A method and device for heat treatment of high-strength steel plates for ocean engineering

Abstract

The application discloses a high-strength steel plate heat treatment method and device for ocean engineering, stress and thickness coupling mapping is carried out on the normalized steel plate, phase change stress distribution is constructed and stress concentration area is located, and a steady-state holding system is established according to the partition differential temperature heating and pre-compensation heating control, so that the fixed process set according to the grade is replaced; the cooling stage synchronously tracks the surface and thickness direction internal cooling rate, analyzes whether the starting time of each layer phase change is coincident, and identifies and classifies the crack sensitive area according to the above; directional metallographic examination and low temperature impact test are carried out on the above area, the performance difference in the thickness direction and the rolling direction is measured, and the process window boundary is adjusted after convergence confirmation according to the historical failure interval and the cooling phase change behavior difference, so as to generate the next heating correction instruction, improve the adaptability of the heat treatment process to the actual state of different batches of steel plates, and improve the control level of the steel plate residual stress and the stability of the mechanical properties in the thickness direction.

Description

Technical Field

[0001] This invention relates to the field of heat treatment technology for metallic materials, and in particular to a heat treatment method and apparatus for high-strength steel plates used in marine engineering. Background Technology

[0002] Marine engineering equipment operates for extended periods in environments with high salt spray, low temperatures, and alternating loads. The high-strength steel plates used not only require high strength and low-temperature impact toughness, but also properties in the thickness direction (Z-direction) that are comparable to those in the rolling direction. Existing steel plates typically obtain the required microstructure through raw material testing, normalizing or quenching, and controlled cooling. The heating, holding, and cooling parameters at each stage are mostly executed according to fixed processes pre-set for the steel grade.

[0003] However, the raw material composition and thickness of steel plates of the same grade vary between different batches. Fixed process parameters cannot simultaneously account for the microstructure transformation rhythm of different areas of the plate. During the cooling stage, uneven temperature distribution can easily lead to areas with high local internal stress, creating potential cracking hazards. Uneven microstructure in the thickness direction also results in significant differences in mechanical properties between this direction and the rolling direction. Current production relies heavily on sampling inspection after heat treatment to detect defects, but the inspection results are difficult to feed back to previous processes, and similar problems repeatedly occur in subsequent batches. Summary of the Invention

[0004] This invention provides a heat treatment method and apparatus for high-strength steel plates used in marine engineering. By performing stress-thickness coupling mapping on the steel plate to locate the residual stress concentration area and implement zoned differential temperature heating, combined with the coincidence analysis of the phase transformation time sequence in the thickness direction during the cooling stage, the potential crack location is predicted and classified. Then, based on the anisotropy test results and historical failure data, the process window boundary is dynamically corrected in a convergent manner and fed back to the heating parameters of the next heat, forming a closed-loop process control system in which detection, heating, cooling and performance evaluation are mutually fed back.

[0005] The first aspect of this invention provides a heat treatment method for high-strength steel plates used in marine engineering, comprising the following steps:

[0006] Based on the detection of chemical composition and thickness distribution of steel plate raw materials, raw material specification data is obtained. Normalizing pretreatment is carried out according to the raw material specification data to generate normalized structure. The phase transformation stress distribution is determined by analyzing the residual stress state of the normalized structure.

[0007] The phase change stress distribution is used to locate the stress concentration area and determine the differential temperature heating regime. The heat dissipation lag trend of the differential temperature heating regime is analyzed to obtain the differential temperature heat compensation increment. The differential temperature heat compensation increment is used to carry out pre-compensation heating control to establish a steady-state heat preservation regime.

[0008] The surface cooling rate is obtained by segmented cooling control according to the steady-state heat preservation system. The core cooling rate gradient is calculated based on the surface cooling rate to obtain the core-surface cooling rate difference. Crack-sensitive areas are output from the core-surface cooling rate difference to screen the areas where phase transformation stress is superimposed and concentrated.

[0009] Grain size level is obtained by directional inspection of metallographic structure in the crack-sensitive area. Abnormal points with low impact toughness are screened for the grain size level and classified into evaluation to obtain low temperature impact performance value. Anisotropy coefficient is obtained by comparing the difference between Z direction and rolling direction based on the low temperature impact performance value.

[0010] Based on the anisotropy coefficient, the failure risk range is formed by tracing the failure ranges of each failure. The process window boundary value is obtained by dynamically tracking the differences in cooling phase change behavior according to the failure risk range. The degree of deviation of heating parameters is calculated based on the process window boundary value, and a heating correction command is output.

[0011] A second aspect of the present invention provides a heat treatment apparatus for high-strength steel plates used in marine engineering, comprising:

[0012] The stress analysis module is used to obtain raw material specification data based on the chemical composition and thickness distribution of steel plate raw materials, perform normalizing pretreatment based on the raw material specification data to generate normalized structure, and determine the phase transformation stress distribution by analyzing the residual stress state of the normalized structure.

[0013] The heat compensation control module is used to locate the stress concentration area by the phase change stress distribution to determine the differential temperature heating regime, analyze the heat dissipation lag trend of the differential temperature heating regime to obtain the differential temperature heat compensation increment, and use the differential temperature heat compensation increment to carry out pre-compensation heating control to establish a steady-state heat preservation regime.

[0014] The crack identification module is used to obtain the surface cooling rate by performing segmented cooling control according to the steady-state heat preservation system, calculate the core cooling rate gradient based on the surface cooling rate to obtain the core-surface cooling rate difference, and screen out the areas where phase transformation stress is superimposed and concentrated from the core-surface cooling rate difference to output the crack sensitive area.

[0015] The performance evaluation module is used to obtain the grain size level by directional inspection of the metallographic structure in the crack-sensitive area, screen out abnormal points with low impact toughness based on the grain size level and classify them into the evaluation to obtain the low temperature impact performance value, and obtain the anisotropy coefficient by comparing the difference between the Z direction and the rolling direction based on the low temperature impact performance value.

[0016] The feedback correction module is used to trace the failure intervals of each failure based on the anisotropy coefficient to form a failure risk interval, dynamically track the differences in cooling phase change behavior based on the failure risk interval to obtain the process window boundary value, and calculate the degree of deviation of heating parameters based on the process window boundary value to output a heating correction command.

[0017] The beneficial effects of this invention are reflected in the following points: First, this invention does not uniformly set the heating process based on the steel grade, but instead first performs stress-thickness coupling mapping on the normalized steel plate to construct a phase transformation stress distribution that reflects the actual distribution of residual stress on the plate surface. Based on this, specific areas with real stress concentration risks are located, and then these areas are subjected to zoned differential temperature heating and pre-compensation heating control, so that heating resources are concentrated on the risk areas themselves, rather than applying the same set of heating parameters to the entire plate. This reduces the probability of residual stress concentration during the heat treatment stage. Second, during the cooling stage, not only is the cooling rate of the steel plate surface and the interior in the thickness direction tracked, but the starting time of each layer of phase transformation in the thickness direction is further analyzed to see if they coincide. The degree of synchronization in the phase transformation sequence is used as the basis for judging the risk of local stress superposition. This judgment dimension is different from the traditional method of evaluating quality based solely on differences in cooling rates or temperature gradients. It can identify and classify potential crack risk locations before the steel plate is taken off the production line, shifting quality control from sampling inspection after heat treatment to the production process. Finally, the performance differences in the thickness direction and rolling direction obtained by directional metallographic inspection and low-temperature impact test are compared with the failure ranges of this batch and historical batches. The adjustment weight of the process window boundary is determined by quantitative calculation of the difference in cooling phase transformation behavior. The boundary value is required to be confirmed only after the fluctuation range of multiple consecutive batches tends to stabilize, so as to avoid repeated changes in the boundary due to the random fluctuation of a single batch. The heating correction command generated accordingly can more accurately reflect the actual risk level of this type of process parameter, improve the pertinence of heating parameter correction, and improve the reliability of process window boundary setting. Attached Figure Description

[0018] Figure 1 This is a schematic flowchart of a heat treatment method for high-strength steel plates for marine engineering according to the present invention.

[0019] Figure 2 This is a schematic diagram of the heating, cooling, and temperature measurement layout of the heat treatment device of the present invention.

[0020] Figure 3 This is a structural block diagram of a heat treatment device for high-strength steel plates used in marine engineering according to the present invention.

[0021] Wherein: 1-Steel plate body; 2-Edge heating unit; 3-Transition heating unit; 4-Central heating unit; 5-Surface temperature measuring thermocouple; 6-One-quarter plate thickness temperature measuring thermocouple; 7-Core temperature measuring thermocouple; 8-Water mist strong cooling spray device; 9-Air slow cooling zone; 10-Controller. Detailed Implementation

[0022] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0023] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0024] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0025] The technical solutions of the embodiments of this application are described below.

[0026] like Figure 1 As shown, this embodiment of the invention provides a heat treatment method for high-strength steel plates used in marine engineering, including the following steps S101-S105:

[0027] Step S101: Obtain raw material specification data based on the chemical composition and thickness distribution of the steel plate raw material, perform normalizing pretreatment based on the raw material specification data to generate normalized structure, and determine the phase transformation stress distribution by analyzing the residual stress state of the normalized structure.

[0028] Specifically, raw material specification data is obtained based on the detection of chemical composition and thickness distribution of steel plate raw materials. The detection of chemical composition and thickness distribution are the direct targets of this step, and their results together constitute two sets of fields for the raw material specification data. After the steel plate raw materials are stored, the detection areas at both ends and the middle of the rolling direction are first ground with a grinding wheel to remove oxide scale and rolling oil stains. Chemical composition detection uses a direct-reading spectrometer to excite discharge in the ground area to determine the content of seven elements: carbon, manganese, chromium, molybdenum, vanadium, nickel, and copper. Among these, the carbon equivalent and the total amount of alloying elements determine the phase transformation temperature range and microstructure transformation rate of the steel grade, which are the core inputs for subsequently determining the normalizing temperature. The accuracy of chemical composition detection directly affects the reliability of this core input; even slight differences in chemical composition between different heat numbers will be reflected in different graded values ​​in the normalizing temperature reference table. Thickness distribution detection uses an ultrasonic thickness gauge, dividing the grid into five equal parts vertically and horizontally, and scanning point by point. The deviation between the measured thickness and the design thickness at each point is recorded. Points with poor coupling are automatically re-measured; those still failing the re-measurement are marked as blind zones. If the number of blind zones exceeds 10% of the total grid points, the thickness distribution detection for that sheet is considered incomplete and requires manual re-inspection. The completeness of the thickness distribution directly determines whether the next step of stress-thickness coupling mapping can be successfully carried out. After the chemical composition and thickness distribution detection are completed, the raw material specification data is archived according to two sets of fields: element content and thickness deviation. Multiple steel plates from the same batch share one element content record, while each retains its own independent thickness deviation record. If the content of any element in the raw material specification data exceeds the upper or lower limit for that steel grade, a corresponding component exceedance indicator is displayed, indicating that the composition sheet for that batch needs to be manually reviewed.

[0029] Normalizing pretreatment is performed based on raw material specifications to generate a normalized microstructure. The element content in the raw material specifications determines the heating temperature, and the thickness deviation determines the holding time. The heating temperature is determined by referring to the temperature comparison table in the normalizing process verification record for this steel grade, based on the element content combination. Combinations with higher element content and higher phase transformation points are assigned a higher temperature to ensure sufficient austenitization. When the same combination crosses the boundary of two grade intervals, a higher grade is used to prevent insufficient microstructure transformation from causing distortion in subsequent stress analysis. The holding time is determined by multiplying the maximum measured thickness in the raw material specifications by the holding coefficient per unit thickness. The larger the thickness, the slower the core reaches the set temperature, and the higher the coefficient value is, ensuring that the core and surface of the plate complete austenitization simultaneously. The heating rate is limited to within 80% of the rated rate to keep the temperature difference between the surface and the core within the allowable temperature gradient zone for this steel grade, avoiding excessively rapid heating and pre-embedded thermal stress in the thickness direction. After heat treatment, the steel is transferred to air cooling, and the start and end times of cooling are recorded. No active intervention is made during the cooling process to obtain the natural microstructure of the normalized steel grade. The cooling rate is also not actively intervened; only the start and end times of cooling are recorded for archiving and traceability. The grain size level of the normalized microstructure is determined by metallographic examination of the furnace-batch samples, exhibiting a uniform mixture of ferrite and pearlite. The grain size level is associated with the elemental content and thickness deviation in the raw material specification data according to the same batch number. These three factors together constitute the basic fields of the heat treatment file for that batch, which are used for subsequent identification of characteristic areas on the plate. If the measured temperature of a batch is lower than the set lower limit due to temporary cooling of the heating furnace, the normalized microstructure record is simultaneously associated with an under-temperature indicator. This indicator prompts priority verification of the grain size of the corresponding area when subsequently identifying characteristic areas on the plate, ensuring that localized coarse grains caused by under-temperature are given priority during the next step of investigating the normalized microstructure.

[0030] In some embodiments, determining the phase transformation stress distribution by analyzing the residual stress state through the normalized structure includes: identifying characteristic regions of the plate surface based on the normalized structure to determine the layout of stress measurement points; measuring the residual stress amplitude of each measurement point according to the stress measurement point layout to obtain a stress amplitude sequence; calculating the stress plateau distribution span between measurement points based on the stress amplitude sequence to determine the stress dispersion; and extracting a large-scale moderate stress zone to form a phase transformation stress distribution using the stress dispersion.

[0031] The layout of stress measurement points is determined based on the identification of characteristic regions on the plate surface using normalized microstructure. The grain size distribution along the plate surface during normalization is the direct basis for delineating characteristic regions: the ends of the rolling direction dissipate heat quickly due to the edge rolling effect, resulting in finer grains, while the middle dissipates heat slowly, resulting in coarser grains. Based on this, the plate surface exhibiting normalized microstructure is divided into three types of characteristic regions: edge, transition, and center. The transition zone is a band-shaped area where the grain size gradually changes between the two. Abrupt grain size changes often correspond to the location with the greatest difference in local shrinkage and the steepest residual stress peaks and valleys during normalizing cooling. Therefore, when the grain size of a certain region differs from the surrounding area by more than two levels, it is marked as a characteristic region in advance, and the measurement points are denser. The stress measurement point layout is based on the characteristic regions: the edge and center each account for 30% of the total measurement points, and the transition zone accounts for 40% to cover the location with the fastest gradient change. The grain size abrupt change area is further densely populated with twice the conventional density, and these additional points do not occupy the conventional quota, thus obtaining finer stress sampling resolution in microstructure-sensitive areas. Each measuring point records its coordinates, the category of its characteristic region, and whether it is densified. The coordinates are based on the geometric center of the steel plate as the origin, and are marked according to the rolling direction and transverse direction, sharing the same coordinate system as the thickness deviation grid. The determination of whether a region represents a grain size abrupt change is based on the statistical analysis of grain size distribution formed from the normalizing microstructure inspection records of the same batch, rather than a fixed, artificially set grade. Steel plates from the same batch with consistent characteristic region divisions share the same set of stress measuring point layout coordinate templates. Only densified points caused by grain size abrupt changes are recorded separately for each sheet. This ensures that the stress measuring point layout covers the microstructure-sensitive areas presented by normalizing without solidifying occasional anomalies into the template, and without losing batch comparability due to differences between sheets. The direction of the densified points is along the steepest grain size gradient, so that the densified points, when connected in series, precisely span both sides of the abrupt change zone, capturing the transition from fine to coarse grains in the normalized microstructure rather than remaining in a single grain size region.

[0032] The residual stress amplitude at each stress measuring point was determined based on the stress measuring point layout to obtain a stress amplitude sequence. The residual stress amplitude at each measuring point reflects the peak-to-valley difference of internal stress left by uneven local shrinkage during normalizing and cooling at that point. Only the magnitude of the amplitude was taken, without distinguishing between tensile and compressive directions. The blind hole method was used for measurement: a blind hole with a diameter of 2 mm and a depth of 2 mm was drilled at each coordinate marked on the stress measuring point layout. Strain rosettes were attached around the holes. After drilling to release the original constraint at that point, the strain recovery measured by the strain rosettes was proportional to the residual stress amplitude at that point. The proportionality coefficient is the release coefficient. The value was determined according to the specimens of the same heat number for the same steel grade to match the elastic modulus and Poisson's ratio corresponding to its alloy composition. Different heat numbers were calibrated separately due to differences in composition. For measuring points marked as densified points in the stress measuring point layout, the drilling sequence was arranged after the regular points. The waiting time between adjacent densified points was twice that of the regular points to ensure that the stress release of the previous hole was completely stable before drilling the next hole, preventing the release fields of adjacent holes from superimposing and artificially inflating the readings. When there are more than three calibration records for the same furnace number, the release coefficient is taken as the average of all calibrations to suppress random errors from a single calibration; otherwise, the most recent individual calibration value is used without interpolation with other furnace numbers. The stress amplitude sequence is filled point-by-point with the converted residual stress amplitude, using the coordinates of the stress measurement point layout as the framework. The position corresponds one-to-one with the coordinates without interpolation completion. Each measurement point in the stress amplitude sequence is filled with the conversion result only once. The calibration count and source furnace number of the release coefficient for each measurement point are recorded simultaneously as a basis for tracing the reliability of that amplitude in the stress amplitude sequence. The more calibration counts, the higher the confidence level of the value at that point. Measurement points with low confidence levels are not given priority when critically merging stress platforms in subsequent divisions.

[0033] The stress dispersion is determined by calculating the span of stress plateau distribution between measuring points in the stress amplitude sequence. If the stress amplitudes of adjacent measuring points in the stress amplitude sequence fall within the same narrow interval and do not fluctuate significantly with the coordinates, a stress plateau is formed. The difference between multiple plates is the key to judging whether the residual stress on the plate surface is uniform, rather than the absolute reading of any single point. The stress plateau is scanned segment by segment in the stress amplitude sequence using a sliding window: five consecutive adjacent measuring points are taken as a group, the range within the group is calculated, and a linear fit is performed on the group along the coordinate direction to obtain the slope. When the range does not exceed 10% of the group mean and the fitted slope is close to zero, it is judged as the same plateau, merged into a segment, and the plateau mean and coordinate span are recorded. If the range meets the standard but the slope is too large, it indicates that the stress rises monotonically along the coordinates and is not a true plateau. The window is narrowed down to three measuring points for re-judgment. Measuring points that still do not meet the requirements are included in the transition zone as a stress gradient zone between two plates and are marked separately. When the difference between the means of adjacent windows falls within half of the measurement error, they are further merged. Measuring point readings with low reliability are not given priority in the merging critical treatment to prevent the wide plateau from being fragmented by the window boundary. After sorting all platforms by their mean values, the highest mean value is denoted as σ_max and the lowest mean value as σ_min. Stress dispersion is calculated using the formula Δσ = σ_max - σ_min, in megapascals (MPa), directly reflecting the range of residual stress across the entire steel plate from the most relaxed to the most stressed region. Stress dispersion is only meaningful when two or more platforms are delineated in the stress amplitude sequence; when all measuring points are combined into a single platform, stress dispersion is recorded as zero. If, after two window reductions, more than 20% of the measuring points on a plate are still included in the transition zone, stress dispersion is still calculated, but the corresponding transition zone percentage is higher, indicating that the risk of this plate depends more on the location of the gradient zone than on a single drop value. A wider transition zone and a greater number of platforms indicate that the stress is distributed in a continuous gradient along the plate surface rather than concentrated in a few areas. This distribution pattern, along with the stress dispersion value, forms the basis for subsequently extracting intermediate stress zones.

[0034] Stress dispersion is used to extract a large-scale medium stress zone to form the phase transformation stress distribution. When the stress dispersion is lower than the lower limit of normal fluctuation for the same thickness specification of the steel grade, the residual stress of the entire plate is judged to be uniform, and the phase transformation stress distribution is recorded as an empty set and archived along with the measured stress dispersion value, indicating that the plate does not require zoned temperature differential treatment. In most cases, after sorting all stress platforms by mean, the highest and lowest platforms are removed. The highest platform often corresponds to overheating at the plate edge or occasional local concentration, and the lowest platform often corresponds to measurement background noise. Neither of them represents the main level. If they are not removed, the judgment of the middle section will be skewed by the extreme values ​​at both ends. Among the remaining platforms, the adjacent platforms with the largest area and mean falling within the range of 40% to 60% of the σ_min stress dispersion are merged into a large-scale medium stress zone. The stress corresponding to this middle section is neither the most relaxed nor the most acute, but has the widest distribution range. It is the area most likely to be superimposed on a large area during phase transformation, thus inducing batch cracking. This value range is derived from the empirical range determined by comparing the phase transformation stress distribution of similar steel plates with the actual crack location in flaw detection. The phase transformation stress distribution consists of two parts: the boundary coordinate set of the intermediate stress zone and the stress mean interval of the plate within the zone. The boundary is taken as the midpoint of the coordinate system where the adjacent measuring point plate belongs to which the system switches. If there are two candidate regions with similar areas but a mean difference of more than 10% within the zone, the phase transformation stress distribution retains these two regions as independent sub-zones to reflect the possible dual concentration trend of the plate. These sub-zones are used for separate positioning during thickness coupling to avoid merging two concentration areas with different origins into one and thus masking one of them. Large-scale intermediate stress zones are usually distributed continuously or quasi-continuously on the plate surface coordinates. Their boundary coordinate set and the stress mean interval within the zone together constitute the core object of the phase transformation stress distribution used for the next step of thickness coupling.

[0035] Step S102: Utilize the phase change stress distribution to locate the stress concentration area and determine the differential temperature heating regime. Analyze the heat dissipation lag trend of the differential temperature heating regime to obtain the differential temperature heat compensation increment. Use the differential temperature heat compensation increment to carry out pre-compensation heating control and establish a steady-state heat preservation regime.

[0036] In some embodiments, the step of using the phase change stress distribution to locate stress concentration areas and determine the differential temperature heating regime includes: establishing a thickness distribution map by performing stress-thickness coupling mapping based on the phase change stress distribution; determining candidate stress concentration areas by analyzing the thickness difference phase change mismatch parts through the thickness distribution map; measuring the mismatch stress superposition level of the candidate stress concentration areas to obtain stress concentration areas; and extracting the partition boundary coordinates based on the stress concentration areas to construct the differential temperature heating regime.

[0037] A thickness distribution map is established based on the stress-thickness coupling mapping of phase transformation stress distribution. Areas with larger plate thicknesses cool more slowly and undergo phase transformation later. If these areas coincide with areas of residual stress concentration, the risk of cracking is highest due to hindered cooling contraction and superposition with existing internal stresses. Therefore, the phase transformation stress distribution and thickness information must be correlated geographically for comprehensive interpretation. The coordinates of the intermediate stress zone boundaries given by the phase transformation stress distribution share the same coordinate system as the thickness deviation grid recorded during the raw material testing stage. This system uses the geometric center of the steel plate as the origin and is divided into five equal parts vertically and horizontally. The two coordinates are directly superimposed without coordinate transformation or interpolation. The total number of coordinates in the thickness distribution map remains consistent with the thickness deviation grid and does not increase or decrease due to stress zone calculations. Coordinates falling within the area enclosed by the phase transformation stress distribution boundary are associated with the corresponding intermediate stress zone number in the thickness distribution map; those falling outside the zone are associated with the background zone number. Coordinates falling exactly on the boundary line between two adjacent zones are associated with the zone number of the closer zone, and if the distances are equal, the smaller number is used. Within the thickness distribution map, coordinates under the same zone number are arranged in ascending order of thickness deviation. When the difference between the maximum thickness deviation within a zone and the adjacent baseline zone exceeds one rolling tolerance grade for that steel grade, it is marked as the location of a sharp thickness change in that zone and prioritized as the key coordinate for the next mismatch identification, preventing extreme thickness values ​​within the zone from being masked by the average trend along the zone and thus missed. When the phase transformation stress distribution is an empty set, all coordinates in the thickness distribution map are assigned to the baseline zone number, and the thickness deviation field is still retained according to the original detection results. Thickness data is not cleared due to missing stress zone assignments, ensuring that even steel plates without stress concentrations retain complete thickness information baselines.

[0038] By analyzing thickness distribution maps, candidate stress concentration areas are identified at locations where phase transformation mismatch occurs due to thickness differences. Areas with different thicknesses should have different cooling and phase transformation rhythms, typically accompanied by stress band switching. When the difference in thickness deviation between adjacent coordinates on the thickness distribution map exceeds the upper limit of the normal rolling tolerance for that steel grade, it is identified as a thickness difference location. If this location still falls within the same stress band number on the thickness distribution map and no switching occurs, it indicates that its phase transformation behavior has not adjusted synchronously with the thickness change, constituting a phase transformation mismatch location, distinguishing it from a simple thickness abrupt change in stress band switching. The stress concentration candidate area only includes coordinates where both thickness difference and phase transformation mismatch are simultaneously established; coordinates with only thickness differences but normal stress band switching are not included, thus separating pure rolling tolerance fluctuations from genuine phase transformation mismatches. Stress concentration candidate areas, concentrated in the base area at the edge of the steel plate and belonging to historically fluctuating thickness difference locations, are not included in the stress concentration candidate area calculation. Instead, they are associated with the edge fluctuation exemption flag and counted separately by steel plate serial number. This allows for differentiation between misjudgments and actual omissions during process review, ensuring that inherent measurement fluctuations at the edge do not drown out genuine mismatch signals within the plate. This exemption flag does not affect the normal interpretation process of other coordinates in the thickness distribution map. Stress concentration candidate areas are recorded in coordinate order, including the thickness difference amplitude and the original stress zone number. When the same mismatch location spans multiple consecutive coordinates, the stress concentration candidate areas are merged into one candidate interval, with the start and end positions of this batch of consecutive coordinates as the interval's start and end points. Within the interval, the thickness difference amplitude is no longer retained point-by-point, thus reducing redundancy while retaining the crucial information of the interval span reflecting the size of the mismatch range.

[0039] The stress concentration zones are obtained by measuring the level of mismatched stress superposition in candidate stress concentration zones. Not all candidate stress concentration zones are retained; the level of mismatched stress superposition L in each zone needs to be measured individually. The average of the absolute deviations of the stress amplitudes at all measuring points within the candidate stress concentration zone from the mean value of the original stress band within that zone is taken, and calculated using the formula L=(1 / n)Σ|σ_i-σ_band|, where σ_i is the stress amplitude at the i-th measuring point within the zone, taken from the stress amplitude sequence of the steel plate, σ_band is the mean value of the original stress band within the zone, n is the total number of measuring points within the zone, and L is in megapascals (MPa). A larger value indicates a more significant mismatch between the stress at that point and the overall level of the original stress band. When the total number of measuring points n in the candidate stress concentration zone is less than 5, L is instead taken as the average of the smaller of the absolute deviations of each measuring point within the zone from the mean values ​​of two adjacent stress bands, still in megapascals. Stress concentration candidate areas where the stress concentration level (L) exceeds the mismatch threshold obtained from the retrospective statistics of crack locations in the flaw detection of this type of steel plate are retained as stress concentration areas. Areas where L does not exceed the threshold are downgraded from stress concentration candidate areas to ordinary thickness difference locations and do not enter the differential heating process, because mismatches below the threshold have not historically corresponded to cracks. When all candidate intervals have L values ​​below the threshold, the stress concentration area is considered an empty set, and the steel plate is transferred to the conventional uniform heating process instead of the differential heating process. For each stress concentration area, the boundary coordinates, the measured L value, and the original stress zone number are retained, and these are numbered in descending order of L value. When constructing the differential heating regime, the boundary of the stress concentration area with the highest degree of mismatch is prioritized, ensuring that heating compensation resources are first allocated to the most likely cracking location. After the stress concentration areas are numbered in descending order of L value, the higher-numbered areas are given priority in allocating the temperature boost level of the heating unit when constructing the differential heating regime in the next step.

[0040] Differential heating regimes are constructed by extracting the boundary coordinates of stress concentration zones. For example... Figure 2As shown, the boundary coordinates of each interval in the stress concentration zone are converted into the physical boundary range of that interval under the actual dimensions of the steel plate body 1. The physical boundary corresponds one-to-one with the side heating unit 2, central heating unit 4, and other partitioned heating units of this type of heating furnace. The number of heating units is the fixed number of partitions configured at the factory for this furnace. When the physical boundary of a stress concentration zone crosses the boundary line of two adjacent heating units, it is incorporated into the heating unit with the larger coverage area according to the principle of proximity expansion, without being split into two segments spanning units, so that a stress concentration zone is always compensated by a single heating unit. The differential temperature heating regime uses the heating unit number as an index to associate each unit with the existence of a stress concentration zone within its coverage area and the corresponding heating temperature increase: the temperature increase is taken from the stress concentration zone covered by the unit with the highest L, and converted according to the classification table of L and temperature increase. The higher the L, the greater the increase. The classification table is listed in the numerical range of L and is revised every six months based on the latest verification test data. Units that do not cover any stress concentration zones are associated with a zero increase indicator in the differential temperature heating regime and are consistent with the conventional normalizing heating temperature. When the temperature rise difference between two adjacent units exceeds the factory acceptance limit for the gradient of adjacent units in the furnace, the differential heating system inserts a transition heating unit 3 between them and takes the arithmetic mean of the values ​​of the two units. This ensures a smooth temperature transition along the plate surface and prevents the heating itself from introducing new temperature gradients and stress concentrations at the unit boundaries. The heating parameters for each differential heating system can be traced back to the corresponding version used at the time, providing a reference for distinguishing between different versions of the differential heating system when analyzing heat dissipation lag trends.

[0041] In some embodiments, the step of analyzing the heat dissipation lag trend to obtain the differential temperature heat compensation increment for the differential temperature heating system includes: identifying the heat dissipation lag trend of steel plates of the same specification in previous iterations based on the differential temperature heating system to form a lag trend parameter; analyzing the heat dissipation lag extrapolation of the plate edge based on the lag trend parameter to determine the predicted heat dissipation amount; calculating the predicted heat dissipation amount to determine the heat compensation correction coefficient for previous underheating prediction deviations; and calculating the total comprehensive heat compensation amount based on the heat compensation correction coefficient to determine the differential temperature heat compensation increment.

[0042] Based on the differential temperature heating regime, the lag trend parameter is formed by identifying the heat dissipation lag trend of steel plates of the same specification in previous heating cycles. Two attributes of each heating unit in the differential temperature heating regime—temperature rise and whether it covers stress concentration areas—determine which type of heat dissipation lag history record to retrieve for that unit: units with a non-zero temperature rise retrieve the differential temperature heating history record, while units with a zero temperature rise retrieve the conventional normalizing history record. These two types are statistically analyzed separately to ensure that the additional heat from the differential temperature unit does not mask the lag baseline level of the conventional unit. The lag trend parameter is the weighted average of the time taken for the same type of heating unit in the same specification of steel plate from the end of heating to the measured plate temperature falling back to the set steady-state temperature. The weights are allocated according to the completeness of the data collection for each batch of steel plates; batches with less than 80% completeness are only listed separately as reference samples. Under the same differential temperature heating regime, even if the temperature rise of the edge units and the internal units is the same, they are statistically analyzed separately in the lag trend parameter. The edge units, with three sides exposed and a high heat dissipation area, have a different temperature stabilization time pattern than the internal units that are heated on all four sides. The lag trend parameter is indexed along two dimensions: heating unit location category and temperature rise interval. Batches of the same steel grade with consistent heating unit division methods share the same lag trend parameter index. When the division method changes, the index is recalculated, and the old index is not extrapolated to the new division. This ensures that each lag pattern corresponds to a specific unit geometric layout, avoiding the use of outdated lag experience data after layout adjustments. The finer the temperature rise level division in the differential heating system, the greater the number of intervals in the lag trend parameter index, ensuring that each level can be matched with the corresponding historical lag time.

[0043] The predicted heat dissipation is determined by analyzing the lag trend parameters to extrapolate the heat dissipation lag at the plate edge. Since the plate edge has three open sides and a high proportion of heat dissipation area, its temperature drops faster and is more difficult to stabilize after heating. Its lag time is typically longer than that of internal units with balanced heating on all four sides. The difference between these two in the lag trend parameters is the baseline value for the plate edge heat dissipation lag extrapolation. This baseline value is linearly amplified according to the actual number of heating units at the edge of the steel plate to obtain the additional heat preservation time lost by the entire steel plate due to faster heat dissipation at the edges. During amplification, the cumulative length of the edge units along the plate length is used as the weight; the denser the edge units, the narrower the plate width, and the higher the proportion of open space per unit area, the greater the extrapolation. The predicted heat dissipation is obtained by multiplying the extrapolated duration by the heat dissipation per unit time. The heat dissipation per unit time is not directly taken as the slope of the air cooling curve, but rather the slope (degrees per second) is multiplied by the specific heat capacity of the steel grade (kilojoules per kilogram of degrees Celsius) and the mass of the steel plate, and converted into the heat power lost per unit time (kilojoules per second). This is then multiplied by the extrapolated duration to obtain the predicted heat dissipation in kilojoules, which represents the extra heat lost through the edge of the plate within the extrapolated duration if no compensation is provided. When the extrapolation is zero, the predicted heat dissipation is also recorded as zero, without fabricating a non-zero minimum value due to the formula form; when the difference between the lag time of the edge and the internal unit in the lag trend parameter is lower than the time resolution of the temperature measurement system of this model, the extrapolation is treated as a measurement error and is not included in the extrapolation calculation. At this time, the predicted heat dissipation is replaced by the lower limit of the heat dissipation of the board edge of the same specification data of this model, and is not obtained by direct extrapolation without relying on the lag trend parameter of the current batch. It is also archived together with the source batch number of the heat dissipation rate used for conversion as the traceability basis when calculating the underheating prediction deviation.

[0044] The heat compensation correction coefficient is determined by calculating the predicted underheating deviation from previous tests based on the estimated heat dissipation. For each batch of steel plates of the same specification, the difference between the actual measured underheating and the predicted value based on the estimated heat dissipation constitutes the underheating prediction deviation: a positive deviation indicates that the actual underheating is more severe than predicted, and the prediction is conservative; a negative deviation indicates that the prediction is aggressive. The deviation value and the predicted heat dissipation are both in kilojoules, and the two can be directly subtracted without additional unit conversion. The heat compensation correction coefficient is the average of the ratio of each underheating prediction deviation to the corresponding predicted heat dissipation. The ratio itself is dimensionless and reflects whether the overall predicted heat dissipation is systematically too high or too low. When calculating the average, individual records with an absolute deviation value exceeding three times the historical standard deviation for that model are excluded to avoid individual batches with abnormal temperature measurements affecting the heat compensation correction coefficient. These excluded records are archived separately, not included in the average, but retained for process review. Once the review confirms an anomaly, they are not recalculated. When there are fewer than 5 records, the heat compensation correction factor is temporarily taken as the general correction factor for the steel grade category, and no specific factor is calculated for this type of steel plate. Once more than 5 records are accumulated, the calculation will be switched to a specific factor. The source of the factor before and after the switch is marked synchronously, and the two types of values ​​are not mixed. After the specific factor calculation is enabled, the general value is only retained for historical reference. A heat compensation correction factor greater than 1 indicates that the predicted heat dissipation of this type of steel plate has underestimated the actual heat loss, and additional heat compensation is required; a factor less than 1 indicates the opposite. The factor itself is saved separately according to the model and the heating unit location category. Different models or different location categories do not borrow from each other to avoid the correction experience of peripheral units being mistakenly applied to internal units, and to avoid the mixing of heat compensation correction factors between different thickness specifications.

[0045] The differential temperature heat compensation increment is determined by calculating the total comprehensive heat compensation based on the heat compensation correction coefficient. The base heat compensation for each heating unit is calculated by multiplying the predicted heat dissipation of that unit by the heat compensation correction coefficient. The unit of the base heat compensation is kilojoules, consistent with the heat dissipation. The base heat compensation of all heating units is added sequentially according to their unit numbers to obtain the total comprehensive heat compensation for the steel plate. No weighting distinction is made between units during the addition process; the weighting differences are already reflected in the predicted heat dissipation values ​​of each unit. The differential temperature heat compensation increment is calculated by converting the total comprehensive heat compensation into the percentage power increase that the furnace model can execute. The conversion relationship is taken from the correspondence table of kilojoule heat compensation and power increase percentage in previous power calibration tests of this furnace model. The correspondence table lists the increase percentage by heat compensation interval, without continuous interpolation. When the power increase percentage obtained from the conversion of the total comprehensive heat compensation exceeds 20%, the differential temperature heat compensation increment is capped at this 20% upper limit. The excess portion is compensated by extending the holding time. The extended holding time is calculated by dividing the excess heat compensation by the rated power of the furnace. The heat compensation correction coefficient is used in the above calculation according to the location category of the heating unit. The heat compensation correction coefficients of different location categories are not interchangeable. If a unit uses the general heat compensation correction coefficient of the steel grade category due to insufficient records, the corresponding differential temperature heat compensation increment is synchronously associated with the general coefficient identifier, indicating that the actual matching accuracy of the heat compensation increment is lower than that of the batch using the specific coefficient. The differential temperature heat compensation increment under the same number is calculated only once in the same heat and is not repeatedly accumulated due to multiple temperature measurement records during the heating stage. When the difference in differential temperature heat compensation increment between adjacent units exceeds the upper limit of the adjacent gradient of the heat, the two are averaged and then redistributed to avoid the heat compensation itself introducing a new temperature difference at the unit boundary, so that each unit only receives one smooth transition compensation parameter in this heat.

[0046] A steady-state heat preservation system is established through pre-compensation heating control using differential temperature heat compensation increments. The percentage increase in power or the extension of heat preservation time for each heating unit in the differential temperature heat compensation increment is pre-issued to the controller 10 of the corresponding heating unit before the start of the heat preservation phase. The controller adjusts its output power or heat preservation termination time accordingly, without passively responding after a deviation in the measured temperature occurs; this is the pre-compensation heating control. During execution, each heating unit reports its measured plate temperature at fixed intervals. When the difference between the measured plate temperature and the unit's expected steady-state temperature exceeds the upper limit of the furnace's normal fluctuation, the corresponding unit receives a small power correction, calculated as half of the temperature deviation. The steady-state heat preservation system is divided into three stages in chronological order: pre-compensation issuance, execution monitoring, and final verification. The issuance stage records the actual differential temperature compensation increment parameters received by each unit and the issuance time. The monitoring stage records each reported temperature and correction action. The final verification stage verifies whether the measured heat preservation time of all heating units reaches the minimum heat preservation time for that steel grade and records the measured temperature at the end of heat preservation for the cooling stage. If the actual heat preservation time of a unit exceeds the upper limit by more than 10% due to additional corrections, the steady-state heat preservation system is linked to an overdue correction flag, and the differential temperature compensation increment issuance parameters for subsequent steel plates in the same batch of that unit are simultaneously reduced by one level. The steady-state heat preservation system can only be considered complete for that furnace batch after all units have passed the final verification. If any unit fails verification, the heat preservation stage is extended, and the cooling step is temporarily suspended. This ensures that each steel plate entering cooling has reached a uniform and sufficient austenitization state in the thickness direction, providing a consistent initial temperature field for the next stage of segmented cooling and preventing local under-temperature from being amplified into new microstructure differences during the cooling stage.

[0047] Step S103: According to the steady-state heat preservation system, the surface cooling rate is obtained by segmented cooling control. The core cooling rate gradient is calculated based on the surface cooling rate to obtain the core-surface cooling rate difference. The core-surface cooling rate difference is used to screen out the areas where phase transformation stress is superimposed and concentrated to output the crack sensitive area.

[0048] Specifically, the surface cooling rate is obtained by segmented cooling control based on the steady-state heat preservation regime. Segmented cooling begins at the moment the steady-state heat preservation regime determines the completion of the heat run: the first stage uses a water mist strong cooling spray device 8 to rapidly lower the plate surface to the intermediate transition temperature set for the steel grade, quickly carrying the surface layer through the high-temperature zone to refine the microstructure and suppress the precipitation of coarse ferrite; the second stage involves slow cooling in the air slow cooling zone 9, allowing the phase transformation to complete under a gentler temperature gradient, avoiding surface embrittlement caused by rapid cooling throughout the phase transformation process. The switching time between the two stages is taken from the measured heat preservation termination temperature of each heating unit recorded in the steady-state heat preservation regime. Areas with higher termination temperatures reach the transition temperature first and enter slow cooling first. Synchronous switching of the entire plate is not required, allowing each area to enter its corresponding cooling rhythm according to its actual temperature. The surface cooling rate is calculated by dividing the temperature drop between two adjacent plate surface temperature measurements by the time interval between the two measurements. The temperature measurement points correspond one-to-one with the heating units defined in the differential temperature heating stage, and each unit is calculated and archived separately. The surface cooling rates of different units are not uniform to preserve the regional differences in plate surface cooling. During the rapid cooling phase, samples are taken every 2 seconds; during the slow cooling phase, samples are taken every 10 seconds. The difference in sampling intervals does not change the calculation method of the surface cooling rate, but only affects the time span corresponding to the temperature drop. If an abnormal jump occurs within one sampling cycle before and after the switch due to reasons such as water film detachment, and the jump amplitude exceeds twice the historical average rate of the unit during the rapid cooling phase, the value of that cycle is marked as a switching disturbance and is not included in the smoothing statistics of the cooling curve. If the surface cooling rate of a certain unit is consistently lower than that of other units in the same furnace during the rapid cooling phase, it is determined that the corresponding nozzle of that unit is partially blocked and associated with a spray anomaly. The independent rate curves for each unit allow such individual cooling anomalies to be identified and traced back for investigation, rather than being masked by the average value of the entire board.

[0049] The core cooling rate gradient is calculated by measuring the surface cooling rate to obtain the core-surface cooling rate difference. The core cooling rate is determined by a thermocouple pre-embedded at the center of the steel plate thickness, with the embedding depth at half the plate thickness. The sampling time of the thermocouple shares the same clock reference as the surface temperature measurement time corresponding to the surface cooling rate, avoiding misalignment of data due to clock drift. The surface cooling rate is calculated using the formula v_s=ΔT_s / Δt, where ΔT_s is the temperature drop between two adjacent plate surface temperature measurements, and Δt is the time interval between the two measurements. The core cooling rate v_c is calculated in the same form, only the measurement location is changed to the center of the plate thickness. Both are in degrees Celsius per second, and the calculation period is consistent. The core-to-surface cooling rate difference is calculated using the formula Δv = v_s - v_c, in degrees Celsius per second. The surface dissipates heat faster, reaching its phase transformation temperature first and undergoing a structural transformation, resulting in volume expansion. At this point, the core remains in a high-temperature austenitic state and has not yet transformed. The surface layer, which transforms first, is constrained by the core, which transforms later, creating opposing tensile and compressive internal stresses along the thickness direction. Δv is positive, and the larger the value, the greater the difference in the timing of phase transformation between the surface and core, and the stronger this constraining stress. This is the direct source of phase transformation sequence misalignment and cracking tendency along the thickness direction. When the thickness is lower than the minimum zone thickness set by the differential heating system for this steel grade, it is difficult to stably drill and embed thermocouples in the center due to the thinness of the plate. The core thermocouple is then embedded to one-quarter of the plate thickness, and the core temperature is inferred from the measured temperature difference between one-quarter and the surface. The calculation method for the core-to-surface cooling rate difference remains unchanged; only the depth markings of the measuring points are updated. The difference in cooling rate between the core and surface is usually significantly greater in the strong cooling stage than in the slow cooling stage. If the Δv in the strong cooling stage is lower than the statistical lower limit of the strong cooling stage of this model in the same batch, it is determined that the spraying of the strong cooling equipment is uneven and the equipment associated with the cooling record of this batch is abnormal.

[0050] In some embodiments, the step of screening the location of concentrated phase transformation stress from the surface cooling rate difference to output a crack-sensitive area includes: collecting layered temperature records in the thickness direction based on the surface cooling rate difference to obtain a layered temperature group; determining the phase transformation initiation time difference of each layer based on the layered temperature group to obtain a phase transformation timing deviation; analyzing the overlapping intervals of phase transformation timing through the phase transformation timing deviation to determine the location of concentrated phase transformation stress; and evaluating the crack initiation probability of the location of concentrated phase transformation stress to output a crack-sensitive area.

[0051] Layered temperature groups were obtained by collecting stratified temperature records along the plate thickness direction based on the surface cooling rate difference. Periods with larger surface cooling rate differences correspond to the stages with the steepest temperature gradients and the most likely phase transitions to be out of sync along the plate thickness. The stratified temperature records from these stages are the most valuable for reference. During data collection, all stratified data from one sampling cycle before and after this period were prioritized. For other periods, one-third of the normal sampling density was used to concentrate the data in the segments that best reflect temporal misalignment without diluting it by smoother segments. The stratified temperatures along the plate thickness were obtained from three depths: surface thermocouple 5, one-quarter plate thickness thermocouple 6, and one-half plate thickness (i.e., core thermocouple 7), all buried in the same batch as the core cooling rate measurement stage. The one-quarter measurement point was arranged together with the core thermocouple during installation. The sampling times at all three depths were aligned to the same clock reference used for the surface cooling rate difference. The stratified temperature groups were indexed by depth level, collecting temperature records from all sampling times of each layer and arranging them chronologically. No interpolation was performed between different depth levels to preserve the true cooling sequence of each layer. When a certain stratum lacks continuous sampling points during the period marked as equipment malfunction by the cardiac surface cooling rate difference, the corresponding stratum in the stratified temperature group should leave a gap for that period, with the gap range consistent with the time marked as equipment malfunction. Values ​​from adjacent strata or moments before and after the current stratum should not be used to fill the gap, to avoid fabricating a non-existent phase transition inflection point. When the number of records in each stratum of the stratified temperature group is inconsistent, the stratum with the fewest records should be used to align the common time range. The alignment process does not change the original timestamps. The aligned common time range is used by each stratum to calculate the phase transition initiation time difference, ensuring that the three depth strata are comparable point-by-point on the same time axis and that comparisons are not misaligned due to missing measurements in individual strata.

[0052] Phase transformation timing deviation is obtained by measuring the time difference of phase transformation initiation in each layer based on the layered temperature group. The phase transformation initiation time of each depth layer is taken as the inflection point in the cooling curve of that layer in the layered temperature group where the temperature briefly rises or the cooling rate suddenly decreases due to the release of latent heat of phase transformation. The criterion is that the cooling rate of three adjacent sampling points decreases by more than a set proportion compared with the previous average cooling rate for that steel grade. The sampling time corresponding to the point where the temperature drops to the target is the phase transformation initiation time of that layer. This criterion captures the physical characteristic that the heat release of phase transformation slows down the cooling. When two consecutive inflection points appear in the same layer, the earlier one is taken. The phase transformation timing deviation is based on the phase transformation initiation time of the surface layer. The time difference between the phase transformation initiation time of the other depth layers and the surface layer is recorded layer by layer. When the surface layer is used as the reference, the time difference is usually positive because the core is lagging behind in thermal conduction. The unit is seconds, consistent with the sampling interval. The larger the time difference, the more the phase transformation of that layer lags behind the surface layer and the more asynchronous the transformation in the thickness direction. When a certain layer has a gap in the corresponding interval of the layered temperature group and the inflection point cannot be determined, its phase transformation start time is marked as pending. The layer corresponding to the phase transformation timing deviation only retains the pending mark and is not filled in by interpolation of the time difference of other layers. When the time difference of each layer of the phase transformation timing deviation is less than the time resolution of the temperature measurement system of this model, it is determined that the phase transformation process in the thickness direction of the plate is approximately synchronous and there is no obvious timing misalignment. Such plates usually have low internal stress. If the surface layer in the layered temperature group has an inflection point feature that is not obvious due to the rapid cooling during the strong cooling stage, a quarter layer of the plate thickness is used as the reference layer for the phase transformation timing deviation and noted in the record. This ensures that the time difference between different steel plates always uses an identifiable inflection point as a common reference and is not directly compared due to different reference layers. Changes in the reference layer also affect the starting reference of the duration window of subsequent layers.

[0053] For example, the step of determining the location of concentrated phase transformation stress by analyzing the overlapping interval of phase transformation time through the phase transformation time deviation includes: obtaining a layered time window by aggregating the duration windows of each layered phase transformation in the plate thickness direction according to the phase transformation time deviation; obtaining a layered overlapping interval by measuring the deviation of the start and end times of adjacent layered phase transformations through the layered time window; obtaining a phase transformation time overlapping interval by performing an overlap ratio analysis on the layered overlapping interval; and determining the location of concentrated phase transformation stress by comparing the phase transformation time overlapping interval with a synchronous judgment threshold.

[0054] Based on the phase transformation timing deviation, the duration window of each layer's phase transformation along the plate thickness direction is collected to obtain the layer time window. The phase transformation timing deviation already provides the start time of each layer's phase transformation. This step further determines the end time of each layer's phase transformation: the end time is the sampling time when the layer's cooling rate recovers to the normal level before the phase transformation. The recovery criterion shares the same proportional threshold as the start inflection point, only in the opposite direction. The judgment window searches backward from the start time, without backtracking to records before the start time, to avoid misjudging normal cooling before the phase transformation as the end of the phase transformation. The layer time window records the start and end times for each layer, with the window width being the difference between the two. A larger width indicates a longer phase transformation duration for that layer, usually corresponding to local alloy segregation or subtle differences in cooling conditions that prolong the phase transformation. A narrow width may indicate that the layer is cooling too quickly, compressing the phase transformation into a concentrated period. In the phase transition timing deviation, the start time of a layer marked as "pending" is itself undetermined. The corresponding layer in the stratified time window is also marked as "pending" and its end time is not assigned separately. The window for that layer is determined after the start inflection point is completed. If the end time of a phase transition in a layer is later than the last time of all samples recorded for that layer, it indicates that the phase transition has not yet ended within the cooling curve recording range. The end time of that layer in the stratified time window is marked as "unclosed," and its width is calculated based on the difference between the last time and the start time within the recording range, representing the lower limit rather than the true width. It is not used for comparison as the actual phase transition duration of that layer. The start and end times of adjacent layer windows together constitute the basic data for determining the overlapping interval. If the width of a layer window is less than half that of the adjacent layer, it indicates that the phase transition in that layer was cooled and compressed too quickly, and local hardening structures are prone to form. This layer is additionally marked as a suspected fast-cooling layer in the overlap judgment.

[0055] Layer overlap intervals are obtained by measuring the deviation of the start and end times of phase transitions between adjacent layers using layer time windows. If there is a temporal intersection between the layer time windows of two adjacent depth levels, the start time of the intersection is taken as the later of the start times of the two windows, and the end time is taken as the earlier of the end times of the two windows. An intersection is determined to exist only when both are valid. Windows of any level that are pending or not closed are not included in the intersection calculation until the data of that layer is completed. Blanks before completion are not replaced by windows of other levels. Layer overlap intervals are calculated pairwise for adjacent layers. If there are three depth levels in the thickness direction, two sets of overlap intervals are calculated: one between the surface layer and the quarter layer, and the other between the quarter layer and the core layer. The overlap between the surface layer and the core layer is calculated directly without crossing layers to avoid omitting the thermal hysteresis information carried by the intermediate layers, since the relationship between the surface layer and the core is actually conducted step by step through the intermediate layers. When the time windows of adjacent layer pairs do not overlap at all, the overlapping interval is recorded as a zero-width interval instead of being skipped. Zero width itself means that the phase transformation time sequence in the thickness direction is completely staggered and there is no superposition. When the width of the overlapping interval is close to the full width of the narrower of the two windows involved in the calculation, it means that the phase transformation process of the narrower layer is almost completely contained by the wider layer. In this case, an additional inclusion relationship identifier is associated when archiving to distinguish it from ordinary partial overlap. The intersection determination of overlapping intervals is directly taken from the original value at the sampling time without rounding. When the starting time of the overlapping interval of an adjacent layer pair is later than the midpoint of each of the two windows, it means that the two only briefly overlap at the end of the phase transformation, and the stress superposition is mainly concentrated in the final stage of the phase transformation. This overlapping interval is associated with the end-stage overlap identifier to distinguish it from overlap that runs through the entire phase transformation process.

[0056] The overlap ratio analysis of the overlapping intervals of the layers is used to obtain the overlapping intervals of the phase transition time. The width of the overlapping interval of the layers is divided by the width of the narrower layer time window in that layer pair to obtain the overlap ratio of that layer pair, expressed as a percentage: the closer the ratio is to 100%, the more completely the phase transition of the narrower layer falls within the phase transition period of the wider layer, the two transform together in the same time period, and the strongest stress superposition is achieved; the closer the ratio is to zero, the more staggered the two time sequences are, and there is almost no superposition. When there are two or more layers in the thickness direction, the overlap ratio is calculated for each layer separately, without cross-layer averaging, because the overlap between the surface and a quarter layer reflects more the uniformity of the cooling medium contact surface, while the overlap between a quarter layer and the core layer reflects more the degree of hysteresis of heat conduction in the thickness direction, and the two are not physically equivalent. The overlapping interval of the phase transformation time sequence is selected from the layer-level pairs with the highest overlap ratio. This interval represents the overlap status of the phase transformation time sequence in the thickness direction of the entire steel plate, because the highest ratio indicates the most concentrated phase transformation and the most likely stress superposition. When there are two layers with the highest overlap ratio, both intervals are retained as candidates without merging or discarding one. Layers with a zero overlap interval width have their overlap ratio recorded as zero, but they are still included in the ratio comparison. A zero ratio is also valid information for judging the degree of phase transformation synchronization in the thickness direction and is not excluded from the comparison range due to its zero width. The start and end times of the phase transformation time sequence overlapping interval, the layer-level pair number, and the overlap ratio value are all stored together for direct use when determining whether a region constitutes a concentrated area of ​​phase transformation stress superposition, without the need to recalculate the ratio.

[0057] The location of concentrated phase transformation stress is determined by comparing the overlapping intervals of the phase transformation time sequence with the synchronous judgment threshold. The overlap ratio of the overlapping intervals of the phase transformation time sequence is compared one by one with the synchronous judgment threshold obtained from the retrospective statistics of crack detection cases for this steel grade. If the ratio exceeds the threshold, it indicates that the phase transformation in the thickness direction is highly synchronous, multiple layers expand together at almost the same time, local stress accumulates in a short period of time and has no time to relax, and the risk of cracking is significant. The more the ratio exceeds the threshold, the stronger the synchronous superposition, and the more priority should be given to this location. The location of concentrated phase transformation stress is only established when the comparison result exceeds the threshold. The coordinate range corresponding to the overlapping interval of the phase transformation time sequence, the corresponding layer number, and the overlap ratio of the interval are taken as the record content of the location. If the ratio is exactly equal to the threshold, it is treated as exceeding the threshold, and no additional exception rule is set for the case of equality. If the overlap ratio of the overlapping interval of the phase transformation time sequence is lower than the threshold, the corresponding location is not judged as a location of concentrated phase transformation stress. The original ratio value is retained as a reference sample for the periodic review of the threshold for this steel grade. The review cycle is consistent with the revision cycle of the corresponding relationship table for this type of heating furnace. If only one set of layer pairs is used for comparison along the thickness direction of the same steel plate, and the proportion of this set does not exceed the threshold, the area of ​​concentrated phase transformation stress is determined to be an empty set, and the steel plate will not be further refined into crack risk coordinates in this stage. If multiple areas of concentrated phase transformation stress exist simultaneously, they are sorted from high to low according to their proportion values. When assessing the probability of crack initiation, the area with the highest proportion is treated first. Areas with the same proportion are processed in sequence according to the number of their respective layer pairs, so that the assessment order remains definite even when the proportions are the same, and inconsistent results are obtained due to different processing orders.

[0058] The crack initiation probability is assessed and output as a crack-sensitive area at locations of concentrated phase transformation stress. The crack initiation probability at each location of concentrated phase transformation stress is obtained by dividing the overlap ratio of that location, the amplitude of the core-surface cooling rate difference of the corresponding layer pair, and the mismatch level L recorded at the corresponding location in the stress concentration zone stage by the benchmark value of the same category, and then weighting them. The weights are the correlation regression coefficients of the three factors and the actual cracking results in previous crack detection cases of this steel grade. The overlap ratio represents the degree of phase transformation synchronization, the amplitude of the core-surface cooling rate difference represents the steepness of the temperature gradient, and L represents the existing residual stress mismatch. When all three are high, the cracking probability increases significantly. When the regression sample is insufficient, the general weight of the steel grade category is temporarily used. Crack-sensitive areas are classified into three levels—high, medium, and low—based on their crack initiation probability: High-level areas have a probability exceeding the lower limit of the top 10% of crack cases in historical statistics; low-level areas have a probability below the upper limit of the bottom 50% of non-crack cases; and medium-level areas fall between these two. The grading boundaries are not continuously interpolated, ensuring the grading directly aligns with the probability distribution of historically occurring and non-cracked cases. When the area of ​​concentrated phase transformation stress is an empty set, the crack-sensitive area is simultaneously classified as an empty set, and no alternative locations are selected from ordinary thickness difference locations or the background area. For high-level crack-sensitive areas, the coordinate range, the corresponding layer number, and the crack initiation probability value are all archived for priority monitoring during processing or inspection. For medium and low-level areas, only the probability value is archived. When the coordinate range of a concentrated phase transformation stress area highly overlaps with another area, the records of both crack-sensitive areas are retained separately, with the overlapping portion accurately reflected in the coordinate range for subsequent location-based sampling. The overlapping coordinates are not merged into a single record, thus concealing the true risk level of one of the areas.

[0059] Step S104: Based on the orientation inspection of the metallographic structure of the crack-sensitive area, the grain size level is obtained. For the grain size level, the abnormal points with low impact toughness are screened and included in the evaluation to obtain the low temperature impact performance value. Based on the low temperature impact performance value, the difference between the Z direction and the rolling direction is compared to obtain the anisotropy coefficient.

[0060] Specifically, grain size levels are obtained by directional metallographic examination of crack-sensitive areas. Higher crack-sensitive areas indicate a greater risk of microstructural anomalies; therefore, metallographic sampling density decreases progressively: three samples are taken from each interval in high-level areas, two from medium-level areas, and one from low-level areas. Sampling locations are evenly distributed within the interval coordinate range to cover possible microstructural fluctuations, avoiding concentration at interval boundaries or a single central point. After grinding and polishing, metallographic samples are etched with a 4% (v / v) nitric acid-alcohol solution to reveal grain boundaries. The etching time follows the standard process specifications for the steel grade. Samples with insufficient or excessive etching are re-etched, and the original number is retained in the inventory with an added version number; these are not used for interpretation. Grain size levels are calculated by counting the number of grain boundaries intersecting with the truncation line under a microscope, field by field, using the cross-section method. At least three fields of view are counted for each sample, and the arithmetic mean is taken as the grain size level. When the results differ by more than one level between fields of view, additional fields are counted, and the average value after expanding the field of view is used; the initial statistical results are not retained. For steel plates with empty crack-sensitive areas, metallographic inspection is still conducted using the standard sampling ratio for that steel grade, with sampling points evenly distributed across the plate surface and the density consistent with lower-level areas. This ensures that even steel plates without sensitive areas retain a structural baseline, preventing the complete skipping of metallographic inspection simply because no sensitive area exists. Grain size levels are graded and archived according to sampling coordinates and their respective crack-sensitive areas. When multiple samples within the same range have inconsistent levels, all original results are retained. Since the dispersion of levels itself is an input for determining mixed crystals, they are not merged into a single representative value, nor are the mode value taken for simplification due to a large number of samples. Metallographic microstructure inspection is performed according to standard etching processes. The roughness of the sample grinding and polishing must meet the requirements for grain boundary clarity for microscopic observation. Inadequate polishing will cause the statistical results of the grain size level cutoff line to deviate from the true value. Therefore, the polishing process must be visually inspected and confirmed one by one before etching.

[0061] In some embodiments, the step of screening for low impact toughness anomalies at the grain size level and classifying them into the low-temperature impact performance value includes: determining a directional sampling scheme based on the grain size level to locate the grain size mixed crystal transition region; conducting a low-temperature impact test using the directional sampling scheme to obtain impact energy measurement values; screening for low impact energy anomalies at the impact energy measurement values ​​and recording the characteristics of the anomalies; and evaluating the abnormal correlation between grain size and impact toughness using the characteristics of the anomalies to obtain the low-temperature impact performance value.

[0062] The directional sampling scheme is determined by locating the mixed-grain transition region based on grain size level. If two or more samples exist within the same crack-sensitive area and the grain size difference is more than one level, this area is identified as a mixed-grain transition region. The plastic deformation capacity differs on both sides of the coarse-fine grain boundary, leading to strain concentration at the interface under impact, which easily becomes the initiation point of brittle fracture and is a high-risk source of abnormal impact toughness. Areas with a grain size difference of less than one level are not considered mixed-grain regions, even if there are many samples within the area. Such fluctuations reflect normal microstructural inhomogeneity rather than process abnormalities and are not included in the directional sampling scheme. The directional sampling scheme focuses on the mixed-grain transition region, with three sampling points arranged along the transition direction from fine to coarse grain size at each location, spaced one-quarter of the region's span. These points fall on the fine-grain side, the transition center, and the coarse-grain side, respectively, to capture toughness changes on both sides and at the interface. The directional sampling scheme also includes control sampling points in areas with uniform grain size and no signs of mixed crystals, with the number not less than one-third of the total number of mixed crystal sampling points. This is used to distinguish whether the abnormal toughness originates from mixed crystals or is a normal fluctuation of the steel grade itself. For steel plates without mixed crystal transition areas, the directional sampling scheme only retains conventional control sampling points, with the number consistent with the standard sampling inspection for that steel grade, without adding any additional points. The criterion for determining mixed crystal transition areas comes from the grade distribution formed by statistical analysis of the grain size level of the same batch, rather than an artificially set fixed threshold. This ensures that steel plates of different thicknesses and different grain size baselines each adopt mixed crystal criteria commensurate with their microstructure characteristics, avoiding the same fixed grade difference standard being too strict for fine-grained steel grades and too lenient for coarse-grained steel grades.

[0063] Impact energy measurements were obtained through low-temperature impact testing using a directional sampling scheme. At each sampling point, a standard Charpy V-notch impact specimen was taken along the thickness direction of the plate, with the notch uniformly aligned to allow crack propagation along the thickness direction to assess resistance to brittle fracture in that direction. The specimen was subjected to impact immediately after being held at the specified low-temperature impact test temperature for the steel grade. Specimens with insufficient or excessive holding time were discarded and re-tested, but included in the batch under the original sampling point number. The pendulum energy of the testing equipment was calibrated according to the range for the steel grade. The pendulum energy deviation was checked with standard specimens before each batch; if the deviation exceeded the allowable range, the test was paused and the equipment recalibrated. Impact energy measurements were recorded individually by specimen number, in joules, following the same numbering rules as the sampling point numbers in the directional sampling scheme. Multiple parallel specimens from the same sampling point were recorded separately, and the average was not taken initially; the average was reserved for anomaly screening as needed. In the directional sampling scheme, control points and mixed crystal points are mixed and placed in the same batch, and randomly arranged for testing according to sample number, without grouping and continuous testing according to sampling point type. This prevents the pendulum from introducing systematic bias due to thermal drift from continuous testing of similar samples. Simultaneously with recording the impact energy measurement value, the sample fracture surface is photographed and archived for verification when screening for anomalies, checking whether the dimples or cleavage characteristics of the fracture surface match the impact energy measurement value. Samples with predominantly dimple fracture surfaces typically correspond to ductile fracture with normal impact energy measurement values, while samples with predominantly small cleavage planes often correspond to brittle fracture with lower impact energy measurement values. The morphological differences between the two provide independent evidence, independent of the numerical values, for subsequent determination of whether anomalies represent a genuine decrease in toughness rather than accidental fluctuations in measurement, making anomaly screening less reliant on statistical criteria.

[0064] Screening for low impact energy measurements reveals anomalies, and their characteristics are recorded. First, the mean and standard deviation of all impact energy measurements within the same sampling batch are calculated. A sample with an impact energy lower than the mean minus two times the standard deviation is identified as an anomaly. This criterion aligns with the normal distribution of low-temperature impact data for that steel grade and is not changed temporarily due to a small batch sample size. If the impact energy measurement itself is lower than the lower limit of acceptable quality specified in the technical conditions for that steel grade, it is directly identified as an anomaly regardless of whether it meets the statistical anomaly criteria. This type of non-compliance takes precedence over statistical criteria, and the judgment remains unchanged even if the overall mean of the batch is high. Samples bordering on twice the standard deviation or exhibiting cleavage characteristics are also included as anomalies. Anomaly characteristics are recorded as follows: the coordinates of the sampling point to which the anomaly belongs, whether it is located in the mixed crystal transition region, the corresponding grain size level, and the fracture morphology type. The deviation of the impact energy measurement from the mean at that point is also included. If any of the four items is missing, the characteristics are incomplete but the sample remains in the list for processing as usable information during correlation assessment. When the number of anomalies in the same batch exceeds 30% of the total number of samples, the anomaly characteristics are additionally associated with a batch-level anomaly identifier, indicating that the overall impact toughness of this batch is low and the heating or cooling process needs to be rechecked rather than just looking at individual points. A batch with no anomalies indicating low impact energy is recorded as an empty set and proceeds directly to the next stage; the normal sample with the lowest impact energy is not forcibly selected as an anomaly. The fracture morphology type recorded in the anomaly characteristics and whether the point is located in the mixed-crystal transition region corroborate each other: anomalies located in the mixed-crystal region with cleavage characteristics have the most reliable correlation between low toughness and grain size inhomogeneity; anomalies located in non-mixed-crystal regions but with cleavage characteristics suggest that there may be other structural anomalies not covered by the mixed-crystal inspection, requiring further attention in subsequent processing or inspection stages.

[0065] Low-temperature impact performance values ​​are obtained by evaluating the anomalous correlation between grain size and impact toughness using outlier features. For outliers located in the mixed-crystal transition region, the anomalous correlation between grain size and impact toughness is calculated as the product of the deviation of the impact energy from the mean at that point and the difference in grain size level within the mixed-crystal region. The larger the deviation and the larger the level difference, the stronger the bond between the decrease in toughness and grain size inhomogeneity. A larger product tends to attribute the low toughness to mixed crystals rather than fluctuations in the material itself. The product itself is not normalized and retains the original dimensional combination. For outliers with incomplete features or located in non-mixed-crystal regions, the anomalous correlation is calculated as the average correlation of similar historical outliers for that steel grade, varying point by point with the recorded deviation, without uniform rounding. The low-temperature impact performance value is based on the average of all impact energy measurements for that batch. Outliers with an anomalous correlation higher than the median of all correlations for that steel grade are removed from the mean calculation, while those lower than the median are still included in the mean, ensuring that low values ​​caused by mixed crystals do not contaminate the performance baseline representing the entire batch. If, after removing samples with low-temperature impact performance values, less than half of the total number of valid samples remain, or if all samples are deemed invalid, the baseline value is recalculated using all data from before the removal, with a note indicating that the removal ratio was too high. This ensures that the baseline value for low-temperature impact performance is never empty. Low-temperature impact performance values ​​are archived according to sampling batch, crack sensitivity zone, and sampling direction, with each direction used independently as input for anisotropy comparison. A baseline value record is retained before and after removing highly correlated anomalies. The difference between the two records reflects the degree of toughness drag caused by mixed crystals; a larger difference indicates that the overall impact toughness performance of that batch is more dependent on whether the influence of mixed crystals is removed.

[0066] Anisotropy coefficients are obtained by comparing the differences between the Z-direction and rolling direction based on low-temperature impact performance values. The determination of low-temperature impact performance values ​​covers two sampling directions: the Z-direction sample is the sample taken along the thickness direction as mentioned above, and the rolling direction sample is taken together at the same sampling point in the orientation sampling stage, perpendicular to the rolling direction. Samples in both directions are prepared under the same batch and heat treatment conditions, differing only in the sampling direction, to ensure that the comparison of directionality is not influenced by heat treatment differences. The number of samples in both directions is kept consistent in principle. The anisotropy coefficient is the ratio of the Z-direction low-temperature impact performance value to the rolling direction low-temperature impact performance value: a ratio less than 1 indicates that the Z-direction toughness is lower than that in the rolling direction. This is consistent with the principle that rolling flattens and elongates grains along the rolling direction, causes inclusions to form layers along the rolling direction, and makes it easier for inclusion layers and grain boundaries to cleave when impacted in the Z-direction. The closer the ratio is to 1, the weaker the directionality of the batch, and the closer the thickness-direction resistance to lamellar tearing is to that in the rolling direction. When the low-temperature impact performance value in the rolling direction is temporarily replaced by a general value of the same batch due to insufficient sample, the anisotropy coefficient-related replacement value should be identified and the reason for its use and the expected completion time should be noted. If the anisotropy coefficient is lower than the lower limit of the qualified level specified in the technical conditions of the marine engineering steel plate of that type, the Z-direction toughness of that batch is judged to be unqualified, and the overall qualification cannot be judged solely based on the qualified rolling direction. The anisotropy coefficient is archived by batch number and together with the crack sensitivity classification, grain size level, and low-temperature impact performance value of that batch, constitutes a complete mechanical property file for that batch, which is available for unified access in the quality assessment process. The file format is consistent with the archiving numbering rules of the previous steps of that batch. For steel plates of the same type with a consistently low anisotropy coefficient, the corresponding differential temperature heating regime and segmented cooling parameters will be traced in the next step, serving as an important input basis for judging whether the process window of that type needs to be systematically adjusted.

[0067] Step S105: Based on the anisotropy coefficient, trace back the failure intervals of each failure to form a failure risk interval. Based on the failure risk interval, dynamically track the differences in cooling phase change behavior to obtain the process window boundary value. Calculate the degree of deviation of heating parameters based on the process window boundary value and output a heating correction command.

[0068] Specifically, failure risk intervals are formed by tracing back past failure intervals based on the anisotropy coefficient. For batches with anisotropy coefficients below the lower limit of the acceptable range for that steel type, their original process parameter combinations (differential heating zone temperature, steady-state holding time, and the rate of each stage of segmented cooling, etc.) are extracted one by one and compared with the process parameters of failed batches recorded in the historical failure database for that steel grade. The extraction covers all key parameters from heating to cooling, without omitting any step. The range of process parameter values ​​for each batch that actually cracks or fails flaw detection in the historical database is recorded as a failure interval, recorded separately according to parameter category. If multiple historical failure intervals of the same parameter overlap, they are merged into a wider interval, with the boundary taken as the outermost value of each interval, and no overlapping narrow intervals are retained. For the current batch with an unacceptable anisotropy coefficient, if its process parameters fall within a certain failure interval, or the distance from the interval boundary is less than 10% of the historical fluctuation range of that parameter category, the failure interval is determined to be related to the current batch. The introduction of a distance threshold also includes batches that are close to the failure boundary but have not yet crossed the boundary in the early warning. The failure risk interval is formed by taking the union of all relevant failure intervals according to parameter category. Different categories are not merged and are archived separately. If there are fewer than two historical relevant failure intervals under the same category, they are still retained as a single interval. Batches with qualified anisotropy coefficients are not traced back to their history, and their failure risk intervals are judged as empty sets. They directly enter the regular process window evaluation process without occupying the comparison resources of the failure database. Once the failure risk interval is formed, it becomes the boundary condition for delineating the batch range when retrospectively tracing the cooling phase change behavior record. This boundary condition is defined separately according to parameter category. The screening of batches in different categories does not interfere with each other, so that the retrospective of cooling phase change behavior always focuses on historical samples with process conditions similar to the current non-conforming batches.

[0069] In some embodiments, the step of dynamically tracking the differences in cooling phase change behavior based on the failure risk interval to obtain the process window boundary value includes: collecting phase change curves of each batch based on the failure risk interval to obtain cooling phase change behavior records; comparing the phase change timing and cooling rate differences of each batch based on the cooling phase change behavior records to determine the phase change behavior difference amount; calculating the boundary offset weight through the phase change behavior difference amount to form a weight update record; and evaluating the boundary convergence stability of the weight update record to obtain the process window boundary value.

[0070] Cooling phase transition behavior records are obtained by collecting phase transition curves from previous batches based on the failure risk interval. The upper and lower limits of the parameters given by the failure risk interval define the range of batches that need to be traced back: for previous batches of the same specification falling within this range, the original curves of the surface cooling rate difference and phase transition timing deviation recorded in stage S103 are retrieved batch by batch and generated as read-only copies. Batches outside the range are not included in the retrieval, and the retrieval action does not modify the original files. The cooling phase transition behavior records are based on batch numbers. Each batch retains the complete curve of the surface cooling rate difference changing over time, the phase transition start time of each depth level, and the crack sensitivity classification result corresponding to the batch. The three timestamps are uniformly aligned to the cooling start time of the batch, so that the phase transition behavior of different batches can be compared on the same time coordinate. When the failure risk interval covers multiple parameter categories, the cooling phase transition behavior records are collected into the same batch data according to parameter category. The same batch can be referenced multiple times because it falls into multiple risk interval categories, and the references are not deduplicated and merged. For parameter categories with fewer than 5 batches in each batch, the cooling phase transformation behavior records are compiled according to the actual number of batches, and the difference between the actual number of batches and the minimum number of batches typically required for this category is noted, indicating that the sample representativeness of this category is limited. The crack sensitivity classification results retained for each batch in the cooling phase transformation behavior records, although not directly involved in the calculation of the phase transformation behavior difference in this step, are archived as supporting evidence of whether the phase transformation behavior of that batch ultimately leads to actual risk. This allows for cross-verification during process review when adjusting the tracing boundary, preventing boundary adjustments from being based solely on numerical indicators and detached from actual quality consequences.

[0071] The difference in phase change behavior is determined by comparing the phase change timing and cooling rate differences of each batch based on the cooling phase change behavior records. First, the difference between the phase change start time and the surface cooling rate of all batches in the cooling phase change behavior records is taken. The batch average is calculated according to the same parameter category as the reference benchmark for that category. This is recalculated every five new batches to track operating condition drift. The reference benchmark is calculated separately for each category and is not used across categories. The difference between the phase change timing and the reference benchmark for each batch, and the difference between the cooling rate and the reference benchmark, are divided by the reference benchmark of the same category to achieve dimensionlessness. The normalized relative deviation of the phase change timing is denoted as Δt_pt, and the relative deviation of the cooling rate is denoted as Δv_c. Both are dimensionless quantities with unified dimensions and can be directly weighted. The phase transformation behavior difference is calculated using the formula D = w1·|Δt_pt| + w2·|Δv_c|, where w1 and w2 are dimensionless weighting coefficients taken from the correlation regression coefficients of timing deviation and rate deviation on the actual cracking results in failure cases of this steel grade. Since D is dimensionless, a larger value indicates a more significant deviation of the phase transformation behavior from the reference baseline. When the cooling rate difference of a certain batch in the cooling phase transformation behavior record is incomplete due to equipment malfunction, the phase transformation behavior difference is calculated using Δt_pt alone, w2 is treated as zero, and a single-item calculation identifier is associated to indicate that the confidence level of the D value is low. The phase transformation behavior difference is archived by batch number and parameter category for categorized retrieval when calculating boundary offset weights. The relative magnitudes of Δt_pt and Δv_c in the phase change behavior difference can also be used to preliminarily determine whether the abnormality of this batch is mainly due to the lag in the phase change timing or the deviation in the cooling rate itself. This provides clues for process review to further distinguish whether the deviation of this batch from the reference benchmark is caused by the heating and heat preservation stage or the segmented cooling stage. This distinction helps to accurately target the boundary adjustment to the process stage that truly needs correction rather than adjusting the entire set of parameters in a general way, and also avoids mistakenly attributing the deviation that belongs to the heating stage to the cooling stage.

[0072] Boundary offset weights are calculated based on the phase transformation behavior difference to form a weight update record. For batches with a large phase transformation behavior difference value (D), although their process parameters still fall within the current process window, their actual phase transformation behavior has significantly deviated from historical normal levels. This indicates that the current window boundary is insufficient to constrain these batches and better exposes boundary positioning deviations. Therefore, a higher weight is given to boundary adjustment. The weight is monotonically increasing with D, rather than linearly proportional; the specific increasing function is derived from the fitting results of previous weight calibration tests for this steel grade. The boundary offset weight is calculated as the proportion of the phase transformation behavior difference of this batch to the sum of the differences of all batches in the same parameter category, ranging from zero to one, with the sum of all batches always being one. It is renormalized after each new batch. The weight update record records the boundary offset weight and the process parameter offset direction for each batch. The offset direction is the positive or negative deviation of the actual parameter value of the batch relative to the center of the current process window. The weight is quantitative, and the direction is directional; both determine which direction and how much the boundary is pushed for this batch. Missing either one indicates an incomplete record. Batch boundary offset weights with zero phase transition behavior differences are recorded as zero, and the entries remain in the weight update record. Zero-weight entries are used to identify stable boundary segments when statistically analyzing consecutive batches with no change. Weight update records are appended sequentially by batch time, without overwriting existing entries. When correcting data from the same batch, new weights are appended as independent entries, and old entries are marked as replaced. The replacement relationship is indexed by both batch number and append time. When the offset direction in the weight update record remains consistent across multiple consecutive batches, it indicates that there is indeed a systematic offset at the current process window boundary rather than random fluctuations. These consecutive records in the same direction will be reflected as cumulative adjustments in the same direction when collecting the boundary change sequence in the next step.

[0073] For example, the step of evaluating the boundary convergence stability of the weight update records to obtain the process window boundary value includes: collecting continuous batches of boundary value fluctuation records from the weight update records to establish a boundary change sequence; calculating whether the rate of change of fluctuation amplitude approaches zero based on the boundary change sequence to obtain a convergence determination result; statistically analyzing the number of batches with no continuous change based on the convergence determination result to establish a convergence trend table; and comparing the convergence trend table with the boundary convergence warning threshold to obtain the process window boundary value.

[0074] A boundary change sequence is established based on the boundary value fluctuation records of multiple consecutive batches collected from the weight update records. The boundary offset weights and directions of multiple consecutive batches in the weight update records are converted batch by batch into the actual adjustment amount of that batch to the process window boundary: the adjustment amount equals the batch weight multiplied by the boundary adjustment step size for that category (step size is taken in units of this parameter), plus the sign of the offset direction. The weight determines the adjustment magnitude, and the direction determines the sign of the adjustment. The step size is calibrated according to the boundary adjustment verification test for that steel grade and is verified every certain number of batches. The boundary change sequence records the converted adjustment amount of each batch in batch time sequence, continuously increasing with the accumulation of batches without rolling over. The total number of sequence positions corresponds one-to-one with the total number of weight update record entries by batch number. In the weight update records, batches with a weight of zero have an adjustment amount of zero. The boundary change sequence normally records a zero value for this position, with consecutive position numbers, for direct use when statistically analyzing batches with no continuous changes. When the adjustment amounts of two adjacent batches in the boundary change sequence are in opposite directions, it indicates that the boundary is oscillating between the two directions and has not yet converged. The oscillation record is retained with its original value; the alternation of positive and negative values ​​itself provides information for judging the convergence progress. Boundary change sequences are established separately according to process parameter categories, and boundary change sequences of different categories are not mixed. The length of the boundary change sequence accumulates continuously with production batches. The longer the sequence, the better it reflects the true trend of boundary adjustment. Short sequences are easily dominated by the random fluctuations of individual batches. Therefore, this step does not set an upper limit on the length of the sequence and allows it to grow naturally with the production process. This also means that the convergence judgment obtained when the number of early batches is insufficient is only for reference and not as the basis for final boundary confirmation. It is re-evaluated after the number of batches has accumulated continuously.

[0075] The convergence determination result is obtained by calculating whether the rate of change of fluctuation amplitude approaches zero based on the boundary change sequence. The boundary change sequence is divided into several consecutive windows with a fixed window length, and adjacent windows overlap by half a window length. The range of all adjustments within the window is calculated for each window as the fluctuation amplitude A_k of that window, where k is the window number numbered chronologically. The window length is taken as the minimum number of batches covering a complete oscillation cycle in the verification test of this steel grade. If it is too short, it is easy to misjudge the occasional fluctuation of a single batch as the overall fluctuation of the window; if it is too long, it will mix the converged segment and the non-converged segment into the same window, thus delaying the zero-reach determination. The rate of change of fluctuation amplitude is calculated using the formula r_k=(A_k-A_{k-1}) / A_{k-1}, where A_k is the fluctuation amplitude of the current window, taken from the range within the window of the boundary change sequence, and A_{k-1} is the fluctuation amplitude of the previous window. r_k is dimensionless; when it approaches zero, it indicates that the fluctuation amplitude no longer changes significantly with the increase of batches, and is a signal of boundary stabilization. When A_{k-1} is zero, r_k is judged as undefined, and the convergence judgment result of this window is marked as pending. The convergence judgment result is recorded window by window, including the actual value of r_k and whether it approaches zero: if the absolute value of r_k is lower than the zero-approaching threshold calibrated by the boundary convergence verification test for this steel grade, it is considered converged; otherwise, it is considered not converged. When the number of batches in the boundary change sequence is less than a complete window, the convergence judgment result is not calculated temporarily, and is judged after a window is filled, so that the convergence judgment is always based on a sufficiently long fluctuation sample. Even if the absolute value of r_k in the convergence determination result is lower than the zero-reaching threshold, it needs to maintain this state for multiple consecutive windows to have process significance. A brief zero-reaching of a single window may just be a small fluctuation that occasionally occurs in the boundary change sequence at that point, rather than the boundary truly entering a stable state. This can be distinguished by cumulative counting when counting the number of batches with no continuous changes in the next step, so as to avoid misjudging an occasional low-fluctuation window as the process window having converged.

[0076] A convergence trend table is established by statistically analyzing the number of batches with no continuous change based on the convergence determination results. The number of windows with consecutive convergence markers in the convergence determination results is accumulated sequentially by batch. Once a non-convergence marker appears, the count is reset to zero and restarted. The zeroing action is recorded as an independent event, along with the zeroing batch number and the count value before zeroing, allowing the length of each continuous convergence segment to be traced individually. The convergence trend table records the maximum number of consecutive convergence windows reached before each zeroing and the start and end batch numbers of that count. Counts for the same parameter category are arranged chronologically without overwriting updates, and each count is an independent segment of the convergence history of the category boundary. Windows marked as pending in the convergence determination results are not included in the continuous convergence count and do not trigger zeroing; they are considered vacant, and the count continues normally before and after the pending window. When the convergence determination results show convergence across multiple consecutive windows but have not yet reached the stable determination batch number, the convergence trend table accurately records the current counting progress, enabling reviewers to determine how far the category is from true convergence. When a consecutive convergence count in the convergence trend table reaches the number of batches required for stability, that count is associated with a stability achievement flag. Once marked, the flag is retained as a condition for triggering updates to the process window boundary values. If the same parameter category in the convergence trend table has historically approached the number of batches required for stability multiple times but was interrupted by a non-convergence flag in the last window, these near-convergence segments that were interrupted are also fully retained in the convergence trend table and are not deleted due to failure to ultimately reach the target. This allows process reviewers to determine which batch intervals the recurring unstable phases during boundary adjustments for this category occur in, providing clues for further investigation into unidentified process disturbance sources. These disturbance sources may originate from the equipment itself rather than improper process parameter settings.

[0077] The process window boundary values ​​are obtained by comparing the convergence trend table with the boundary convergence warning threshold. The latest count marked with a stability achievement indicator in the convergence trend table, along with the cumulative adjustment value of the boundary change sequence within the corresponding batch interval, is added to the original set value of the current process window boundary for that category to obtain the candidate boundary value. This accumulation only considers the batch interval given in the convergence trend table and does not include batches outside the interval. The candidate boundary value is compared with the boundary convergence warning threshold: the warning threshold is the upper limit of the process window adjustment allowed by the technical conditions for that steel grade. When the candidate boundary value exceeds this threshold, the process window boundary value is taken from the warning threshold itself, and the excess is recorded for review to prevent a single adjustment from exceeding the permitted range of the technical conditions. When the convergence trend table has not yet shown a stability achievement indicator, the process window boundary value uses the current effective value without adjustment, ensuring that the batches in production always have a stable comparison benchmark. When the same category in the convergence trend table has repeatedly achieved and then revoked the stability achievement indicator, the historical versions of the process window boundary values ​​increase accordingly, with each version corresponding to a true stability determination. The process window boundary values ​​are maintained separately according to process parameter categories, and the categories are not linked. Each update of the process window boundary value simultaneously records the specific values ​​before and after the update, the entry number of the convergence trend table, and the remaining margin of the candidate boundary value relative to the warning threshold, forming a complete boundary evolution chain. This allows any process window boundary value that takes effect at any historical moment to be traced back to the convergence evidence at that time. If it is found that a certain version of the boundary value still causes batch non-compliance, it is possible to accurately locate which update introduced the deviation, rather than blaming it on the entire boundary maintenance mechanism. It is also convenient to revise the convergence criteria on which the update is based separately.

[0078] Heating correction commands are output based on the deviation of heating parameters calculated according to the process window boundary values. In the next heat's differential temperature heating regime, the planned temperature increase for each heating unit is compared one by one with the process window boundary values ​​of its respective parameter category: the deviation is calculated as the difference between the planned value and the boundary value divided by the boundary value itself, expressed as a percentage. A positive value indicates the plan exceeds the boundary, and a negative value indicates it is below the boundary. When the deviation exceeds the adjustment trigger ratio specified for that steel grade, the planned temperature increase for that heating unit is reset according to the process window boundary values ​​and fed back to the differential temperature heating regime generation stage before actual execution in this heat, allowing the correction to be directly invoked in the next heat without affecting the current heat. Heating correction commands record whether a reset has been triggered, the planned value before triggering, and the boundary value after triggering for each unit; for units that have not been triggered, only the deviation is recorded, and no correction action is generated. Both are listed side-by-side in the same command according to unit number. When the process window boundary value changes due to the update of the convergence trend table within the same heat, the heating correction command still uses the boundary value read at the start of the current heat and does not switch the comparison benchmark midway, ensuring that all heating units in the same heat use the same version of the boundary value. The heating correction command summarizes the comparison results of all triggered and non-triggered units according to the heating unit number. When multiple units in the same heat are triggered to reset simultaneously, they are sorted from largest to smallest absolute value of deviation, and the unit with the most serious deviation is corrected first, avoiding the simultaneous large adjustment of multiple adjacent units and the introduction of a new temperature gradient in the next heat. If the planned value of a unit still exceeds the gradient upper limit after the boundary value is reset, the command synchronously inserts the median increase of the transition unit to keep the corrected differential temperature distribution smooth, and then sends it to the differential temperature heating system generation stage of the next heat along with the process window boundary value version number on which it is based.

[0079] To implement the heat treatment method for high-strength steel plates used in marine engineering corresponding to the above method embodiments, in order to achieve the corresponding functions and technical effects. See also Figure 3 , Figure 3 This diagram illustrates a structural block diagram of a heat treatment apparatus 300 for high-strength steel plates used in marine engineering, according to an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The heat treatment apparatus 300 for high-strength steel plates used in marine engineering, according to an embodiment of this application, includes:

[0080] The stress analysis module 301 is used to obtain raw material specification data based on the chemical composition and thickness distribution of steel plate raw materials, perform normalizing pretreatment to generate normalized structure according to the raw material specification data, and determine the phase transformation stress distribution by analyzing the residual stress state of the normalized structure.

[0081] The heat compensation control module 302 is used to locate the stress concentration area by the phase change stress distribution to determine the differential temperature heating regime, analyze the heat dissipation lag trend of the differential temperature heating regime to obtain the differential temperature heat compensation increment, and carry out pre-compensation heating control to establish a steady-state heat preservation regime by means of the differential temperature heat compensation increment.

[0082] Crack identification module 303 is used to obtain the surface cooling rate by performing segmented cooling control according to the steady-state heat preservation system, calculate the core cooling rate gradient based on the surface cooling rate to obtain the core-surface cooling rate difference, and screen out the parts where phase transformation stress is superimposed and concentrated from the core-surface cooling rate difference to output the crack sensitive area.

[0083] The performance evaluation module 304 is used to obtain the grain size level based on the directional inspection of the metallographic structure in the crack sensitive area, screen out abnormal points with low impact toughness based on the grain size level and classify them into the evaluation to obtain the low temperature impact performance value, and obtain the anisotropy coefficient by comparing the difference between the Z direction and the rolling direction based on the low temperature impact performance value.

[0084] The feedback correction module 305 is used to trace the failure intervals of each failure based on the anisotropy coefficient to form a failure risk interval, dynamically track the differences in cooling phase change behavior based on the failure risk interval to obtain the process window boundary value, and calculate the degree of deviation of heating parameters based on the process window boundary value to output a heating correction command.

[0085] The aforementioned heat treatment apparatus 300 for high-strength steel plates used in marine engineering can implement a heat treatment method for high-strength steel plates used in marine engineering as described in the above-described method embodiments. The options in the above method embodiments are also applicable to this embodiment and will not be detailed here. The remaining contents of this application's embodiments can be referred to the contents of the above method embodiments, and will not be repeated in this embodiment.

[0086] The above embodiments are not an exhaustive list based on the present invention, and there may be many other embodiments not listed. Any substitutions and improvements made without departing from the concept of the present invention are within the protection scope of the present invention.

Claims

1. A heat treatment method for high-strength steel plates used in marine engineering, characterized in that, include: Based on the detection of chemical composition and thickness distribution of steel plate raw materials, raw material specification data is obtained. Normalizing pretreatment is carried out according to the raw material specification data to generate normalized structure. The phase transformation stress distribution is determined by analyzing the residual stress state of the normalized structure. The phase change stress distribution is used to locate the stress concentration area and determine the differential temperature heating regime. The heat dissipation lag trend of the differential temperature heating regime is analyzed to obtain the differential temperature heat compensation increment. The differential temperature heat compensation increment is used to carry out pre-compensation heating control to establish a steady-state heat preservation regime. The surface cooling rate is obtained by segmented cooling control according to the steady-state heat preservation system. The core cooling rate gradient is calculated based on the surface cooling rate to obtain the core-surface cooling rate difference. Crack-sensitive areas are output from the core-surface cooling rate difference to screen the areas where phase transformation stress is superimposed and concentrated. Grain size level is obtained by directional inspection of metallographic structure in the crack-sensitive area. Abnormal points with low impact toughness are screened for the grain size level and classified into evaluation to obtain low temperature impact performance value. Anisotropy coefficient is obtained by comparing the difference between Z direction and rolling direction based on the low temperature impact performance value. Based on the anisotropy coefficient, the failure risk range is formed by tracing the failure ranges of each failure. The process window boundary value is obtained by dynamically tracking the differences in cooling phase change behavior according to the failure risk range. The degree of deviation of heating parameters is calculated based on the process window boundary value, and a heating correction command is output.

2. The method according to claim 1, characterized in that, The determination of phase transformation stress distribution through residual stress state analysis of the normalized microstructure includes: The layout of stress measurement points is determined based on the characteristic regions of the normalized structure identification plate. Based on the stress measurement point layout, the residual stress amplitude of each measurement point is determined to obtain the stress amplitude sequence; The stress dispersion is determined by calculating the span of the stress plateau distribution between the measurement points for the stress amplitude sequence. The stress dispersion is used to extract a large-scale medium stress zone to form a phase transition stress distribution.

3. The method according to claim 1, characterized in that, The step of determining the differential temperature heating regime by locating the stress concentration region using the phase transformation stress distribution includes: A thickness distribution map is established based on the described phase transformation stress distribution using stress-thickness coupling mapping. The thickness distribution map is used to analyze the thickness difference phase transition mismatch locations to determine candidate stress concentration areas; The stress concentration candidate region is obtained by measuring the level of mismatched stress superposition. Based on the stress concentration zone, the boundary coordinates of the partitions are extracted to construct the differential temperature heating system.

4. The method according to claim 1, characterized in that, The step of analyzing the heat dissipation lag trend of the differential temperature heating system to obtain the differential temperature compensation increment includes: Based on the differential temperature heating system, the heat dissipation lag trend of steel plates of the same specification in previous tests is identified to form lag trend parameters. The predicted heat dissipation amount is determined by analyzing the hysteresis extrapolation of the heat dissipation hysteresis at the board edge based on the aforementioned hysteresis trend parameters. The heat compensation correction coefficient is determined by calculating the deviation of previous underheating predictions based on the predicted heat dissipation. The differential temperature heat compensation increment is determined by calculating the total comprehensive heat compensation amount based on the aforementioned heat compensation correction coefficient.

5. The method according to claim 1, characterized in that, The step of identifying crack-sensitive areas from the location of concentrated phase transformation stress by screening the surface cooling rate difference includes: The layered temperature group is obtained by collecting the layered temperature records along the thickness direction of the plate based on the aforementioned cardiac surface cooling rate difference. The phase transition timing deviation is obtained by measuring the time difference of the phase transition initiation of each layer based on the aforementioned layer temperature group. The location of concentrated phase transition stress superposition is determined by analyzing the overlapping intervals of the phase transition timing using the phase transition timing deviation analysis. The probability of crack initiation is evaluated at the location where the phase transformation stress is superimposed and concentrated, and the crack sensitive area is output.

6. The method according to claim 1, characterized in that, The process of identifying and evaluating low impact toughness anomalies at the grain size level to obtain low-temperature impact performance values ​​includes: Based on the described grain size level, the directional sampling scheme is determined to locate the grain size mixed crystal transition region; The impact energy measurement value is obtained by conducting a low-temperature impact test using the aforementioned directional sampling scheme. Screen the impact energy measurements for abnormal points with low impact energy and record the characteristics of the abnormal points; The low-temperature impact performance value is obtained by evaluating the anomalous correlation between grain size and impact toughness using the aforementioned anomaly point features.

7. The method according to claim 1, characterized in that, The process window boundary value is obtained by dynamically tracking the differences in cooling phase transition behavior based on the failure risk range, including: Based on the failure risk range, the cooling phase transition behavior record was obtained by collecting the phase transition curves of each batch. The difference in phase change behavior is determined by comparing the phase change timing and cooling rate differences of each batch based on the cooling phase change behavior records. The boundary offset weight is calculated by measuring the difference in phase transition behavior to form a weight update record; The weight update record is used to evaluate the boundary convergence stability and obtain the process window boundary value.

8. The method according to claim 5, characterized in that, The step of determining the location of concentrated phase transition stress by analyzing the overlapping intervals of phase transition time through the phase transition time deviation includes: The layering time window is obtained by aggregating the phase transition time deviation of each layer in the thickness direction. The overlapping interval of the layers is obtained by measuring the deviation between the start and end times of the phase transition between adjacent layers through the layered time window. The overlapping ratio analysis of the layered overlapping intervals is used to obtain the phase transition time series overlapping intervals; The overlapping intervals of the phase transition time are compared and the synchronous judgment threshold is used to determine the location of the concentrated phase transition stress.

9. The method according to claim 7, characterized in that, The step of obtaining the process window boundary value by updating the weight record to evaluate the boundary convergence stability includes: A boundary change sequence is established by collecting consecutive batches of boundary value fluctuation records for the weight update records; The convergence determination result is obtained by calculating whether the rate of change of fluctuation amplitude approaches zero based on the boundary change sequence. A convergence trend table is established by statistically analyzing the number of batches that remain unchanged consecutively based on the convergence determination results. The process window boundary value is obtained by comparing the convergence trend table with the boundary convergence early warning threshold.

10. A heat treatment device for high-strength steel plates used in marine engineering, characterized in that, include: The stress analysis module is used to obtain raw material specification data based on the chemical composition and thickness distribution of steel plate raw materials, perform normalizing pretreatment to generate normalized structure based on the raw material specification data, and determine the phase transformation stress distribution by analyzing the residual stress state of the normalized structure. The heat compensation control module is used to locate the stress concentration area by the phase change stress distribution to determine the differential temperature heating regime, analyze the heat dissipation lag trend of the differential temperature heating regime to obtain the differential temperature heat compensation increment, and use the differential temperature heat compensation increment to carry out pre-compensation heating control to establish a steady-state heat preservation regime. The crack identification module is used to obtain the surface cooling rate by performing segmented cooling control according to the steady-state heat preservation system, calculate the core cooling rate gradient based on the surface cooling rate to obtain the core-surface cooling rate difference, and screen out the areas where phase transformation stress is superimposed and concentrated from the core-surface cooling rate difference to output the crack sensitive area. The performance evaluation module is used to obtain the grain size level by directional inspection of the metallographic structure in the crack-sensitive area, screen out abnormal points with low impact toughness based on the grain size level and classify them into the evaluation to obtain the low temperature impact performance value, and obtain the anisotropy coefficient by comparing the difference between the Z direction and the rolling direction based on the low temperature impact performance value. The feedback correction module is used to trace the failure intervals of each failure based on the anisotropy coefficient to form a failure risk interval, dynamically track the differences in cooling phase change behavior based on the failure risk interval to obtain the process window boundary value, and calculate the degree of deviation of heating parameters based on the process window boundary value to output a heating correction command.

Images