Steel billet porous combustion uniform-temperature closed-loop control method

Abstract

The invention relates to the technical field of industrial furnace thermal control and combustion regulation, and discloses a steel billet porous combustion uniform-temperature closed-loop control method, which is used for solving the problem that in the traditional method, the temperature of steel billet porous combustion is difficult to control on the premise of not adding a complex contact temperature measuring device. Effective feedback information capable of representing the temperature difference change of the billet in the length direction and the width direction is obtained; the method comprises the steps that firstly, a freezing furnace zone, a burner group, temperature measurement mapping and a threshold version are converted; the time base of multi-source temperature measurement, execution quantity and furnace condition data is unified, and grading admission is carried out; constructing an observation window according to a tapping event to extract a length and width temperature difference, and combining with a burner to carry out consistency verification; according to the mapping version, generating a limited small-step adjusting quantity, a stable boundary, a safety boundary, furnace pressure and temperature measurement quality gating issuing, triggering version switching or rollback and degradation recovery through a probe and recheck, and realizing stable convergence of the temperature difference.

Description

Technical Field

[0001] This invention relates to the field of industrial furnace thermal control and combustion regulation technology, specifically a closed-loop control method for uniform temperature control of porous combustion in steel billets. Background Technology

[0002] Existing billet reheating furnaces generally employ zoned heating and synchronous adjustment of burner groups for heating system control. Controlled variables typically include furnace temperature setpoints for the preheating, heating, and soaking sections, as well as parameters such as gas flow rate, combustion air flow rate, and furnace pressure. To adapt to changes in production cycle and steel grade specifications, DCS closed-loop control is often used in engineering practice, with thermal models or predictive algorithms used in the secondary system to correct setpoints for stable control of furnace temperature and heating load. For example, the published invention patent application CN101929736B discloses a furnace temperature control method based on regenerative burners, which uses simulated continuous control and... The combination of digital discrete burn-out control and furnace temperature as feedback quantity to achieve temperature regulation; for example, the published invention patent application CN101256418B discloses a comprehensive control method for the outlet temperature of a heating furnace, which makes the outlet temperatures of multiple branches approach uniformity and achieves load regulation through branch balance control and state feedback predictive control; although the above methods can improve the stability of furnace temperature or outlet temperature regulation, the feedback quantity still mostly comes from macroscopic quantities such as furnace temperature, zone temperature or single outlet temperature, which cannot characterize the temperature difference change of the billet along the length and width directions, and it is difficult to directly suppress the billet uniform temperature index through closed loop; In actual production, the uniform temperature index of the steel billet is jointly determined by the surface and internal temperature fields of the billet, and is also affected by the heat release distribution of each zone's burner group. This influence relationship is characterized by strong coupling, long lag time, and significant nonlinearity. Limited by the arrangement of temperature measurement points and the measurement uncertainties of non-contact temperature measurement under conditions such as field-of-view obstruction and emissivity fluctuations, existing methods generally struggle to obtain effective feedback information representing the temperature difference changes along the length and width directions of the steel billet without adding complex contact temperature measurement devices. Consequently, it is difficult to establish a correspondence between the steel billet temperature difference and the actual effective heat release contribution of each zone's burner group. Due to the lack of reliable evidence for this correspondence, existing closed-loop methods... Traditional closed-loop control systems often use furnace temperature or zone temperature as feedback, resulting in control adjustments primarily focused on furnace temperature stability rather than directly constraining billet temperature difference. Consequently, during operational disturbances, changes in production cycle time, or switching of heating modes, excessive temperature differences at the head, middle, and tail of the billet, or between the edges and middle, can easily occur, leading to increased uniform temperature fluctuations and prolonged convergence time. Therefore, a closed-loop uniform temperature control technology for zoned combustion regulation in billet reheating furnaces is needed. This technology aims to achieve online acquisition and consistency verification of billet temperature difference indicators without adding complex contact temperature measurement arrangements, and to guide the closed-loop distribution of heat release from the zoned burner groups, thereby improving billet uniform temperature stability and shortening convergence time after disturbances. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a closed-loop control method for uniform combustion temperature of steel billets in porous structures. This method solves the problem in traditional methods of obtaining effective feedback information that represents the temperature difference changes of steel billets along their length and width without adding complex contact temperature measuring devices.

[0004] To achieve the above objectives, the present invention provides the following technical solution: A closed-loop control method for uniform combustion temperature of steel billets in porous structures includes: S1: Establish a session to define the freezer zone rules, burner group numbers, temperature measurement point mapping, uniform temperature index caliber, and version effective boundaries. S2: Unify the time base of zone temperature measurement, furnace exit temperature measurement, combustion execution quantity and furnace condition data, make access judgment on data validity and record the status of occlusion, drift, loss and alarm. S3: Establish an observation window based on the furnace exit event, generate records of temperature difference in the length and width directions, summarize the burner group actions within the window and complete the consistency verification. S4: Based on the temperature difference status, call the mapping version to determine the adjustment action, perform boundary gating and then issue it, and associate the observation window with the evidence number; S5: When the triggering conditions are met, insert a probe window and enter the review window to complete the version switch or rollback and set the prohibition period. In case of an anomaly, enter the downgrade process and return to the session after the recovery conditions are met.

[0005] Furthermore, a session is established to define the freezer zone rules, burner group numbers, temperature measurement point mapping, uniform temperature index definitions, and version validity boundaries, including: The session number is generated from the timestamp, furnace number, and shift sequence number; The session configuration entry is assigned a version number, written to the session snapshot, and registered in the audit index; Before freezing, perform consistency and integrity checks. If a test is missing, mark the observation as restricted and disable probe permissions. If the boundary is abnormal, switch to conservative actions. If session establishment fails, the previous version will be used and the reason code will be logged.

[0006] Furthermore, the time base for zone temperature measurement, furnace exit temperature measurement, combustion execution data, and furnace condition data will be unified, including: The sampling frames are accessed in a rolling manner according to a fixed control cycle and carry the session number, caliber version number, gating threshold version number and threshold source identifier type code; The time base parameters are taken from the session freeze item, and the timestamps from each source are aligned to the gateway clock; If the alignment error exceeds the limit, mark the time as abnormal and switch to the conservative channel; if the processing timeout occurs, record the timeout event.

[0007] Furthermore, the validity of the data is assessed for entry and the status of occlusion, drift, loss, and alarms is recorded, including: Access determination employs point-level, segment-level, and session-level gating and outputs quality-labeled frames. Point-level classification is based on the effective sampling ratio and field of view coverage, combined with emissivity stability, drift, and loss criteria. At the section level, a restricted list is generated based on the execution permission and saturation flags; The session level triggers a degradation strategy code based on the furnace pressure fluctuation level and alarm mapping, and registers abnormal events for sampling mutations and point conflicts.

[0008] Furthermore, an observation window is established using the furnace exit event as an anchor, generating records of temperature differences along the length and width, including: The observation window ends at the release timestamp and traces back according to the list of key segments at the end of the session freeze; Window validity gating references the admission standard frame and verifies tracking to advance monotonicity; When the window is valid, representative values ​​are extracted from the head and tail regions and the left, middle and right regions according to the frozen sampling ratio. The head and tail representative values ​​are taken as the median value of several consecutive control cycles, and the length temperature difference and width temperature difference records are generated and assigned temperature difference quality level.

[0009] Furthermore, the summary of burner group actions within the window is compiled and a consistency verification is performed, including: The action summary organizes the combustion execution changes into non-repeating action segments within the minimum stable holding time according to the burner group number, and records the action type code, amplitude level, saturation mark and gating clipping mark; Consistency verification uses a lag level table and improvement thresholds to determine the timing, direction, and boundaries. If the verification is invalid, write the invalidity reason code, and associate the verification record with the window number.

[0010] Furthermore, based on the temperature difference status, the mapping version is invoked to determine the adjustment action, which is then issued after boundary gating, and the observation window and evidence number are associated, including: The temperature difference status is formed by summing the near-end temperature differences and carries the exceedance level and trend indicator; Based on the priority and lag level of the mapped version, limited action candidates are generated and filtered out using small steps, combined with the section temperature margin and execution saturation. Candidate actions are cut or frozen and then issued after being subject to stability lower limit, safety upper limit, furnace pressure, alarm and temperature measurement quality gating; The control output is associated with the window number and verification number, and the execution rejection, feedback missing and timeout events are registered as disabled, downgraded and timeout events respectively.

[0011] Furthermore, when the triggering condition is met, a probe window is inserted and the verification window is entered, including: The trigger entry is frozen within the session and the rule number is registered. The trigger type is continuous over-limit, insufficient convergence, mode switching or evidence quality degradation, and it must simultaneously meet the quantity threshold and the quality threshold. Once the trigger is established, a limited number of burner groups are selected to perform restricted small-step perturbations. The probe window duration is set according to the hysteresis level, and probe records are generated and registered in the evidence index under stricter gating conditions.

[0012] Furthermore, after completing the version switch or rollback and setting a prohibition period, in case of an anomaly, the process enters the downgrade procedure and returns to session execution after the recovery conditions are met, including: When the probe records meet the conditions of consistent timing, consistent direction, and consistent boundaries, candidate mapping versions are generated only for the burner group entries participating in the probe within the range of adjacent lag levels or adjacent priority levels. Switching or rollback is performed at the effective boundary of the version, a prohibition period is set, and the same rule number is prohibited from being triggered again. When the probe or verification fails, or when a serious alarm occurs, temperature measurement is lost for a long time, execution is refused continuously, or the furnace pressure fluctuation level exceeds the limit, the process will switch to limited downgrading and will exit downgrading after the quality level is restored, the furnace pressure level is restored, and the confirmation record is valid.

[0013] Compared with the prior art, the present invention provides a closed-loop control method for uniform combustion temperature of steel billets in porous structures, which has the following beneficial effects: 1. This invention, through a session-based approach, freezes the furnace zone, burner group number, temperature measurement mapping, uniform temperature index caliber, and gate threshold version. It uses the unified time of zone temperature measurement, furnace exit temperature measurement, combustion execution quantity, and furnace condition data as a benchmark, and implements point-level, segment-level, and session-level access control. This allows for explicit labeling and constraints on issues such as obstruction, unstable emissivity, drift, loss, and alarms. An observation window is constructed using the furnace exit event as an anchor. The frozen caliber extracts the temperature difference along the length and width directions, as well as the quality grade, and verifies the temporal, directional, and boundary consistency of burner group action segments within the window. This establishes a relationship between temperature difference changes and effective adjustment actions. The evidence chain is traced; the control side generates limited small-step adjustment candidates based on temperature difference status and mapping version. After passing through minimum stability boundary, maximum safety boundary, furnace pressure, alarm, and temperature measurement quality gating, when the trigger condition is met, the version is switched or rolled back through the probe window and the verification window. In case of anomalies, the amplitude is limited and downgraded and the session is rolled back. Under production change and operating condition disturbances, the adjustment process is kept verifiable, rollbackable, resistant to misjudgment, and stably meets the uniform temperature control requirements. This solves the problem in traditional methods that it is difficult to obtain effective feedback information that can represent the temperature difference change of the billet along the length and width directions without adding complex contact temperature measurement devices.

[0014] 2. This invention unifies the numbering and caliber of temperature difference observation, action execution, and version management. It assigns the same window number and verification number to the furnace exit temperature measurement, billet tracking, zone temperature measurement, and burner group execution status of each billet. This ensures that adjustments are locatable and traceable even during cycle fluctuations, execution saturation, or localized temperature anomalies, reducing attribution confusion and strategy drift caused by reliance on single-point furnace temperatures or empirical rules. Addressing the characteristics of reheat furnaces, such as high thermal inertia and strong segment coupling, it reduces the risk of oscillations caused by the combined action of multiple actuators by limiting actions and small step amplitudes, and implementing stable lower limits, safe upper limits, and furnace condition-related gating constraints in a fixed sequence. In cases of persistent over-limits and insufficient convergence, it provides a rollback path for version updates through probe windows and verification windows with limited disturbances, avoiding instability caused by frequent trials. Ultimately, this ensures stable iteration of the adjustment strategy and long-term stable operation of the uniform temperature control. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the process for a closed-loop control method for uniform combustion temperature of steel billets in the form of a porous structure according to the present invention. Detailed Implementation

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

[0017] Example 1: Figure 1 A closed-loop control method for uniform combustion temperature of steel billets in porous combustion chambers is presented, including: S1: Establish a session, distinguishing between freezing furnace zone rules, burner group numbers, temperature measurement point mapping, uniform temperature index caliber, and version validity boundaries. The specific implementation is as follows: At the start of production changeover, when the heating system changes from heating up to homogenization or from standby to recovery, or when the operation and maintenance window arrives, a session establishment process is initiated and a session number is generated, serving as the unique index for this operation caliber and version switch. The session number is formed by concatenating a timestamp, furnace number, and shift sequence number. The timestamp is accurate to the second, the furnace number is a fixed field code, and the shift sequence number is an incrementing sequence number of at least two digits, such as 20260207153000 plus one furnace number and at least two sequence numbers. This setting is based on avoiding numbering conflicts caused by multiple furnaces operating in parallel and cross-shift handover, and supporting traceability. When the session is established, the operation caliber parameters are encapsulated as session configuration entries and frozen within the session. These entries include at least the following fields: furnace area division table, zone rules, burner group list, temperature measurement mapping table, homogenization index caliber table, version effective boundary rules, gate control threshold table, and resource budget table. Each field generates a version number and is associated with and fixed to the session number. The furnace zone division table should include at least the following fields: furnace length direction section boundaries, section numbering method, etc., to express the range of the preheating zone, heating zone, and soaking zone. The section boundaries can be numbered in steps or at equal distances. The number of sections in the furnace length direction can be 6 to 12, preferably 8 to 10. The selection is based on matching the section length with the billet dwell time under the stepping cycle, so that the adjustment of a single section forms a recognizable response within the observation window, and reduces the difficulty of coupling attribution caused by the synchronous adjustment of adjacent sections. The zoning rules should include at least the following fields: left zone, middle zone, and right zone coverage ratio, etc. The coverage ratio of the left and right zones can be 0.25 to 0.35, and the middle zone takes the remaining ratio, for example, 0.30 for the left zone, 0.40 for the middle zone, and 0.30 for the right zone. The ratio setting is based on consistency with the burner arrangement bandwidth and ensures that the furnace exit field of view has stable coverage in the left, middle, and right regions, avoiding the loss of representative values ​​due to occlusion when the zoning is too narrow. The burner group list generates burner group numbers by combining section numbers and zone numbers, and establishes parameter entries for each burner group, including at least the fields of allowable adjustment type, minimum stability boundary, maximum safety boundary, cycle boundary, and control unit caliber. The allowable adjustment type can be selected as gas adjustment, combustion air adjustment, reversing or pulse cycle adjustment, and opening / closing or switching to a conservative position. The minimum stability boundary and maximum safety boundary are used to constrain combustion stability and equipment safety, respectively. For example, the lower limit of gas adjustment is 5 to 15 control units, and the upper limit is 80 to 100 control units. The combustion air adjustment boundary is set in conjunction with the gas boundary. The cycle boundary is, for example, 2 to 12 cycle units. The control unit is a percentage point of the gas valve opening or the minimum resolution unit of the gas flow setting, given by the actuator calibration table and frozen in the session. The setting principle of the cycle boundary refers to the shortest allowable interval of reversing or pulse execution and the combustion stability establishment time to avoid frequent cycle switching and suppress furnace pressure fluctuations. The temperature mapping table is used to establish the correspondence between temperature measurement zones and temperature measurement sources. Each temperature measurement zone is associated with at least one zone temperature measurement point and one non-contact temperature measurement field of view at the furnace outlet. It includes at least the following fields: point number, installation location, representative zone, data source priority, and occlusion sensitivity marker. The data source priority can be selected as furnace outlet field of view first, side wall points second, and furnace apex third, because the furnace outlet field of view is more representative of the billet surface temperature and has higher consistency with the temperature uniformity index. The occlusion sensitivity marker is used for point risk classification and is associated with the gating threshold table. The temperature uniformity index caliber table freezes the definition of the temperature difference in the length direction and the temperature difference in the width direction, the sampling area ratio, and the target interval within the session. The length direction temperature... The temperature difference is taken as the temperature difference between the head and tail regions of the same billet exiting the furnace, with the head-to-tail coverage ratio ranging from 0.10 to 0.20, for example, 0.15. The temperature difference in the width direction is taken as the temperature difference between the left, middle, and right zones, with the left and right zone coverage ratio ranging from 0.20 to 0.35, for example, 0.25. The above ratio range is set based on the fact that the exit field of view is more susceptible to occlusion and edge effects at the ends and edges. Using this range can ensure that the effective pixel coverage meets the minimum requirements for representative value extraction and reduce fluctuations. The target range, for example, is a temperature difference of no more than 15℃ in the length direction and no more than 12℃ in the width direction. Its source is the achievable range after superimposing the process allowable fluctuations and the uncertainty of non-contact temperature measurement. Version effectiveness boundary rules are used to constrain the timing of switching between the mapped version and the caliber version. The billet queue boundary is prioritized as the effectiveness boundary, followed by the step event boundary, to avoid incomparable evidence due to caliber switching occurring during the operation of a single billet in the furnace. The gating threshold table is frozen in a field-based manner within the session, including at least the following fields: time alignment allowable error, effective sampling percentage threshold for temperature measurement points, furnace exit field of view coverage threshold, emissivity stability duration threshold, temperature drift judgment duration threshold, sampling loss cycle threshold, furnace pressure fluctuation level upper limit, alarm level mapping table, and degradation strategy code. For example, the time alignment allowable error is set to 5 seconds; the effective sampling percentage stability threshold is set to 0.92, the unusable threshold is set to 0.85, and the usable field of view coverage threshold is set to 0. 8. The occlusion threshold is set to 0.6; the emissivity instability duration threshold is set to 60 seconds; the drift suspicion duration threshold is set to 20 minutes, and the confirmation threshold is set to 60 minutes; the short loss threshold is set to 3 control cycles, and the long loss threshold is set to 12 control cycles; the upper limit of the furnace pressure fluctuation level is set to three levels; the furnace pressure fluctuation level can be selected from the fluctuation level code given by the furnace condition frame, or mapped to the furnace pressure fluctuation amplitude and duration in the gating threshold table; the alarm level is determined according to the alarm code and the level mapping table, and the severe level triggers the downgrade strategy code; the threshold source identifier is synchronously written into the gating threshold table and frozen in the session, and the source type can be selected to include actuator calibration, safety policy, process procedure and temperature measurement system acceptance criteria; the threshold update is completed by freezing in a new session, and the reason for the change and the source type are recorded in the audit index; The resource budget table is used to constrain the time resources for session establishment and runtime. It includes at least the following fields: session establishment timeout threshold, degradation policy code, policy triggering conditions, timeout reason code, and list of unfrozen fields. For example, the session establishment timeout can range from 10 to 60 seconds, with 20 seconds being preferred. The setting is based on the cycle time window for on-site production changes and mode switching, as well as the upper bound of gateway data entry and verification time. Degradation policy codes can be selected such as pausing temperature difference driving and retaining only conservative control of zone temperature, freezing version updates and probe triggers, and limiting action steps to no more than one level. A corresponding relationship is established with the triggering conditions for traceable invocation. Before freezing the session, consistency and integrity checks are performed. The consistency check requires that the furnace area division table segment number and the burner group list be consistent, and the zone division rules... Consistent with the burner group zone numbering, each zone in the temperature measurement mapping table must have at least one temperature measurement source entry and include a priority definition; integrity checks require each table to have a version number, generation timestamp, and responsible source identifier, and the resource budget table must include a timeout threshold and degradation policy code; if a zone is found to be missing a temperature measurement source, the zone is marked as observation-restricted and written into an event entry, and the probe window permission for that zone is set to prohibited; observation-restricted zones are not used as evidence sources in mapping version updates, and only conservative gear actions or hold actions are allowed; if a burner group boundary is found to be missing or exceeds the plant safety policy, the corresponding burner group is marked as only allowing conservative actions and written into an event entry as a mandatory constraint for subsequent boundary gating; Freezing a session writes configuration entries to a session snapshot. The session snapshot includes at least the session number, version number of each table, freeze timestamp, effective boundary, and initial status code, and generates a read-only checksum. The audit index includes at least the reason for session establishment, key threshold snapshots, and a list of missing and restricted fields. The source of key thresholds is identified by actuator calibration, safety policies, process procedures, or temperature measurement system acceptance criteria. When a session fails, it runs along the version number of the previous session and triggers a degradation policy code, recording the timeout reason code and a list of unfrozen fields.

[0018] S2: Unify the time base of zone temperature measurement, furnace exit temperature measurement, combustion execution data, and furnace condition data; determine the validity of the data and record obstruction, drift, loss, and alarm status. Specifically, the implementation is as follows: After the session snapshot is frozen, the data unification and access determination process begins, running in a fixed control cycle. The control cycle is between 2 and 10 seconds, for example, 5 seconds. The reason for this selection is that it is on the same order of magnitude as the minimum stable holding time of actuators such as gas valves and dampers, and matches the refresh cycle of field communication to reduce the risk of frequent switching. The input uses a set of sampling frames, which, according to their source, include at least the following types: zone temperature measurement frames, furnace temperature measurement frames, combustion execution frames, furnace condition frames, and billet tracking frames. Each frame includes at least the following fields: sampling timestamp, source number, valid marker, quality field, session number, caliber version number, gate threshold version number, and threshold source identifier type code. The threshold source identifier type code is taken from the gate threshold table of the frozen session. To align the observation window according to the output event, the sampling frame undergoes time base unification before admission judgment. The time base parameter is taken from the session freeze item, mapping the timestamps from various sources to the same gateway clock. The alignment error is allowed to be between 2 and 10 seconds, for example, 5 seconds, with the upper limit of the error derived from the gating threshold table of the session freeze. When the alignment error exceeds the upper limit, the sampling frame is marked as time abnormal and transferred to the conservative channel, but a quality label frame is still generated. The quality label frame includes at least the fields of usable code, time abnormal code, and alarm code, with the usable code set to unusable. This sampling frame is only used for recording and alarming and does not participate in the construction and mapping verification of the observation window, in order to reduce the risk of window mismatch caused by time mismatch. After time alignment is completed, the admission decision process begins. This decision employs a three-tiered gating system: point-level, section-level, and session-level, ensuring that subsequent temperature difference records and burner group action summaries have a verifiable input basis. Point-level gating targets a single temperature measurement point or the furnace exit field of view, using a continuous time window for stability assessment. The time window length is set between 300 and 900 seconds, for example, 600 seconds. The rationale for this setting is to cover at least one step-up and one common cycle disturbance period, filtering out short-term noise and maintaining representativeness. Point-level stability is represented by the effective sampling percentage. An effective sampling percentage below 0.92 is considered unstable and downweighted, while below 0.85 is considered unusable and disabled. Downweighting indicates that the point is not used as the primary source for extracting representative temperature difference values, but only for trend comparison and anomaly alerts. The threshold setting ensures that the number of effective samples within the time window meets the minimum statistical requirements, avoiding misjudgments due to insufficient samples. Point-level gating simultaneously performs occlusion detection, emissivity stability detection, drift detection, and loss detection. Occlusion detection is based on the field-of-view coverage field, which includes at least the effective coverage ratio of the field of view or the converted value of the occlusion marker. A coverage rate below 0.6 is considered occlusion, 0.6 to 0.8 is considered suspicious and only used for trend recording, and a coverage rate above 0.8 is considered usable. The coverage rate threshold is used to meet the minimum effective pixel coverage requirement for representative value extraction. When the emissivity stability marker is continuously unstable for more than 60 seconds, the temperature difference observation for the corresponding time period is marked as unusable to reduce the interference of measurement uncertainty on the temperature difference fluctuation judgment. Drift detection is for long-term offset of temperature measurement points. A continuous unidirectional offset for more than 20 minutes is marked as suspicious, and an offset for more than 60 minutes is marked as confirmed and kept disabled until the maintenance confirmation entry is released. Loss detection is for sampling interruption. Continuous frame loss for more than 3 control cycles is marked as short loss, and frame loss for more than 12 control cycles is marked as long loss and a downgrade suggestion code is given. The reason for this setting is that short-term jitter is tolerable, but long-term loss will destroy the consistency of the closed-loop criteria. Section-level gating is based on the furnace zone, and the gating conclusions of the temperature measurement source and the combustion execution gating conclusions of the zone are used for judgment. The temperature measurement gating conclusion includes at least an availability code, a priority code, and a restricted code. The combustion execution gating conclusion includes at least an execution permission flag, an execution frame integrity flag, and a saturation flag. When the primary priority temperature measurement point is unavailable or the secondary priority temperature measurement point is available, the zone is marked as restricted availability, and the evidence level is lowered in subsequent observation windows. When there is no combustion execution frame or the execution permission flag is not present, the action candidates for the zone are cleared and a prohibition reason code is written. The control output remains at the previous valid value, and other available zones are selected. When the execution saturation flag is true, the corresponding burner group is written to the saturation list. Subsequent actions are selected to avoid the burner group or only allow the reduction action. The judgment criteria are the saturation criterion field in the actuator feedback field or the gating threshold table. The session-level gating integrated furnace condition frame includes fields such as furnace pressure fluctuation level code, safety status, and alarm set. The furnace pressure fluctuation level can be directly given by the furnace condition frame or determined by the furnace pressure fluctuation amplitude and duration according to the level mapping field of the gating threshold table. The furnace pressure fluctuation amplitude is the maximum deviation within the most recent 60 to 300 seconds, and the duration is the cumulative duration of continuous exceedances. When the furnace pressure fluctuation level exceeds the session freeze limit, it enters the degradation strategy code entry, suspends the probe window and version update, and limits the action amplitude to level one or maintains it. If the alarm level is severe or the execution permission is marked as no, it will maintain it; otherwise, it will be limited to level one. The alarm set is classified according to the alarm code and level mapping table. The alarm code and level mapping table is written into the gating threshold table and frozen with the session. When a severe level occurs, the boundary gating is tightened and the degradation strategy code entry is entered. The billet tracking frame includes at least the billet number, timestamp, position index, and segment index. When the billet position advances backward or skips segments, the corresponding time period is marked as a tracking anomaly, and this time period is prohibited from participating in the observation window construction to avoid misbinding of the billet number and segment action summary. The admission determination generates a quality label frame, which maintains the same timestamp as each control cycle. The quality label frame contains at least the following fields: available code, occlusion code, drift code, missing code, alarm code, and degradation suggestion code. Extended fields are written into the segmented zone restricted list, saturation list, and time anomaly list. The gate threshold version number and the time reference parameter version number are recorded to associate the admission caliber and time alignment caliber. The quality label frame serves as the input for the subsequent observation window aligned with the furnace exit event, ensuring that the observation window generates temperature difference records and action summaries only when the input meets the consistency conditions, providing input for mapping verification. Anomalies and boundary handling include at least time anomalies, sampling mutations, and point conflicts. Time anomalies are handled according to conservative channel rules, and the action step is limited to one level. Sampling mutations are judged as anomalies when the single-cycle jump exceeds 20% of the sensor range. The range is taken from the extended field of the temperature measurement mapping table or the point threshold sub-table of the gating threshold table. Anomalies are not included in the admission statistics and mutation events are registered. Point conflicts are judged as conflicts when the difference between adjacent points exceeds 50°C and lasts for 3 control cycles. This threshold is higher than the upper limit of the target temperature difference for session freezing and is used to distinguish between the real temperature difference and point anomalies. In case of a conflict, the higher priority data source is retained and the other point is downgraded. In terms of resource constraints, the single-cycle admission judgment processing delay is set to no more than 200 milliseconds, so that the processing time accounts for no more than 5% of the control cycle. If it exceeds this, the previous control output is maintained and a timeout event is registered.

[0019] S3: Establish an observation window using the furnace exit event as an anchor, generate temperature difference records in the length and width directions, summarize the burner group actions within the window, and complete consistency verification. The specific implementation is as follows: After the data admission judgment is passed, the observation window is constructed and the consistency is verified. The boundary of the observation window is the furnace exit event. The input includes furnace exit event entries, billet tracking entries, admission standard frames, combustion execution frames, zone temperature frames and furnace exit temperature measurement frames. The furnace exit event entries include billet number, furnace exit timestamp, queue position, steel grade specification code field, session number and caliber version number. The observation window ends at the furnace exit timestamp and traces backward to cover the critical end section before furnace exit. The range of the critical end section is taken from the furnace area division table frozen in the session or the list of critical end sections in its extended fields. The window length is set from 10 minutes to 60 minutes, for example, 30 minutes. The reason for this setting is that the impact of the end heating section and the soaking section on the furnace exit temperature difference has a lag characteristic, and the window needs to cover at least one typical dwell time of the soaking section while taking into account cycle fluctuations. The upper limit of the window length is constrained by the heating regime and cycle configuration. The start and end boundaries of the window are taken as the furnace timestamp minus the window length and the furnace exit timestamp. The sampling frames within the window are aligned with the previous unified time reference. The window validity gating directly references the admission standard frame. The system includes fields such as available code, obstruction code, emissivity stability code, loss code, and alarm code, and uses the same gating threshold version number. It also verifies that the billet tracking entry advances monotonically within the window segment. Monotony determination is based on the segment number or step number, allowing no more than one segment of jitter per cycle. Continuous regression or segment skipping is considered a tracking anomaly. If any of the following occurs: obstruction, unstable emissivity, long loss, severe alarm, or tracking anomaly, the window is registered as invalid, and an invalidity reason code is written before verification stops to prevent invalid windows from being included in the evidence chain. The invalidity reason code includes at least the following categories: obstruction, unstable emissivity, long loss, severe alarm, tracking anomaly, furnace pressure exceeding limits, and excessive trimming. When a window is active, a temperature difference record entry is generated. This entry uses the billet number as the primary key and includes at least the furnace exit timestamp, length-direction temperature difference, width-direction temperature difference, temperature difference quality grade, and window number. The sampling area for the length-direction temperature difference is determined by the proportion of the session-frozen area. The coverage ratios for the head and tail regions are each set to 0.10 to 0.20, for example, 0.15. This setting avoids edge radiation distortion and occlusion of sensitive areas while maintaining the representativeness of the head-tail difference. The representative temperature values ​​for the head and tail regions are determined using the median value of three consecutive control cycles. The control cycle is taken from the session-frozen value of the preceding admission process; for example, a 5-second control cycle corresponds to a 15-second time span. This setting is to suppress transient noise and also... The response to the disturbance is considered; the temperature difference in the width direction is formed by taking representative values ​​from the left, middle and right bands to form the difference. The coverage ratio of the left and right bands is each taken as 0.20 to 0.35, for example 0.25, and the middle band takes the remaining ratio. The reason for this setting is to be consistent with the furnace width band adjustment object so as to associate the corresponding burner group action; the temperature difference quality level is determined by the combination of the occlusion code, emissivity stability code, drift code and alarm code of the admission standard frame, and is divided into three levels: A, B and C. The value of A can be used for version update and probe verification, the value of B can be used for closed-loop adjustment but does not enter the version update evidence, and the value of C is only recorded and does not drive control; the quality level judgment rules and improvement thresholds refer to the session frozen gating threshold table and its version number. The source identifier type code of the threshold item is taken from the gating threshold table; An action summary entry is generated within the window. The action summary uses the burner group number as an index and organizes the changes in the combustion execution frame within the window into action fragments. Each action fragment includes at least the following fields: start timestamp, end timestamp, action type code, amplitude level, saturation flag, and gating clipping flag. The action type code takes values ​​from four categories within the allowed adjustment type set of the session freeze: gas adjustment increase / decrease, combustion air adjustment increase / decrease, reversal or pulse beat speed, and opening / closing or switching to a conservative position. The amplitude level is divided into five levels and mapped to the session freeze step table. For example, level one corresponds to a gas adjustment change of 1 to 3 control units, level two to 3 to 6 control units, and level three to 6 to 10 control units. The reason for this setting is... Match the minimum resolvable adjustment amount of the actuator and reduce the risk of overshoot; the gear boundary assignment is determined according to the gear mapping rules of the step table, and the boundary point is not repeatedly assigned; the saturation mark is derived from the combustion execution frame, and the gating clipping mark is derived from the gating record, used to distinguish between active adjustment and clipped actions and to remove incomparable segments; the action summary is recorded in units of action change segments, and segments are not repeatedly generated within the minimum stable holding time of the same burner group. The number of action segments is controlled at an upper limit of 2 to 6 per minute per burner group, for example, 4. The upper limit setting is matched with the control cycle and the minimum stable holding time to control the scale of evidence; the window length, action summary upper limit and hysteresis level classification are given by the session freeze item and recorded with the version number; The consistency verification logically determines the correspondence between temperature difference records and action summaries within a window. This determination includes at least temporal consistency, directional consistency, and boundary consistency. Temporal consistency uses a lag level table from the session freeze, mapping the burner group's section to three lag levels: short lag (2-10 minutes), medium lag (10-25 minutes), and long lag (25-60 minutes). This is because the influence path and dwell time from different sections to the furnace outlet vary. Lag level boundary points are assigned according to the lag level table, with the allowable range trimmed by the window's start and end boundaries; portions exceeding the window are not included in the determination. The determination requires that the action segment's end time and the temperature difference trend change time fall within the corresponding lag level's allowable range. Within a given range, the time of temperature difference trend change is taken as the first time stamp of the continuous billet records with temperature difference quality level A or B within the window that meets the improvement or deterioration criteria upon exiting the furnace; directional consistency is verified based on the action type code and the temperature difference trend marker, which is formed based on the temperature difference records of the most recent 3 to 8 billets within the window; the improvement and deterioration criteria reference the session freeze threshold, with the improvement threshold set to 2 to 6℃, for example, 3℃, because it is of the same order of magnitude as the temperature measurement uncertainty and can distinguish between random fluctuations and effective responses; boundary consistency requires that the furnace pressure fluctuation level within the window does not exceed the session freeze upper limit, there are no serious alarms, and the action segment is not clipped below the minimum stability boundary; if boundary consistency fails, the window is registered as invalid for verification and no consistency conclusion is output. Verification outputs verification record entries. Verification record fields must include at least the window number, participating burner group set, consistency conclusion code, invalidity reason code, and availability level. They can also be expanded to record temperature difference quality level and major invalidation markers. The consistency conclusion code must include at least categories such as pass, fail, and pending. The availability level must include at least levels such as adjustable, updateable, and record-only. Verification records are associated with the action summary and temperature difference record via the window number, used to reference the evidence number and window number when determining subsequent adjustment actions. In the case of occlusion with a temperature difference quality level of C, the window is deemed invalid, and the occlusion reason code is recorded. When a billet tracking anomaly causes the furnace exit event to be unable to be associated with the end section, window generation is paused, and the tracking recovery process is triggered. Verification records are not output before recovery to avoid misbinding. Regarding time and resource constraints, window construction and verification are completed within 5 to 30 seconds after the burner is removed from the furnace, for example, 10 seconds. The reason for this setting is that the control cycle needs to promptly reference the latest temperature difference status and verification conclusions. The sampling frames and action segments within the window are organized using a rolling cache and incremental aggregation method. The rolling cache content includes at least the combustion execution frame summary index, action segment index, and corresponding timestamp fields within the window length coverage area. The index key value is a combination of the burner group number and the start and end boundaries of the time period, which is used to avoid full rescanning based on the burner removal event. The memory usage of a single verification is constrained by the rolling window strategy. When the number of action segments exceeds the cache limit, priority is given to retaining the critical segments at the end and short-latency segments, and the pruning mark and pruning reason code are recorded to maintain real-time performance and evidence representativeness.

[0020] S4: Based on the temperature difference status, the mapped version is called to determine the adjustment action, boundary gating is performed, and the action is issued, and the observation window and evidence number are associated. The specific implementation is as follows: After temperature difference recording and consistency verification are completed, the adjustment action determination and boundary gating issuance proceed. The input data includes at least the latest temperature difference status entry, mapping version entry, action candidate table, gating threshold table, burner group boundary table, quality label frame, furnace condition frame, and zone temperature frame. Each entry has a session number and version number referencing the session freeze caliber. The gating threshold table and burner group boundary table constrain the action amplitude and executable boundary, and the control output is associated with the observation window number and verification record number. The latest temperature difference status entry is formed by summarizing the temperature difference records of the most recent N billets. The value of N ranges from 3 to 8, for example, 5. The reason for setting this value is to suppress occasional anomalies and maintain response speed under cycle disturbances. The value range and default value of N are given by the session-frozen gating threshold table or the status summary configuration item and are recorded with the version number. The status summary configuration item fields include at least the range of the number of summary samples, the length of the trend formation window, the reference identifiers for improvement and deterioration thresholds, and the number of cycles for trend stability determination. Temperature difference status types include length overlimit, width overlimit, and bidirectional overlimit. The system categorizes actions into five levels: insufficient convergence, normal convergence, and inadequate convergence, each carrying an over-limit level and a trend indicator. The over-limit level is determined by the session freeze gating threshold table, for example, Level 1 (0-3℃), Level 2 (3-6℃), Level 3 (6-10℃), and Level 4 (over 10℃). The classification boundary rules are given by the gating threshold table, and boundary points are not assigned repeatedly. Level 1 over-limit is used for prompts and trend constraints, and the action amplitude does not exceed one level. The trend indicator values ​​include three categories: improvement, stagnation, and deterioration, recorded according to the session freeze threshold and the version number, and are used to limit the number and amplitude of actions. The mapping version entry is indexed by the version number and includes at least the following fields: priority, lag level, and set of allowed action types for the burner group's impact on the temperature difference in the length and width directions. Allowed action types include fields such as gas regulation increase / decrease, combustion air regulation increase / decrease, cycle speed, and burner group opening / closing switching from the set of allowed adjustment types for session freeze. The impact priority is an enumerated field, ranging from level one to five, with lower priority preferred. Levels one to three are selected first, and levels four to five are selected when there is continuous exceedance and the available level of the verification record meets the freezing conditions of the gate threshold table. The lag level maps the burner group's segment to three levels: short lag, medium lag, and long lag, based on the lag level table for session freeze, and provides an identifier for the allowed intervention stage within each level. Stage determination includes at least the temperature difference state type and trend marker: bidirectional exceedance or continuous exceedance corresponds to the convergence stage; temperature difference entering the target range and trend stabilization corresponds to the compensation stage; and normal temperature difference with improving or stable trends corresponds to the steady-state stage. Stage criteria are derived from the frozen items in the gate threshold table and recorded with the version number, constraining simultaneous intervention in multiple segments. The action candidate list is arranged in order. First, burner groups with the same direction as the target and an impact priority of 1-3 are selected based on the temperature difference status. Short-latency and medium-latency burner groups are preferred based on the lag level. Unadjustable objects are eliminated by combining the section temperature margin level and the execution saturation flag. The section temperature margin level is determined based on the margin range of the zone temperature frame relative to the process upper and lower limits. The section temperature margin level is divided into three levels: sufficient, limited, and insufficient. The process upper limit, process lower limit, and margin classification threshold data come from the process boundary configuration items and version number records frozen in the session. The process boundary configuration item fields include at least the process upper limit, process lower limit, margin classification threshold, applicable heating system code, and effective boundary fields of the section. Actions with insufficient margin level cannot be increased in load, but can only be reduced in load or maintained. The reason for this setting is to prevent temperature difference-driven actions from exceeding the process boundary. The execution saturation flag is generated by the combustion execution frame. When the execution saturation flag is true, the action type of the burner group is reduced or maintained, and the priority of this type of action in the action candidate list is reduced by one level. The number of burner groups allowed to be issued simultaneously in each control cycle is 1 to 4, for example, 2. This upper limit is given by the session freeze gating threshold table and recorded with the version number. It is used to reduce multi-actuator coupling and facilitate verification and attribution. The action amplitude adopts a small step size, with the default being level 1 or 2. When the temperature difference status is insufficient convergence and simultaneously meets the following conditions: furnace pressure fluctuation level is lower than the session freeze upper limit, alarm level is not higher than the concern level, temperature difference quality level is A or B, gating record has no freeze mark and no mark indicating that it has been trimmed to the minimum stability boundary, level 3 is allowed, and level 4 or 5 is prohibited. The correspondence between the step size and the actual adjustment amount refers to the step table of the session freeze and is recorded with the version number. The action duration is 10 to 60 seconds, for example, 30 seconds. The duration value is an integer multiple covering 2 to 12 control cycles and is given by the session freeze item, so that the action impact is identifiable and the execution jitter caused by frequent switching is reduced. Boundary gating is executed after the action candidates are determined. The gating sequence is frozen within the session, performing minimum stability boundary checks, maximum safety boundary checks, furnace pressure fluctuation gating, alarm gating, and temperature measurement quality gating in sequence. The minimum stability boundary check is based on the lower limit of the burner group boundary table. If a candidate action causes the gas or combustion air regulation to fall below the lower limit, the range is reduced or changed to hold. The maximum safety boundary check is based on the upper limit. If a candidate action causes an exceedance, it is reduced to the upper limit and the reduction reason code is recorded. The furnace pressure fluctuation gating is based on the furnace pressure fluctuation level of the furnace condition frame. When the level reaches the session freeze upper limit, load increase actions are prohibited, and only load decrease or hold is allowed. The alarm gating is based on the alarm level. If a severe alarm occurs... When the level is high, the action is frozen and a downgrade suggestion code is output. At the same time, the action candidates for this cycle are cleared and the freeze reason code is recorded. The temperature measurement quality gating is based on the quality label frame and the temperature difference quality level. When the temperature difference quality level is C or the occlusion mark of the furnace exit field of view meets the continuous threshold, the temperature difference drive is paused and switched to the partition temperature conservative control channel. The occlusion continuity judgment and the continuous judgment of the temperature difference quality level adopt the control cycle counting caliber. The continuity threshold adopts the number of continuous cycles in the admission judgment freeze item or the gating threshold table. For example, 12 consecutive control cycles or 60 consecutive seconds are judged as continuous occlusion. The values ​​of the relevant thresholds, boundary upper and lower limits and the number of continuous cycles are all from the session freeze table entries and are recorded with the version number. Gated output generates gated record entries. The gated record fields must include at least the following fields: marker, trimming marker, trimming reason code, freeze marker, downgrade suggestion code, and control cycle timestamp. Action-based output generates control output entries. The control output fields must include at least the following fields: timestamp, burner group number, action type, amplitude level, duration, associated observation window number, associated verification record number, and associated temperature difference record identifier. The associated temperature difference record identifier is a combination of billet number and furnace exit timestamp or temperature difference record entry number. The action is associated with the evidence index. Anomalies and boundary handling include three categories: execution rejection, execution saturation, and missing feedback. Execution rejection is determined by whether the execution allow flag is set to false. During this cycle, the current control output is maintained and the rejection event is recorded. Simultaneously, the corresponding burner group is added to the temporary disabled list. The disabled duration ranges from 30 to 300 seconds, for example, 120 seconds. This setting is to cover the common window for a single self-check or communication reconnection in the execution link. The disabled duration is given by the session freeze item and recorded along with the version number. Rejection events must include at least the timestamp of occurrence, burner group number, rejection reason code, and the current gating threshold version number. Execution saturation... The system uses a saturation flag as a true condition to reduce the amplitude level and select the next priority burner group. When continuous saturation reaches the session freeze threshold, a segment-level degradation suggestion code is written. The continuous saturation threshold is set to 6 to 20 control cycles, for example, 10 control cycles. It is frozen by the gating threshold table and recorded with the version number. Feedback loss is determined when the temperature difference quality level is continuously C for more than 5 minutes, and the system enters the degradation process and freezes the version update entry. This 5-minute threshold is used for control strategy degradation and is independent of the loss threshold in the admission determination. The relevant threshold references the session freeze item and is recorded with the version number. Regarding time and resource constraints, action selection and gating issuance are completed within 200 milliseconds in each control cycle. This time limit is set to limit the proportion of processing time in the control cycle and match the on-site execution response. When a timeout occurs, the control output of the previous cycle is used and the timeout event is registered. At the same time, a gating record entry is output for traceability without generating a new action. The timeout event field includes at least the timeout duration, timeout count, affected burner group set, and gating threshold version number, and is written to the audit index.

[0021] S5: When the triggering condition is met, insert a probe window and enter the review window to complete the version switch or rollback and set a prohibition period. In case of an anomaly, enter the downgrade process and return to session execution after the recovery conditions are met. The specific implementation is as follows: During the controlled operation phase of the mapping, trigger entries drive the probe window and the verification window. Trigger entries are frozen in the session and registered with rule numbers. They are used for controlled updates and rollback of the mapping version without increasing the contact temperature measurement setup. Trigger types include continuous over-limit, insufficient convergence, mode switching, and decreased evidence quality, and must simultaneously meet both quantity and quality thresholds. The continuous root count threshold, over-limit level threshold, improvement threshold, stability duration threshold, window quantity threshold, proportion threshold, and duration threshold reference the session-frozen trigger entries and are registered according to the version number. Trigger entries should include fields such as rule number, trigger type code, continuous root count threshold, over-limit level threshold, improvement threshold, trend criterion flag, stability duration threshold, window quantity threshold, proportion threshold, duration threshold, quality threshold, furnace pressure level threshold, effective boundary rule, and start / stop flag. Continuous over-limit triggering uses the continuous billet counting caliber. For example, if 6 consecutive billets exceed the limit in the length or width direction and the over-limit level is not lower than level 2, the temperature difference quality level must be not lower than B. Insufficient convergence triggering is when the temperature difference improvement of 5 consecutive billets is insufficient to meet the session freeze improvement threshold, such as 3°C, and the trend is marked as flat or deteriorating, and the furnace pressure fluctuation level must not exceed level 3. Mode switching triggering is when the switch from heating to homogenization ends and the stable window is entered and the stable duration threshold of session freeze is reached, such as exceeding 10 minutes. The stable window is counted cumulatively according to the control cycle, and the serious alarm count and temperature measurement long loss count are 0 within the stable window. A decline in evidence quality is triggered when, for example, the proportion of temperature difference quality grade A in the most recent 12 observation windows is less than 0.5 and the downward trend continues for more than 30 minutes. The downward trend determination references the frozen trend criterion marker of the trigger entry, such as the proportion monotonically decreasing within a continuous control cycle or the decrease reaching a threshold. The relevant cycle number and magnitude threshold are fixed with the rule number. After the trigger condition is met, a trigger event entry is generated. The fields include at least the rule number, trigger type code, trigger count, set of associated window numbers, current mapping version number, trigger timestamp, current furnace pressure grade, and current quality grade summary, and are written into the audit index for rollback and review reference. After the trigger event is confirmed, the probe window is entered. The probe window length is 2 to 12 minutes, for example, 6 minutes. The window length matches the lag level of the participating burner group and adopts the value given by the trigger entry. Short lag can be 2 to 12 minutes, and medium lag can be near the lower limit of 10 to 25 minutes. The allowed range of lag level refers to the lag level table of session freeze, and the minimum and maximum lag durations are selected according to the segment where the burner group is located. Within the probe window, 1 to 2 burner groups can be selected to perform restricted actions, for example, 1 group. The restricted action amplitude cannot exceed level 2, and the action duration cannot exceed 60 seconds. 30 seconds, and adhere to the boundary gating sequence and trimming rules; the probe window admission conditions adopt stricter gating, the furnace pressure fluctuation level does not exceed level two, the alarm set does not contain the severe level, the temperature difference quality level is not lower than B, the furnace exit field coverage reaches the usable threshold and the emissivity is marked as stable, the billet tracking advances monotonically within the window without skipping segments or retreating; if any admission condition is not met, the probe action is not executed, a probe rejection record is generated and written to the event entry, the rejection record includes at least the rule number, rejection reason code, associated trigger event number, gating threshold version number, quality label summary and other fields; During probe window execution, the actions of the burner group, gating and trimming records, and temperature difference status changes within the window are encapsulated into probe record entries. Each probe record includes at least the following fields: probe number, trigger event number, burner group number set, action summary, gating record, probe window start and end timestamps, pre-probe temperature difference status code, post-probe temperature difference status code, and validity conclusion code. The probe number is generated and archived by combining the session number and an incrementing sequence number. The validity conclusion code is determined by the frozen caliber and includes at least the timing consistency condition, direction consistency condition, and boundary consistency condition. The timing consistency condition is determined by the probe action result. After the beam, within the allowable range specified in the hysteresis level table, the temperature difference status improves or the improvement trend strengthens; the direction consistency condition is that the direction of temperature difference improvement is consistent with the expected direction of the target temperature difference in the mapping version; the boundary consistency condition is that the probe action is not clipped below the minimum stability boundary and there are no serious alarms, no temperature measurement long loss, and no tracking anomalies within the window; the probe record is written into the evidence package index, and the index includes at least the probe number, associated window number, associated verification record number, associated control output number, etc.; the control output number is generated by combining the control output entry number or timestamp with the burner group number and archived with the session number. After the probe record is valid, a candidate mapping version is generated and enters the review window. The candidate mapping version does not take effect until the review is passed. The candidate mapping version only makes minor adjustments to the burner group entries participating in the probe. The adjustment items can be the impact priority and the lag level. The change in the impact priority shall not exceed one level of the original priority. The enumeration order of the lag level is fixed as short lag, medium lag, and long lag. Changes to adjacent levels are only allowed from short to medium or medium to long and their reverse. The number of burner groups involved in a single candidate change shall not exceed 10% of the total number of burner groups, for example, 5%. This percentage threshold is given by the version change strategy item frozen in the session and registered with the version number. The review window is expressed in terms of the number of continuous billets N, where N is 3 to 15, for example, 5. The range and default value of N are frozen by the trigger entry or the gating threshold table and registered with the version number. The review window admission requirements are that the temperature difference quality level is not lower than B and the furnace pressure fluctuation level is not higher than level three. No new probe windows are inserted into the review window. The review and judgment process employs two sets of conditions: target range judgment and fluctuation judgment. The target range judgment requires that the temperature difference status within the review window does not exhibit a sudden increase in deterioration level. The increase criterion uses the level difference threshold from the gating threshold table; for example, an increase of more than one level in deterioration level compared to the previous billet is considered a jump, and the temperature difference status of the billet at the end of the window must not exceed the limit. The fluctuation judgment requires that the temperature difference fluctuation in the same direction within the review window does not exceed the upper limit, which is taken from the gating threshold table, for example, 10℃, and that the fluctuation level does not increase continuously. After the review is passed, the system switches to the candidate mapping version at the version effective boundary. The billet queue boundary is used first, followed by the step event boundary. Version switching is written to the audit index, and the audit index fields include at least the old version number, new version number, trigger reason code, trigger event number, probe number, review window number, effective boundary timestamp, and key threshold snapshot. If the review fails, it will revert to the old version and set a prohibition period, which is 10 minutes to 120 minutes, for example, 30 minutes. The threshold is taken from the trigger entry and recorded with the version number. During the prohibition period, probe windows with the same rule number are prohibited from being triggered again. The normal closed loop runs according to the old version. Conditions for abnormal entry into the degradation process are frozen within the session, including at least the following: probe window gating failure, serious alarm in the review window, persistent long temperature measurement loss, continuous execution rejection, and furnace pressure fluctuation level exceeding the upper limit. After entering the degradation process, temperature difference driven actions are suspended, conservative control of zone temperature is maintained, and the action step is limited to one level. At the same time, version updates and probe triggers are frozen. Recovery conditions are frozen within the session, including at least the following: temperature difference quality level recovers to B or above and lasts for 180 seconds, furnace pressure fluctuation level drops to level three or below and lasts for 300 seconds, serious alarm is cleared, and there is a confirmation record. The confirmation record is the confirmation field of the event entry, including at least the confirmation timestamp, confirmation type code, work order number, or confirmer identifier, and is associated with the alarm code. Evidence packages are retained for 72 hours on a rolling basis. Trigger events, probe records, review records, and version switch records are given high priority and associated with the audit index. When overwriting, the probe number and version switch index key are retained first.

[0022] In this embodiment, the billet reheating furnace initiates a session when changing production or transitioning to homogenization. A session number is generated, consisting of a timestamp, furnace number, and shift sequence number. Within the session, the furnace area division, width-wise zoning, burner group numbering, temperature measurement mapping, homogenization index caliber, gate threshold, and version validity boundary are frozen. During operation, zone temperature measurement, furnace exit temperature measurement, combustion execution, furnace condition, and billet tracking data are collected according to a fixed control cycle. After unifying the time base, point-level, section-level, and session-level access judgments are performed, recording obstruction, drift, loss, and alarm statuses, and generating quality labels. When each billet exits the furnace, an observation window is constructed using the exit time as an anchor. The length and width temperature differences are extracted according to the frozen caliber, and the data from each burner group within the window are summarized. The action segment completes the consistency verification of timing, direction, and boundary, forming a window number and verification number; the control side obtains the temperature difference status based on the recent temperature difference of multiple steel billets, calls the mapped version to select a small group of burners to generate adjustment actions, and issues them after passing the minimum stability boundary, maximum safety boundary, furnace pressure fluctuation, alarm and temperature measurement quality gating, and associates the control output with the window number and verification number; when the continuous over-limit, insufficient convergence, mode switching, or evidence quality decline reaches the trigger threshold, a probe window is inserted to execute a restricted small-step disturbance to generate a candidate version, which is switched according to the effective boundary after passing the review window, and if it fails, it rolls back and enters the prohibition period; when an anomaly occurs, it enters the degraded amplitude limit operation, and returns to the session operation after the recovery conditions are met.

[0023] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0024] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above embodiments can be implemented in whole or in part by a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions of the embodiments of this application are implemented in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted wirelessly or wiredly from one website, computer, server, or data center to another website, computer, server, or data center. Wired methods include optical fiber, twisted pair, coaxial cable, etc. Wireless methods include infrared, microwave, etc. Available media include any available media that can be accessed by a computer or data storage devices such as servers and data centers that contain one or more sets of available media. Available media can be magnetic media (floppy disks, hard disks, magnetic tapes), optical media (DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0025] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0026] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A closed-loop control method for uniform combustion temperature of steel billets in porous structures, characterized in that, include: S1: Establish a session to define the freezer zone rules, burner group numbers, temperature measurement point mapping, uniform temperature index caliber, and version effective boundaries. S2: Unify the time base of zone temperature measurement, furnace exit temperature measurement, combustion execution quantity and furnace condition data, make access judgment on data validity and record the status of occlusion, drift, loss and alarm. S3: Establish an observation window based on the furnace exit event, generate records of temperature difference in the length and width directions, summarize the burner group actions within the window and complete the consistency verification. S4: Based on the temperature difference status, call the mapping version to determine the adjustment action, perform boundary gating and then issue it, and associate the observation window with the evidence number; S5: When the triggering conditions are met, insert a probe window and enter the review window to complete the version switch or rollback and set the prohibition period. In case of an anomaly, enter the downgrade process and return to the session after the recovery conditions are met.

2. The closed-loop control method for uniform combustion temperature of steel billets according to claim 1, characterized in that, Establish a session to distinguish the freezer zone rules, burner group numbers, temperature measurement point mapping, uniform temperature index caliber, and version validity boundaries, including: The session number is generated from the timestamp, furnace number, and shift sequence number; The session configuration entry is assigned a version number, written to the session snapshot, and registered in the audit index; Before freezing, perform consistency and integrity checks. If a test is missing, mark the observation as restricted and disable probe permissions. If the boundary is abnormal, switch to conservative actions. If session establishment fails, the previous version will be used and the reason code will be logged.

3. The closed-loop control method for uniform combustion temperature of steel billets in porous structures according to claim 1, characterized in that, The time base for zonal temperature measurement, furnace exit temperature measurement, combustion execution quantity, and furnace condition data will be unified, including: The sampling frames are accessed in a rolling manner according to a fixed control cycle and carry the session number, caliber version number, gating threshold version number and threshold source identifier type code; The time base parameters are taken from the session freeze item, and the timestamps from each source are aligned to the gateway clock; If the alignment error exceeds the limit, mark the time as abnormal and switch to the conservative channel; if the processing timeout occurs, record the timeout event.

4. The closed-loop control method for uniform combustion temperature of steel billets according to claim 1, characterized in that, The system performs an admissions test on data validity and records occlusion, drift, loss, and alarm status, including: Access determination employs point-level, segment-level, and session-level gating and outputs quality-labeled frames. Point-level classification is based on the effective sampling ratio and field of view coverage, combined with emissivity stability, drift, and loss criteria. At the section level, a restricted list is generated based on the execution permission and saturation flags; At the session level, a degradation strategy code is triggered based on the furnace pressure fluctuation level and alarm mapping, and abnormal events such as sampling mutations and point conflicts are registered.

5. The closed-loop control method for uniform combustion temperature of steel billets in porous structures according to claim 1, characterized in that, An observation window is established using the furnace exit event as an anchor, generating records of temperature differences along the length and width, including: The observation window ends at the release timestamp and traces back according to the list of key segments at the end of the session freeze; Window validity gating references the admission standard frame and verifies tracking to advance monotonicity; When the window is valid, representative values ​​are extracted from the head and tail regions and the left, middle and right regions according to the frozen sampling ratio. The head and tail representative values ​​are taken as the median value of several consecutive control cycles, and the length temperature difference and width temperature difference records are generated and assigned temperature difference quality level.

6. The closed-loop control method for uniform combustion temperature of steel billets according to claim 1, characterized in that, Summarize the burner group actions in the summary window and complete the consistency verification, including: The action summary organizes the combustion execution changes into non-repeating action segments within the minimum stable holding time according to the burner group number, and records the action type code, amplitude level, saturation mark and gating clipping mark; Consistency verification uses a lag level table and improvement thresholds to determine the timing, direction, and boundaries. If the verification is invalid, write the invalidity reason code, and associate the verification record with the window number.

7. The closed-loop control method for uniform combustion temperature of steel billets in porous structures according to claim 1, characterized in that, Based on the temperature difference status, the mapping version is invoked to determine the adjustment action, which is then issued after boundary gating and associated with the observation window and evidence number, including: The temperature difference status is formed by summing the near-end temperature differences and carries the exceedance level and trend indicator; Based on the priority and lag level of the mapped version, limited action candidates are generated and filtered out using small steps, combined with the section temperature margin and execution saturation. Candidate actions are cut or frozen and then issued after being subject to stability lower limit, safety upper limit, furnace pressure, alarm and temperature measurement quality gating; The control output is associated with the window number and verification number, and the execution rejection, feedback missing and timeout events are registered as disabled, downgraded and timeout events respectively.

8. The closed-loop control method for uniform combustion temperature of steel billets in porous structures according to claim 1, characterized in that, When the triggering condition is met, insert a probe window and enter the verification window, including: The trigger entry is frozen within the session and the rule number is registered. The trigger type is continuous over-limit, insufficient convergence, mode switching or evidence quality degradation, and it must simultaneously meet the quantity threshold and the quality threshold. Once the trigger is established, a limited number of burner groups are selected to perform restricted small-step perturbations. The probe window duration is set according to the hysteresis level, and probe records are generated and registered in the evidence index under stricter gating conditions.

9. The closed-loop control method for uniform combustion temperature of steel billets in porous structures according to claim 1, characterized in that, After completing the version switch or rollback and setting a prohibition period, in case of an anomaly, the process enters the downgrade procedure and returns to session execution after the recovery conditions are met, including: When the probe records meet the conditions of consistent timing, consistent direction, and consistent boundaries, candidate mapping versions are generated only for the burner group entries participating in the probe within the range of adjacent lag levels or adjacent priority levels. Switching or rollback is performed at the effective boundary of the version, a prohibition period is set, and the same rule number is prohibited from being triggered again. When the probe or verification fails, or when a serious alarm occurs, temperature measurement is lost for a long time, execution is refused continuously, or the furnace pressure fluctuation level exceeds the limit, the process will switch to limited downgrading and will exit downgrading after the quality level is restored, the furnace pressure level is restored, and the confirmation record is valid.

Images