into the state reconstruction result under multi-dimensional thermal balance constraints. When calculating the auxiliary state index, the historical state accumulation and the current thermodynamic deviation are obtained. Based on the recursive accumulation and relative deviation, the furnace lining heat storage index, energy balance index and molten pool disturbance index are determined. Among them, the furnace lining heat storage index reflects the historical cumulative level of heat stored inside the furnace body refractory material after undergoing a heating or heat preservation process; Its recursive process is based on the law of heat conservation. It converts the net heat input per unit time of the current cycle into a dimensionless heat storage increment relative to the equivalent heat capacity of the furnace lining and the preset reference heat input, and then adds it to the heat storage accumulation value of the previous cycle, instead of directly adding the temperature difference, power or heat loss. The net heat input per unit time is obtained by subtracting the heat loss per unit time from the heat input per unit time generated by gas or electric heating. The heat loss per unit time is determined by the difference between the flue gas temperature and the ambient temperature, the opening and closing status of the furnace door, and the deviation of the furnace pressure relative to the stage reference furnace pressure. Specifically, it is equal to the preset heat transfer coefficient multiplied by the difference between the flue gas temperature and the ambient temperature, plus the furnace door heat dissipation coefficient multiplied by the furnace door opening and closing status value, plus the furnace pressure deviation correction coefficient multiplied by the absolute value of the deviation of the furnace pressure relative to the stage reference furnace pressure, thus converting it into the same heat flow scale. When performing the recursive calculation of the heat storage index, the net heat input per unit time is also accumulated as the net input energy of the cycle according to the current monitoring cycle duration, and then divided by the equivalent heat capacity of the furnace lining and the preset reference energy scale to make it dimensionless, so that the recursive term has the same dimension as the dimensionless heat storage index of the previous cycle. The energy balance index characterizes the deviation between the energy supply and exhaust conditions by comparing the deviation of the flue gas oxygen content, furnace pressure and exhaust temperature from their respective stage reference values. In the specific calculation, the absolute values ​​of the differences between the measured values ​​of the above three parameters and the stage reference values ​​are divided by the rated fluctuation tolerance of the corresponding parameters, and then the result is weighted and summed after being dimensionless. The molten pool disturbance index refers to the severity of unstable fluctuations in the liquid surface caused by the impact of feeding or local boiling. Specifically, it is obtained by normalizing the liquid surface change calculated from the furnace infrared temperature variance and the difference between the peak and trough of the liquid surface, as well as the cumulative number of feedings, by dividing each by their respective historical maximum upper limit. The results are then weighted and summed according to preset proportions, such as a weight of 0.5 for liquid surface change, 0.3 for temperature variance, and 0.2 for the number of feedings. If the furnace condition is determined solely by instantaneous temperature, it is easy to misjudge heat loss during the shutdown and heat preservation phase. However, after introducing the furnace lining heat storage index in this embodiment, the system can identify the effect of furnace lining heat release on maintaining the temperature plateau, thus avoiding issuing incorrect re-ignition commands to the combustion control system. Therefore, a quantitative expression of complex thermal inertial processes in the furnace can be realized based on a state reconstruction mechanism with physical consistency, avoiding the limitations of single variable evaluation and solving the problem of process control mismatch caused by the lack of thermal history and disturbance perception in existing technologies. Based on the reconstructed five state indices, the process action sequence and the rate of change of each index are obtained. According to the preset stage division logic and stage statistical thresholds, the current operating condition of the copper melting furnace is determined, and corresponding abnormal warning information is generated. The rules for determining the first and second thresholds are as follows: The first threshold is set based on the minimum process action characteristic value that can cause the furnace temperature change rate to exceed the preset threshold, as recorded from historical charging condition characteristic samples. The second threshold is based on the maximum background noise level of the molten pool disturbance index during the historical pure steady-state heat preservation stage, i.e., without any charging disturbance. For example, it is calibrated using the mean plus three times the standard deviation rule. Only when both of these physical thresholds, which are reflected by different dimensions, are exceeded simultaneously is it determined to be the charging disturbance stage. This can effectively intercept single-factor false triggering caused by environmental vibration interference. The lower limit of the shading judgment refers to the minimum jump amplitude of the furnace infrared temperature that is not a real physical temperature drop caused by smoke and dust shading under specific operating conditions, such as an instantaneous temperature drop exceeding 120°C. Meanwhile, the system has established a clear abnormality type mapping matrix: if the equivalent temperature of the furnace or the equivalent temperature of the copper liquid exceeds the upper limit threshold of the corresponding stage and exceeds the time limit, that is, if it continues to exceed the set time, the set time is calibrated according to the short-term fluctuation tolerance limit allowed by the process in this working condition stage, for example, it is set to 3 consecutive monitoring cycles, triggering an overheating prompt; if the energy balance index exceeds the upper limit threshold, triggering an energy supply imbalance prompt. If the furnace lining heat storage index exceeds the upper limit and continues to do so, it is determined that the furnace body is saturated with heat storage, triggering a high furnace lining heat storage warning to guide the cessation of high-power heating and heat preservation; if the molten pool disturbance index continues to exceed the upper limit during non-feeding operation, a molten pool instability warning is triggered. The system has built-in operation and maintenance mapping rules: when the effectiveness score of the furnace infrared measuring point is lower than the set lower limit for a set number of consecutive preset cycles, the system outputs maintenance suggestions to clean the infrared temperature measuring lens; when the unit smelting energy consumption exceeds the historical benchmark average and the flue gas temperature continues to exceed the stage upper limit threshold, the system outputs maintenance suggestions to check the furnace body seal and repair the furnace lining. This embodiment matches independent thresholds for different operating conditions and combines them with multi-source state constraints. For example, if the furnace infrared temperature drops suddenly but the flue gas temperature remains stable, it can intercept false cooling alarms and improve the accuracy of operating condition perception. Therefore, dynamic tracking and anomaly cause differentiation in the production stage can be achieved based on multi-dimensional state feature collaborative judgment, avoiding interference from a large number of invalid alarms to on-site operators, and solving the problem that the sensitivity and accuracy of early warning cannot be balanced due to the uniform threshold setting in the existing technology. When outputting the final evaluation results, the energy input and material flow data within the statistical period are obtained, and the unit smelting energy consumption is calculated based on the calorific value utilization rate and effective smelting quality. At the same time, all the aforementioned calculation tasks are coordinated and allocated according to the processing capabilities of the edge terminal and cloud platform and the timeliness requirements of the task. The input energy within the statistical period is not directly taken as the instantaneous value of gas flow or electric power at a certain moment. Instead, the product of gas flow and gas lower heating value and electric power are accumulated over time according to the monitoring period, and then multiplied by the corresponding gas utilization coefficient and electric energy utilization coefficient to obtain the accumulated energy that matches the effective smelting quality. Edge devices undertake tasks that meet preset timeliness requirements, such as second-level resampling and rapid local early warning of sudden temperature drops; cloud platforms undertake computational tasks that require long-term prior recording support, such as the rolling adaptive update of the measurement point response delay search range, state calculation coefficients, and phased reference thresholds. To address the rolling correction of the search range for the response delay of the measurement point, the cloud platform constructs a three-dimensional feature vector based on the equivalent furnace temperature, flue gas temperature, and input energy of the previous steady-state cycle. After dimensionless normalization of each physical quantity in the three-dimensional feature vector using historical global parameter extreme values, the Euclidean distance is used to calculate the distance difference between the normalized feature vector of each cycle in the previous record library and the current normalized three-dimensional feature vector. The first preset number of cycles with the smallest distance difference are extracted as historical similar working condition cycles, and the set of response lag delays occurring within them is obtained. The upper and lower boundaries of their statistical distribution are used as the preset upper and lower limits of the delay for the new stage and then issued. To address the rolling correction of the state calculation coefficients, measured data from the historical pure steady-state insulation stage are extracted, and the conversion coefficients for gas and electric heating are updated based on the heat balance equation. The adaptive update of the reference threshold is specifically implemented through an exponential smoothing update mechanism. The statistical mean of the data under each steady-state condition in the current period is extracted, multiplied by the update step size, such as 0.05, and then added to the old threshold of the previous period to generate a new threshold. In the control flow for calculating the statistical mean, to avoid the pollution of the long-term threshold caused by abnormal jump values ​​due to changes in environmental parameters exceeding the preset range, the system first triggers the abnormal sample cleaning branch when the time window closes. The volatility is extracted by calculating the standard deviation of the sample sequence within the window, and high-risk noise points are forcibly removed based on the boundary of the mean plus or minus three times the standard deviation. The net mean is recalculated only for qualified samples that fall within the confidence boundary. The mean after removing samples that exceed the limit is used as the filter increment source, thereby realizing the anti-disturbance quantitative rolling tracking of the long-term aging trend of the furnace or environmental drift.

[0023] When deployed in an industrial site, the raw data streams collected by the field sensors at high frequency are continuously fed into the edge gateway; the edge device completes the timestamp processing and validity calculation in real time, and once it detects indicators that endanger safety, such as furnace pressure exceeding the limit, it immediately triggers an audible and visual warning in the field workshop; the synchronized and aligned corrected data is packaged and uploaded to the cloud platform; After receiving the data, the cloud platform combines the statistical patterns of historical furnace cycles, continuously adjusts various conversion factors and thresholds, calculates the unit smelting energy consumption of the current furnace cycle, and pushes maintenance guidance information such as high furnace lining heat storage and suggestions to shorten the holding time for copper tapping to the plant-level management dashboard. The new process adjustments are captured by the sensors again, thus forming a complete industrial internet monitoring closed loop from data perception and anomaly interception to energy efficiency optimization.

[0024] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A copper melting furnace operation monitoring system based on the Industrial Internet, characterized in that, This includes field sensors, edge devices, and cloud platforms: The multi-source data acquisition module is deployed on edge devices to collect information transmitted through the network, process operation actions, and measurement data including at least furnace temperature, furnace wall temperature, flue gas temperature, heat input and heat loss parameters through field sensors and external input. The process action recording module generates a process action sequence based on the process operation actions; The data time synchronization module corrects the sampling time and response delay of the measuring points according to the network transmission information and process action sequence, and maps the measurement data to the same preset monitoring cycle; The data anomaly correction module performs validity calculations and deviation corrections on the synchronized measurement data. The furnace state calculation module calculates the equivalent temperature of the furnace, the equivalent temperature of the molten copper, the furnace lining heat storage index, the energy balance index, and the molten pool disturbance index based on the preset relationship between the above temperature and heat parameters. The operating condition and anomaly determination module determines the operating condition stage and anomaly type based on the furnace equivalent temperature, copper liquid equivalent temperature, furnace lining heat storage index, energy balance index and molten pool disturbance index calculated above. The energy consumption assessment and operation and maintenance output module outputs operation and maintenance information including furnace status, abnormal warnings, energy consumption assessment results and maintenance suggestions, measurement point reliability, abnormal heat loss or stage switching prompts; The edge and platform division of labor execution modules will assign computational tasks from data collection to anomaly correction to edge devices, and computational tasks from status calculation to operation and maintenance output to the cloud platform.

2. The copper melting furnace operation monitoring system based on the Industrial Internet according to claim 1, characterized in that, The multi-source data acquisition module collects data through the field sensors, including furnace infrared temperature, furnace wall contact temperature, copper liquid contact temperature, flue gas temperature, ambient temperature, furnace pressure, flue gas oxygen content, gas flow rate, electric power, furnace door opening and closing status, feeding action, copper tapping action, and molten pool level data, wherein the gas flow rate and the electric power are at least one of them. The edge device receives externally input network latency, packet loss rate, jitter information, low calorific value of gas, initial furnace material quality, feeding quality, copper output quality, and loss quality. For furnace types without gas heating, the gas flow rate is zero; for furnace types without electric heating, the electric power is zero. Each data item includes a sampling timestamp, upload timestamp, source device identifier, measurement point identifier, process location identifier, unit identifier, and data validity identifier.

3. The copper melting furnace operation monitoring system based on the Industrial Internet according to claim 2, characterized in that, The process action recording module, according to a unified preset monitoring cycle, converts the feeding action, furnace door opening / closing status, electrical power, gas flow rate, and copper tapping action into a process action sequence using binary events, hierarchical coded values, or normalized changes. The process action sequence is then used to record the first... The sequence value for each monitoring cycle is equal to the sum of the products of the feeding action, the furnace door opening / closing status, the electrical power, the gas flow rate, and the copper output action within that cycle, respectively, and the corresponding first to fifth preset weighting coefficients. It is a positive integer representing the monitoring cycle number.

4. The copper melting furnace operation monitoring system based on the Industrial Internet according to claim 3, characterized in that, The data time synchronization module determines the network latency based on the difference between the upload timestamp and the sampling timestamp, and calculates the adjacent period difference of each measurement point data and the response degree of the process action sequence under different candidate delays within a preset latency lower limit to a preset latency upper limit and within a preset length sliding window. The response degree is equal to the sum of the absolute values ​​of the products of the adjacent period difference at each moment within the sliding window and the sequence value of the process action sequence. The candidate delay that maximizes the response level is determined as the measurement point response delay, and the measurement point data is mapped to the time corresponding to the current physical time minus the measurement point response delay using an interpolation operator. Continuous measurement signals are interpolated linearly, while discrete event signals are mapped using nearest neighbor mapping.

5. The copper melting furnace operation monitoring system based on the Industrial Internet according to claim 4, characterized in that, The data anomaly correction module calculates the validity score of each measuring point. The validity score is equal to the sum of the first to fourth score components multiplied by their corresponding first to fourth evaluation weight coefficients. The first score component is determined by missing points, saturation points, and jump points exceeding the preset unit time change limit within the sliding window. The second score component is determined by upload jitter and packet loss rate. The third score component is determined by the deviation between the synchronized measurement value and the verification reference value calculated based on other related measuring points and the preset heat balance relationship. The fourth score component is determined by the degree of fluctuation after removing the change trend. The sum of the four weight coefficients is one. The data anomaly correction module updates the measurement point deviation. The measurement point deviation in the current period is equal to the measurement point deviation in the previous period multiplied by the forgetting factor, plus the difference between the synchronized measurement value and the verification reference value multiplied by one minus the forgetting factor. The forgetting factor is greater than or equal to zero and less than one. The corrected value after deviation correction is obtained by subtracting the measurement point deviation of the current cycle from the synchronized measurement value. If the validity score is less than the set lower limit, the measured value of the corresponding measuring point is replaced by the verification reference value; otherwise, the correction value is used.

6. The copper melting furnace operation monitoring system based on the Industrial Internet according to claim 5, characterized in that, The furnace state calculation module calculates the equivalent temperature of the furnace, the equivalent temperature of the molten copper, the heat storage index of the furnace lining, the energy balance index, and the disturbance index of the molten pool based on the correction values ​​of each measuring point and the corresponding validity scores.

7. The copper melting furnace operation monitoring system based on the Industrial Internet according to claim 6, characterized in that, The equivalent temperature of the furnace is obtained by normalizing the corrected furnace infrared temperature, furnace wall contact temperature, flue gas temperature, and heat input equivalent temperature according to their respective effectiveness scores, determining their weights, and then calculating the weighted sum. The heat input converted temperature is equal to the gas flow rate, the lower heating value of the gas, and the current monitoring cycle duration. The product of these three factors, multiplied by the first conversion factor, plus the power and the current monitoring cycle duration, is then added. The product is then multiplied by the second conversion factor; The compensation amount from flue gas temperature to furnace temperature is equal to the third conversion base, plus the product of the energy balance index and the fourth conversion coefficient, plus the product of the furnace door opening / closing status and the fifth conversion coefficient. The equivalent temperature of the molten copper is equal to the corrected contact temperature of the molten copper multiplied by the first normalization weight, plus the difference between the equivalent temperature of the furnace and the compensation amount multiplied by the second normalization weight, plus the product of the furnace lining heat storage index and the corrected furnace wall contact temperature multiplied by the third normalization weight, and then minus the product of the molten pool disturbance index and the preset liquid surface disturbance temperature drop constant multiplied by the fourth normalization weight. The sum of the first to fourth normalization weights is one, and they are determined by normalization based on the effectiveness scores of the copper liquid contact measurement points, furnace measurement points, furnace wall measurement points, and molten pool disturbance information, respectively.

8. The copper melting furnace operation monitoring system based on the Industrial Internet according to claim 7, characterized in that, The furnace lining heat storage index is derived from the furnace lining heat storage index of the previous monitoring period, the temperature difference between the furnace chamber and the furnace wall, the heat input per unit time, and the heat loss per unit time. The heat loss per unit time is determined based on the difference between the flue gas temperature and the ambient temperature, the opening and closing status of the furnace door, and the deviation of the furnace pressure from the preset staged reference furnace pressure. The energy balance index is calculated based on the deviation of the flue gas oxygen content from a preset staged reference value, the deviation of the furnace pressure from the staged reference furnace pressure, and the deviation of the flue gas temperature from a preset staged reference value. The molten pool disturbance index is obtained by normalizing and weighting the scores based on the variance of the furnace infrared temperature within the sliding window, the change in molten pool surface data, and the number of feeding operations calculated by the feeding action.

9. The copper melting furnace operation monitoring system based on the Industrial Internet according to claim 8, characterized in that, The operating condition and anomaly determination module divides the copper melting furnace operating condition into the charging disturbance stage, rapid heating stage, melting and homogenization stage, holding and waiting for copper tapping stage, copper tapping stage or abnormal fluctuation stage based on the process action sequence, furnace equivalent temperature change rate, copper liquid equivalent temperature change rate, furnace lining heat storage index, energy balance index and molten pool disturbance index. If the feeding action component in the process action sequence is greater than or equal to the first threshold and the molten pool disturbance index is greater than or equal to the second threshold, then the current working condition stage is determined to be the feeding disturbance stage. If the decrease in the furnace infrared temperature is greater than the lower limit of the occlusion judgment set based on the historical statistical values ​​of each working condition stage, and the change in the furnace equivalent temperature and the flue gas temperature is not greater than their respective upper limits of fluctuation, and the effectiveness score of the furnace infrared temperature corresponding to the measuring point is less than the set lower limit of the score, then an infrared temperature measurement distortion abnormality prompt message is generated. The operating condition and anomaly determination module is also used to: based on the phased upper limit threshold and duration set according to the historical statistical values ​​of each operating condition stage, if the corresponding upper limit threshold is exceeded and the duration is longer than the set time, generate abnormal prompt information such as overheating, energy supply imbalance, excessive heat storage in the furnace lining, or instability of the molten pool.

10. The copper melting furnace operation monitoring system based on the Industrial Internet according to claim 9, characterized in that, The energy consumption assessment and operation and maintenance output module calculates the input energy within the statistical period based on the gas flow rate, the lower heating value of the gas, the electric power, and the preset gas utilization coefficient and electric energy utilization coefficient. It calculates the effective smelting mass based on the initial furnace material mass, the charging mass, the copper output mass, and the loss mass. The ratio of the input energy to the effective smelting mass is used as the unit smelting energy consumption. The edge terminal and platform division of labor execution module completes data access, timestamp processing, unified periodic resampling, process action sequence generation, and initial validity calculation on the edge device, and outputs a rapid warning when the validity score of each measuring point is less than the set lower limit, the furnace pressure exceeds the limit, or the temperature drops suddenly. The cloud platform completes the search range for the response delay of the measuring points, the state calculation coefficient, the staged reference values ​​of the flue gas temperature and the staged reference values ​​and thresholds of the furnace pressure for each working condition, as well as the rolling correction of energy consumption assessment and operation and maintenance information.

Images