Tower drum assembly sequence adaptive optimization method and system

Abstract

The present application relates to the technical field of welding process control, and particularly relates to a tower drum assembly welding sequence adaptive optimization method and system. The method comprises the following steps: obtaining cooling state data; performing heat field feature extraction according to the cooling state data to obtain heat field feature data; performing heat stress risk nucleus mapping processing according to the heat field feature data to obtain heat stress risk data; generating heat balance control targets according to the heat stress risk data to obtain control target data; generating initial welding actions according to the control target data to obtain initial welding action data; and reconstructing the heat stress welding sequence of the initial welding action data to obtain welding reconstruction data. Through the closed-loop control of heat field structure recognition and heat stress risk mapping, the present application realizes real-time adaptive reconstruction of the tower drum assembly welding sequence under low-temperature and gust disturbance conditions, effectively suppresses the risk of circumferential heat tearing and axial heat accumulation superposition, and reduces the occurrence probability of cold cracking and residual stress concentration.

Description

Technical Field

[0001] This invention relates to the field of welding process control technology, and in particular to an adaptive optimization method and system for tower assembly welding sequence. Background Technology

[0002] Wind turbine towers are typically assembled by segmented rolling followed by on-site or factory welding. Due to the large tower diameter, significant variations in plate thickness, and long weld lengths, the welding process demands high precision in heat input control and welding sequence arrangement. During construction in northern winters or at high altitudes, ambient temperatures can drop below -20°C, accompanied by gusts of wind. Low temperatures accelerate post-weld cooling, increasing the probability of hardened structures; gusts create non-uniform convection zones along the tower's circumference, leading to accelerated cooling on one side while continuous heat input on the other, resulting in circumferential thermal tearing. Simultaneously, as the circumferential weld advances continuously along its height, residual heat in the lower weld layer has not dissipated, while the upper area to be welded cools rapidly under wind cooling, easily forming axial heat accumulation steps, placing the weld in unfavorable bottom-hot, top-cold boundary conditions. Existing welding sequence control methods are mostly based on experience-based segmented symmetrical welding or fixed skip welding patterns, making it difficult to promptly detect localized thermal stress risks and adjust the sequence, easily leading to cold cracking, residual stress concentration, or deformation superposition problems. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes an adaptive optimization method and system for tower assembly welding sequence, thereby resolving at least one of the aforementioned technical problems.

[0004] This application provides an adaptive optimization method for the welding sequence of tower assembly, including the following steps:

[0005] Step S1: Obtain cooling status data; extract thermal field features based on the cooling status data to obtain thermal field feature data;

[0006] Step S2: Perform thermal stress risk center mapping processing based on thermal field characteristic data to obtain thermal stress risk data;

[0007] Step S3: Generate thermal balance control targets based on thermal stress risk data to obtain control target data;

[0008] Step S4: Generate initial welding actions based on control target data to obtain initial welding action data; reconstruct the welding sequence using thermal stress on the initial welding action data to obtain welding reconstruction data.

[0009] This invention enables real-time adaptive adjustment of the tower welding process under low temperature and gust wind disturbance conditions. Compared with traditional fixed-sequence or single-point temperature control methods, this method can simultaneously identify circumferential thermal tear structures and axial thermal accumulation steps, and transform the risk quantification results into executable welding action constraints and reconstruction strategies. This suppresses the superposition effects of local rapid cooling and thermal accumulation, reduces the probability of hardened structure formation and residual stress concentration, and improves the stability and forming consistency of tower welding under low temperature conditions.

[0010] Preferably, step S1 specifically includes:

[0011] S11: Obtain cooling status data;

[0012] S12: Calculate the cooling rate and thermal non-uniformity based on the cooling status data, and obtain the cooling rate data and thermal non-uniformity data respectively.

[0013] S13: Obtain welding process data, and perform thermal characteristic calculations based on cooling status data and welding process data to obtain thermal characteristic data. The thermal characteristic calculations include circumferential thermal calculations and axial thermal calculations, and the thermal characteristic data includes circumferential thermal characteristic data and axial thermal characteristic data.

[0014] S14: Integrate cooling rate data, thermal characteristic data, and thermal non-uniformity data to obtain thermal field characteristic data.

[0015] This invention couples cooling status data with welding process data for modeling, extracting responsive indicators such as cooling rate and thermal non-uniformity, and calculating circumferential and axial thermal characteristics to achieve a three-dimensional characterization of the thermal field's spatial structure. Compared to methods relying solely on single-point temperature or simple gradient analysis, this invention can distinguish the different contributions of heat input behavior and environmental disturbances to the thermal field evolution, improving the accuracy of identifying thermal tear zones and thermal accumulation zones. It provides a stable and structured foundation of thermal field characteristics for risk mapping and welding sequence reconstruction, enhancing the targeting and reliability of welding control under low-temperature conditions.

[0016] Preferably, the circumferential thermal calculation specifically involves:

[0017] The circumferential sector temperature field is constructed based on the cooling status data to obtain the circumferential temperature field data;

[0018] S131: Construct the circumferential sector temperature field based on the cooling status data to obtain circumferential temperature field data;

[0019] S132: Perform air-cooled coupling partitioning on the circumferential temperature field data to obtain air-cooled partitioning data;

[0020] S133: Perform circumferential thermal gradient localization on the air-cooled partition data to obtain circumferential thermal gradient data;

[0021] S134: Based on the welding process data, extract the heat input from the opposite side and detect the sudden change in cooling rate to obtain the heat input data from the opposite side and the sudden change in cooling rate data, respectively.

[0022] S135: Based on the circumferential thermal gradient data, the thermal input data on the opposite side and the abrupt change data of the cooling rate are used to determine the formation of thermal tearing rings, and circumferential thermal characteristic data are obtained.

[0023] This invention constructs a circumferential sector temperature field and performs wind-cooling coupling zoning to identify non-uniform convection structures under gusts. Combined with contralateral heat input extraction and cooling rate abrupt change detection, it achieves structural coupling analysis of "lower-side heating and upper-side cooling" or "unilateral rapid cooling," thereby determining the formation location and intensity of thermal tear rings. Compared to traditional single-point temperature or simple circumferential symmetric control methods, this invention can capture hidden circumferential gradient peaks and their spatial correspondence with welding behavior, improving the accuracy of identifying thermally unstable structures in low-temperature welding and providing a clear triggering basis for welding sequence reconstruction.

[0024] Preferably, the axial thermal calculation is as follows:

[0025] S136: Construct the axial height zone temperature field based on the cooling status data to obtain axial temperature field data;

[0026] S137: Perform axial thermal gradient calculation on the axial temperature field data to obtain axial thermal gradient data;

[0027] S138: Perform temperature step discrimination on the axial thermal gradient data to obtain temperature step data;

[0028] S139: Extract residual heat accumulation and gust cooling data based on axial temperature field data to obtain residual heat accumulation data and gust cooling data respectively.

[0029] S1310: Based on welding process data and temperature step data, advance alignment is performed to obtain advance alignment data;

[0030] S1311: Axial thermal characteristic data are obtained by quantifying axial thermal accumulation based on residual thermal accumulation data, gust cooling data, and propulsion alignment data.

[0031] This invention constructs an axial height temperature field and performs gradient and temperature step discrimination to identify the "lower stagnation and upper rapid cooling" thermal fault structure formed during continuous welding. Combined with residual heat accumulation and gust cooling extraction, it achieves a separation analysis of the superposition effects of thermal inertia differences and environmental disturbances. Furthermore, through a advancement alignment mechanism, it spatiotemporally correlates thermal structure changes with the welding behavior sequence, thereby quantitatively characterizing axial heat accumulation. Compared to traditional methods that only focus on circumferential symmetry or single-point temperature control, this invention can identify abnormal thermal boundaries in the height direction in advance, avoiding continued welding under unfavorable thermal conditions and improving the scientific rigor and stability of welding sequence adjustments in low-temperature environments.

[0032] Preferably, the temperature step discrimination is specifically as follows:

[0033] a. Extract the axial gradient peak segments from the axial thermal gradient data to obtain gradient peak segment data;

[0034] b. Perform upper-side rapid cooling feature discrimination and lower-side stagnation feature discrimination on the gradient peak segment data to obtain upper-side rapid cooling feature data and lower-side stagnation feature data respectively;

[0035] c. Based on the upper rapid cooling characteristic data and the lower heat stagnation characteristic data, the axial step structure is inspected to obtain temperature step data.

[0036] This invention extracts peak segments from the axial thermal gradient and separately identifies the upper-side rapid cooling characteristics and the lower-side stagnation characteristics, transforming simple gradient anomalies into step structures with clear physical meaning. Compared to methods that judge anomalies solely based on gradient peak values, this invention can distinguish between ordinary gradient changes caused by overall cooling and structural thermal faults—"hot at the bottom, cold at the top"—formed by continuous welding and gust disturbances, thus avoiding misjudgments. Simultaneously, through bidirectional verification of rapid cooling and stagnation, the stability and noise resistance of step identification are improved, providing a more reliable triggering basis for welding sequence adjustment and heat input control.

[0037] Preferably, step S2 specifically includes:

[0038] S21: Perform thermal field input encoding based on thermal field characteristic data to obtain thermal field input data;

[0039] S22: Obtain thermal setpoint data, and perform dual-path deviation calculation based on thermal field input data and thermal setpoint data to obtain dual-path deviation data, wherein the dual-path deviation calculation includes steady-state deviation calculation and transient deviation calculation.

[0040] S23: Competitively fuse the dual-path bias data to obtain competitive situation data;

[0041] S24: Based on the competitive situation data, the heat stress risk potential is propagated to obtain heat stress risk data.

[0042] This invention employs input encoding of thermal field characteristics and introduces a thermal setpoint for dual-path deviation calculation, separating and modeling steady-state thermal imbalance and transient disturbances to achieve differentiated identification of short-term fluctuations such as continuous thermal anomalies and gusts of wind. Furthermore, a competitive fusion mechanism prioritizes different deviation paths to avoid misjudgments caused by a single indicator dominating the analysis. Combined with risk potential propagation, local anomalies are expanded into spatially correlated thermal stress distribution structures. Compared to traditional single-threshold alarm methods, this approach dynamically represents the evolution trend of thermal field imbalance, improving the sensitivity and stability of risk identification and providing a structured risk basis for adaptive reconfiguration of welding sequences.

[0043] Preferably, step S3 specifically includes:

[0044] S31: Decompose the risk situation based on the heat stress risk data to obtain the control domain data;

[0045] S32: Generate the zoned thermal control boundary based on the control domain data to obtain the thermal target zone data;

[0046] S33: Perform target budget allocation on the thermal target zone data to obtain control target data.

[0047] This invention decomposes thermal stress risk data into a risk state, transforming continuously distributed risk potentials into a control domain structure with spatial boundaries, thus transitioning from "risk identification" to "control object location." It constructs partitioned thermal control boundaries, assigning different temperature zones, cooling rates, and gradient constraints to different risk types, forming differentiated control strategies. Through target budget allocation, limited preheating, insulation, and waiting resources are distributed according to risk intensity and spatial correlation, avoiding over-control or resource waste. Compared to traditional uniform threshold adjustment methods, this invention enhances the targeting and executability of control strategies, improving the overall stability and control efficiency of the low-temperature welding process.

[0048] Preferably, the initial welding action is generated as follows:

[0049] S41: Perform target analysis on the control target data to obtain motion constraint data;

[0050] S42: Parameterize the welding action based on the action constraint data to obtain the welding action data;

[0051] S43: Generate symmetrically paired welding action data by mirroring the data;

[0052] S44: Perform thermal constraints based on the symmetric pairing data to obtain thermal constraint data;

[0053] S45: Assemble the action sequence based on thermal constraint data and symmetrical pairing data to obtain initial welding action data.

[0054] This invention achieves a structured mapping from abstract thermal control objectives to executable welding behaviors by parsing and transforming control targets into action constraints. Parametric modeling of welding actions allows for adjustable segment lengths, spacing, and advancement rhythms. Mirror-symmetric pairing generation combined with thermal constraint superposition effectively suppresses the risks of circumferential thermal unevenness and local thermal superposition. The assembly of action sequences forms an initial welding scheme that satisfies spatial symmetry and thermal boundary constraints. Compared to traditional empirical arrangement methods, this invention improves the interpretability of weld sequence generation and the consistency of the thermal field, laying a stable foundation for real-time reconfiguration.

[0055] Preferably, the thermal stress welding sequence reconstruction specifically includes:

[0056] S46: Perform thermal stress activation positioning on the initial welding action data to obtain conflict action data;

[0057] S47: Subsequence splitting of conflict action data to obtain subsequence data;

[0058] S48: Perform heat propagation coupling feedforward deduction based on subsequence data to obtain action partition impact data;

[0059] S49: Reorder the constraints based on the action partition impact data to obtain the welding reconstruction data.

[0060] This invention, by locating the thermal stress activation of the initial welding action, can accurately identify key weld segments that conflict with the current thermal risk structure, avoiding indiscriminate adjustments to the overall sequence; by splitting the sub-sequence, the reconstruction granularity is refined to controllable units, improving scheduling flexibility; by combining feedforward extrapolation with thermal propagation coupling, the impact of each action on the thermal boundaries of adjacent zones is predicted, realizing the transformation from "post-event response" to "pre-event control"; by generating a welding reconstruction sequence through constraint reordering, the risk of thermal tearing and thermal accumulation is suppressed while meeting control objectives, enhancing the stability and foresight of welding sequence adjustment in low-temperature environments.

[0061] Preferably, this application also provides an adaptive optimization system for tower assembly welding sequence, used to execute the adaptive optimization method for tower assembly welding sequence as described above, the adaptive optimization system for tower assembly welding sequence includes:

[0062] Industrial computing nodes, data acquisition interfaces, and industrial communication interfaces;

[0063] The industrial computing node includes a processor and a memory. The memory stores an executable program, and the processor calls the executable program to implement the following functional modules:

[0064] The thermal field feature extraction module is used to acquire cooling status data through the data acquisition interface; and to extract thermal field features based on the cooling status data to obtain thermal field feature data.

[0065] The heat stress risk center mapping processing module is used to obtain the heat field feature data from the heat field feature extraction module, and perform heat stress risk center mapping processing based on the heat field feature data to obtain heat stress risk data.

[0066] The thermal balance control target generation module is used to obtain the thermal stress risk data from the thermal stress risk central mapping processing module, and generate thermal balance control targets based on the thermal stress risk data to obtain control target data;

[0067] The welding sequence generation and reconstruction module is used to obtain the control target data from the thermal balance control target generation module, generate initial welding actions based on the control target data to obtain initial welding action data, and reconstruct the initial welding action data by thermal stress welding sequence to obtain welding reconstruction data.

[0068] The welding reconstruction data is sent to the welding equipment control system via the industrial communication interface to execute the corresponding welding sequence.

[0069] The beneficial effects of this invention are as follows: In S1, cooling conditions and welding process data are coupled to extract circumferential and axial thermal features, establishing a thermal field description basis with spatial structural significance; in S2, steady-state stagnation and transient rapid cooling are separated and modeled through dual-pathway deviation and risk potential propagation, forming a spatially correlated thermal stress risk distribution; in S3, the risk situation is transformed into zoned thermal control boundaries and resource budgets, ensuring the physical executability of control objectives; in S4, through parameterized action generation and heat propagation feedforward deduction, the welding sequence is transformed from empirical arrangement to verifiable scheduling. This invention can not only identify circumferential thermal tear rings and axial thermal accumulation steps, but also complete welding sequence reconstruction before risks form, reducing the probability of cold cracking and residual stress superposition under low-temperature conditions, and improving the stability and consistency of the tower welding process. Attached Figure Description

[0070] Other features, objects, and advantages of this application will become more apparent from the following detailed description of the non-limiting embodiments, taken with reference to the accompanying drawings:

[0071] Figure 1 A flowchart illustrating the steps of an adaptive optimization method for tower assembly welding sequence according to an embodiment is shown.

[0072] Figure 2 A flowchart illustrating the steps of a thermal field feature extraction method according to an embodiment is shown.

[0073] Figure 3 A flowchart illustrating the steps of a heat stress risk center mapping processing method according to an embodiment is shown;

[0074] Figure 4 A flowchart illustrating the steps of a method for generating a thermal balance control target according to an embodiment is shown.

[0075] Figure 5 A flowchart illustrating the steps of an initial welding action generation method according to an embodiment is shown. Detailed Implementation

[0076] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0077] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. Functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0078] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0079] Please see Figures 1 to 5 This application provides an adaptive optimization method for the welding sequence of tower assembly, comprising the following steps:

[0080] Step S1: Obtain cooling status data; extract thermal field features based on the cooling status data to obtain thermal field feature data;

[0081] In one embodiment, temperature sampling points are arranged at 15-degree intervals along the circumferential direction in the vicinity of the tower's circumferential and longitudinal seams, and height zones are divided at 200-millimeter intervals along the height direction. The system synchronously collects real-time temperature values, corresponding time information, and ambient wind speed data for each measuring point. Based on the temperature changes over continuous time, the degree of temperature change per unit time at each measuring point is calculated according to the correspondence between the temperature difference between two adjacent sampling times and the sampling time interval. The temperature differences between adjacent circumferential sectors and between adjacent height zones are statistically analyzed to form circumferential thermal gradient characteristics and axial thermal gradient characteristics, respectively; and the temperature dispersion of each measuring point within the same height zone is statistically analyzed to obtain a thermal non-uniformity index. The cooling rate, circumferential thermal gradient, axial thermal gradient, and thermal non-uniformity are combined to construct thermal field characteristic data.

[0082] Step S2: Perform thermal stress risk center mapping processing based on thermal field characteristic data to obtain thermal stress risk data;

[0083] In one embodiment, the system pre-sets a target temperature range and an allowable cooling rate range as thermal control benchmarks. The system performs steady-state deviation analysis on the thermal field characteristics of each zone, determining the degree of deviation between the current temperature and the target temperature range; simultaneously, it performs transient deviation analysis, determining the deviation of the temperature change rate from the normal baseline over a short period. When the cooling rate exceeds the allowable range, and the circumferential or axial thermal gradient simultaneously exceeds a set limit, the cold crack risk level of that zone is increased. Based on the number and duration of trigger events, the cold crack risk is divided into three levels: low, medium, and high. For example, a low level is defined as only a slight exceedance of the cooling rate limit for a short duration; a medium level is defined as both exceeding the cooling rate and gradient limits; and a high level is defined as the superimposed temperature being significantly lower than the lower limit and existing for multiple consecutive time windows. The level range is formed by calibrating historical construction data and process safety boundaries and is solidified into rule thresholds. When the temperature remains above the upper limit and the cooling rate is below the normal baseline, the overheating risk level is increased, graded based on the magnitude and duration of the temperature exceeding the upper limit, thermal accumulation characteristics, and the degree of slow cooling. A short-term temperature spike is defined as low-level; a sustained high temperature coupled with a persistently low cooling rate is defined as medium-level; and a continuous heat input or axial stacking structure is defined as high-level. The system determines the dominant risk type based on a preset risk competition fusion (the system calculates the intensity of the cold crack channel and the overheat channel in each zone, performs interval normalization comparison of the two channel intensities; if the intensity of one channel is higher than the other and the difference exceeds a set difference threshold, it is determined to be the dominant channel; if the difference does not reach the threshold, it is marked as a mixed state, and both risk levels are retained. The risk levels here include two parallel indicators: cold crack risk level and overheat risk level). Combining the adjacency relationships between zones, a spatial expansion correction is performed on the risk levels (centered on the dominant risk zone, a layer of adjacency traversal is performed along the circumferential adjacent sectors, axially adjacent height bands, and opposite paired zones; adjacent zones are judged to determine if there are signs of deviation with the same trend; if so, their risk level is increased by one level or marked as a transitional diffusion zone; if the data quality of adjacent zones is low, the increase is reduced. This expansion only performs one adjacency level propagation), forming thermal stress risk distribution data.

[0084] Step S3: Generate thermal balance control targets based on thermal stress risk data to obtain control target data;

[0085] In one embodiment, the system divides the tower into zones based on thermal stress risk data, forming high-risk, transition, and stable zones. For example, when the risk potential value of a zone exceeds a preset high-risk threshold and remains at a high level for multiple consecutive time windows, it is defined as a high-risk zone, indicating a significant tendency for cold cracking or overheating instability, requiring priority intervention and control. When the risk potential value is in the middle range, or reaches a high level but for a short duration, it is defined as a transition zone, which has potential risks but has not yet formed a stable instability trend. When the risk potential value is below a set low-risk threshold and there is no obvious trend of continuous amplification, it is defined as a stable zone, where the current thermal state is within a controllable range. For zones where cold cracking risk is dominant, the minimum control temperature requirement for that zone is increased, and the maximum allowable cooling rate per unit time is correspondingly tightened. For zones where overheating risk is dominant, the maximum allowable temperature range is lowered, and the continuous welding length is limited. Control resources are allocated hierarchically based on the available preheating equipment power and acceptable waiting time on-site, for example, prioritizing preheating measures or inserting waiting windows for high-risk zones. The system generates control target data that includes the target temperature range, the upper limit of the cooling rate control, and the limit of the continuous welding length.

[0086] Step S4: Generate initial welding actions based on control target data to obtain initial welding action data; reconstruct the welding sequence using thermal stress on the initial welding action data to obtain welding reconstruction data.

[0087] In one embodiment, the system constrains the continuous circumferential or axial welding segments based on the limitations on temperature range, cooling rate, and continuous welding length in the control objectives. For example, it limits the welding angle of a single circumferential seam or allows only one segment of welding to be completed in a single height zone, forming an initial welding action sequence. The system constructs mirror welding pairs according to the principle of circumferential symmetry to reduce the impact of circumferential thermal unevenness. The generated action sequence undergoes a heat propagation feedforward evaluation to analyze its impact on the temperature difference and cooling state of adjacent zones after execution. This evaluation includes the system first reading the zone location, angle range, height band, expected duration, and heat input level of the action / action sequence to be evaluated, and simultaneously acquiring the circumferential temperature field, axial temperature field, cooling rate, and circumferential / axial gradient peak information within the time window prior to the action's execution as the initial thermal state. The system then determines the affected neighborhood, selecting adjacent circumferential sectors, adjacent axial height bands, and paired zones on the opposite side as the propagation targets, centered on the zone where the action is located. The system also determines the propagation priority direction based on the windward / leeward / shear zone labels and step / ring zone markers. Based on this, the system performs trend extrapolation for each neighboring zone: if the action zone has a high temperature and gradient... If a neighboring region is identified, the system determines that the temperature rise trend and gradient increase trend are both enhanced. If the neighboring region is in a strong wind and cold zone, the system determines that the cooling acceleration trend is enhanced and the risk of temperature difference widening is increased. If the neighboring region has a step front or a marked high-risk core area, the system determines that the risk potential is more sensitive to the increasing trend. The system compares the above trend results with the thermal target zone of the neighboring region. If any situation occurs where the temperature may exceed the limit, the cooling rate may exceed the upper limit, or the gradient may exceed the upper limit, the action is marked as a thermal conflict action, and an adjustment strategy is given according to the trigger type: if the risk of exceeding the limit is too high and cold cracking occurs, preheating or downgrading is inserted and the continuous segment is broken up; if the risk of exceeding the limit is too high and overheating occurs, execution is delayed and waiting or the segment length is shortened. The adjusted sequence then enters the next round of the same feedforward evaluation until the conflict mark is eliminated or the preset verification round is reached. If the prediction result / evaluation result obtained from the heat propagation feedforward evaluation process causes a local zone to exceed the control target requirements, the action is delayed or waiting, preheating, or other adjustment actions are inserted. Through local reordering and multiple rounds of verification, welding reconstruction data that meets the thermal constraint conditions is output.

[0088] Preferably, step S1 specifically includes:

[0089] Step S11: Obtain cooling status data;

[0090] In one embodiment, temperature acquisition points are deployed near the circumferential or longitudinal seam of the tower to be welded, and a partition index is established according to the combination of circumferential sectors and axial height bands. Each partition is numbered and managed. For example, in the circumferential direction, the circumference is divided into 24 sectors at 15 degrees, numbered S01 to S24 respectively; in the axial direction, each height band is divided at 200 millimeters. Assuming the weld seam's coverage height is two meters, it can be divided into ten height bands, numbered H01 to H10. The system establishes the partition index using circumferential numbering + axial numbering. For example, S03-H05 represents the partition formed by the intersection of the third circumferential sector and the fifth height band. The system collects the surface temperature time series and corresponding time information of each partition, and simultaneously acquires environmental disturbance parameters, including at least wind speed and ambient temperature. Windward angle and shading status can be collected as needed. For outliers in the collected data, the system performs amplitude limiting and median filtering. When a single point temperature experiences a sudden change exceeding a set range in adjacent time intervals, it is replaced by the linear trend of the previous and subsequent frames of data, and a data quality flag is generated simultaneously. After processing, structured cooling status data is formed.

[0091] Step S12: Calculate the cooling rate and thermal non-uniformity based on the cooling situation data, and obtain the cooling rate data and thermal non-uniformity data respectively.

[0092] In one embodiment, a sliding time window approach is used to calculate the cooling rate for the temperature time series of each partition. Within each window, the trend formed by the time interval corresponding to the temperature difference between the start and end times of the window is used as the basic estimate, and the median value of multiple estimates within the window is taken. Thermal non-uniformity is characterized by the degree of temperature dispersion of each circumferential sector within the same height band, and the maximum temperature difference and overall fluctuation are statistically analyzed. The system outputs cooling rate data and thermal non-uniformity data, and retains the corresponding time window index information.

[0093] Step S13: Obtain welding process data, and perform thermal characteristic calculations based on cooling status data and welding process data to obtain thermal characteristic data. The thermal characteristic calculations include circumferential thermal calculations and axial thermal calculations, and the thermal characteristic data includes circumferential thermal characteristic data and axial thermal characteristic data.

[0094] In one embodiment, the system acquires welding process data, including the current height zone number and circumferential angle of the welding torch, the start and end positions of the welding segment, whether the advancement is continuous, and the heat input level indicator formed by the discrete distribution of current, voltage, and travel speed. In circumferential thermal calculation, the temperature distribution of circumferential sectors is constructed. Combined with the ambient wind speed (obtained in real time by wind speed sensors deployed on-site, or data from the construction site meteorological monitoring system), the windward, leeward, and shear zones of each zone are determined. (The system acquires the tower azimuth angle and real-time wind direction data, and calculates the relative relationship between the wind direction and the center angle of each circumferential sector. When the wind direction is directly opposite to or close to the normal direction of a sector, it is determined to be a windward zone; when the wind direction is opposite to the sector direction or is in a leeward area; when there is a significant difference in wind cooling intensity between adjacent sectors, and the wind direction changes significantly at their boundary, the boundary area is determined to be a shear zone.) The peak areas of circumferential temperature changes are located. Combined with the heat input status and abnormal cooling rate of the opposite zone, the characteristic segments forming thermal tear rings are identified, and circumferential thermal characteristic data are output. In axial thermal calculation, the temperature distribution of axial height zones is constructed (the representative temperatures of each height zone within the same time window are arranged in order of height to form an axial temperature sequence). The temperature change trend between adjacent height zones is analyzed (the temperature difference and direction of change between adjacent height zones are compared to determine the heating or cooling trend, and the cooling rate is combined to determine whether there is an abnormal steep change). Temperature step phenomenon is identified (a continuous structure with stagnation at the bottom and rapid cooling at the top is identified as a temperature step). Combined with residual heat accumulation, the effect of gust cooling, and the alignment of the welding torch (if there are residual heat accumulation characteristics at the bottom, gust cooling characteristics at the top, and the welding torch advancement direction forms a leading edge relationship with the step position, then the axial heat accumulation characteristic is confirmed to be formed), axial heat accumulation characteristics are formed, and axial thermal characteristic data are output.

[0095] Step S14: Integrate the cooling rate data, thermal characteristic data, and thermal non-uniformity data to obtain thermal field characteristic data.

[0096] In one embodiment, the system aligns cooling rate data, circumferential thermal characteristic data, axial thermal characteristic data, and thermal non-uniformity data according to a unified time window and partition index, constructing a feature set for each partition. The integrated output thermal field characteristic data includes cooling rate indices for each partition, circumferential and axial gradient peak characteristics (including peak position, intensity, and distribution range), thermal tear ring markers, temperature step markers and their intensity (a high-intensity step is defined as one with a significant gradient peak, a long duration of heat retention, and a significant quenching amplitude; a medium-intensity step is defined as one with the above characteristics at a moderate level; and a low-intensity step is defined as one that only meets the minimum determination criteria; numerical labels can also be used to mark high, medium, and low), and overall thermal non-uniformity indices.

[0097] Preferably, the circumferential thermal calculation specifically involves:

[0098] S131: Construct the circumferential sector temperature field based on the cooling status data to obtain circumferential temperature field data;

[0099] In one embodiment, the circumferential weld neighborhood is divided into several sectors along the circumferential direction, for example, partitioned at fixed angular intervals. Each sector corresponds to one or more temperature sampling points. When multiple sampling points exist within the same sector, they are weighted according to their relative positions to the weld center (the system records the distance category of each sampling point relative to the weld centerline during point placement, for example, classifying sampling points into three categories: weld center neighborhood, heat-affected zone neighborhood, and far-field zone; where the weld center neighborhood is assigned the highest weight, the heat-affected zone is assigned a medium weight, and the far-field zone is assigned a low weight). Sampling points closer to the weld center are assigned a higher weight, forming the representative temperature value of that sector. The temperature time series of each sampling point is first subjected to amplitude limiting and median filtering; resampling is performed at uniform time intervals to keep the data of each sector synchronized in the time dimension. The system outputs circumferential temperature field data containing sector number, corresponding time, and temperature value.

[0100] S132: Perform air-cooled coupling partitioning on the circumferential temperature field data to obtain air-cooled partitioning data;

[0101] In one embodiment, a wind-cooling intensity index is constructed for each circumferential sector. This index is determined by the real-time wind speed, the relative angle between the tower's azimuth and the wind direction (corresponding to the degree of windwardness), and the exposure status corresponding to the enclosure or insulation coverage. The system synchronously analyzes the wind-cooling intensity index with the temperature change trend of the sector within a short time window: when the wind-cooling intensity and cooling rate are both high, it is identified as a strong windward convection zone; when both the wind-cooling intensity and cooling rate are at a low level, it is identified as a leeward stagnant zone; when there is a significant change in wind-cooling intensity and a significant difference in cooling rate between adjacent sectors, their boundary is marked as a transition shear zone. The system outputs wind-cooling partition data containing partition labels for each sector and information on the corresponding boundary sectors.

[0102] S133: Perform circumferential thermal gradient localization on the air-cooled partition data to obtain circumferential thermal gradient data;

[0103] In one embodiment, based on the completed air-cooled zoning determination, the temperature difference between adjacent circumferential sectors is compared with its corresponding arc length to form a circumferential temperature change intensity sequence. This change sequence is then smoothed. The system statistically analyzes the gradient peak value and its duration at each location, using the number of sectors continuously exceeding a set standard as the peak width index. During system positioning, gradient peaks are preferentially searched within the transition shear zone or its adjacent boundary; when multiple gradient peaks exist, the peak with the larger temperature change amplitude and wider duration is selected as the dominant peak. The system outputs circumferential thermal gradient data containing the sector location of the gradient peak, peak intensity, peak width range, and corresponding time window information.

[0104] S134: Based on the welding process data, extract the heat input from the opposite side and detect the sudden change in cooling rate to obtain the heat input data from the opposite side and the sudden change in cooling rate data, respectively.

[0105] In one embodiment, the system determines the opposite sector, which is approximately 180 degrees different in angle from the current welding sector and within a one-sector deviation range, based on the current welding sector location and historical welding trajectories. If the opposite sector has been continuously welding or has a high heat input level in recent consecutive time windows, and its temperature decrease trend is significantly slower than that of surrounding sectors, then it is determined that there is an influence from the opposite side heat input, and the system outputs opposite side heat input data including duration, heat input level, and sector location. For cooling rate mutation detection, in sectors near the identified circumferential gradient peak, a short-time change amplitude analysis is performed on the cooling rate sequence; when the change amplitude exceeds a set range and is consistent with the upward trend of air cooling intensity, cooling rate mutation data including mutation amplitude, occurrence time, and location is output.

[0106] S135: Based on the circumferential thermal gradient data, the thermal input data on the opposite side and the abrupt change data of the cooling rate are used to determine the formation of thermal tearing rings, and circumferential thermal characteristic data are obtained.

[0107] In one embodiment, the located circumferential thermal gradient peak is used as the basis for judgment, and evidence of heat input on the opposite side and evidence of abrupt changes in cooling rate are jointly analyzed. The banding determination adopts a method that simultaneously satisfies multiple conditions combined with continuity constraints: when the gradient peak intensity reaches a set standard and its duration is not less than the minimum bandwidth requirement, and the sector where the peak is located experiences an abnormal change in cooling rate, and the opposite sector has a continuous heat input state within a recent time window, the conditions for thermal tearing banding are determined. These conditions must remain met within multiple consecutive time windows to confirm banding, thereby reducing misjudgments caused by instantaneous environmental disturbances. The system outputs circumferential thermal characteristic data, including the central sector of the banding, bandwidth range, intensity level, opposite heat input location, and corresponding trigger identifier.

[0108] Preferably, the axial thermal calculation is as follows:

[0109] S136: Construct the axial height zone temperature field based on the cooling status data to obtain axial temperature field data;

[0110] In one embodiment, the vicinity of the tower weld is divided into several height zones along the height direction. These zones can be divided at fixed intervals or based on plate thickness ratios. Each height zone corresponds to a set of temperature sampling points within its range. At a uniform time interval, the system weights the temperatures of the sampling points within the same height zone, assigning higher weights to sampling points closer to the weld center and heat-affected zone, thus forming the representative temperature value for that height zone. For missing or abnormal sampling points, interpolation is performed using the temperature change trends of adjacent height zones to complete the data. The system outputs axial temperature field data containing the height zone number, corresponding time, and temperature value, while retaining information on the effective sampling quantity for each height zone.

[0111] S137: Perform axial thermal gradient calculation on the axial temperature field data to obtain axial thermal gradient data;

[0112] In one embodiment, the system calculates the intensity of axial temperature change based on the temperature difference between adjacent height zones and their corresponding height intervals, forming an axial gradient sequence. The system smooths the gradient sequence using a sliding time window and uses the median value within the window as the representative gradient. The system statistically analyzes the gradient peak values ​​and their durations at the boundaries of each height zone, using the number of boundaries continuously exceeding a set standard as the peak width index. The system outputs axial thermal gradient data containing gradient peak magnitude, peak width range, corresponding boundary positions, and occurrence time window information.

[0113] S138: Perform temperature step discrimination on the axial thermal gradient data to obtain temperature step data;

[0114] In one embodiment, analysis windows are constructed upwards and downwards, centered on the boundary where the identified axial thermal gradient peak is located: the lower window covers several height bands below the boundary, and the upper window covers several height bands above the boundary. The system analyzes the temperature change trend within the lower window. When the temperature decays slowly, plateaus, and the cooling rate remains below the normal baseline for an extended period, heat stagnation is identified. Simultaneously, the upper window is monitored. When a significant increase in the cooling rate and a rapid decrease in temperature occur within a short period, rapid cooling is identified. A temperature step is confirmed only when the aforementioned heat stagnation and rapid cooling characteristics occur simultaneously at the same boundary and are repeatedly satisfied within multiple consecutive time windows. The system outputs temperature step data including the step's location, direction, intensity, and duration.

[0115] S139: Extract residual heat accumulation and gust cooling data based on axial temperature field data to obtain residual heat accumulation data and gust cooling data respectively.

[0116] In one embodiment, for residual heat accumulation extraction, the system analyzes the temperature drop process within the area below the identified temperature step or the height zone where welding has been completed. When the time required for temperature drop is significantly prolonged and the cooling rate remains at a low level for a longer period, residual heat accumulation is identified, and accumulation intensity data including the duration of the platform and the magnitude of high temperature maintenance is output. For gust cooling extraction, the system monitors abnormal increases in the cooling rate within a short time window above the step or within the height zone to be welded, and compares this with the trend of wind speed changes. When an increase in wind speed and a significant increase in cooling rate occur simultaneously, it is identified as a gust cooling event, and gust cooling data including the magnitude of the change, the degree of synchronization, the time of occurrence, and the location are output.

[0117] S1310: Based on welding process data and temperature step data, advance alignment is performed to obtain advance alignment data;

[0118] In one embodiment, the system extracts the height advancement sequence from the welding process data, records the height band number and time information corresponding to each welding operation, analyzes whether the advancement direction shows a continuous upward or downward trend, and counts the number of height bands crossed per unit time and the direction switching interval to characterize the advancement speed and rhythm. The system performs spatiotemporal matching between the advancement sequence and temperature step data. When the temperature step is located near the leading edge height band of the current advancement direction, and the step appears within the time interval from the completion of the lower side welding to the entry into the upper side welding, it is determined that there is an alignment relationship between the advancement and the step. The system outputs advancement alignment data including the advancement direction, advancement speed, alignment deviation, and alignment establishment indicator.

[0119] S1311: Axial thermal characteristic data are obtained by quantifying axial thermal accumulation based on residual thermal accumulation data, gust cooling data, and propulsion alignment data.

[0120] In one embodiment, the system combines residual heat accumulation intensity, gust cooling amplitude, and advancement alignment relationship to construct an axial heat accumulation risk level, which characterizes the risk level of the lower heat retention and upper rapid cooling structure. A graded evaluation method is used for quantification: when the accumulation duration is long and the high temperature is maintained for a large range, while the rapid cooling abrupt change is obvious, and the advancement direction aligns with the temperature step and the advancement speed is fast, it is judged as a higher risk level; if the advancement alignment relationship is not established, the risk level is reduced accordingly to avoid misjudging natural environmental cooling as being caused by welding advancement factors. The system outputs axial thermal characteristic data, including step risk level, step location, accumulation area range, rapid cooling area range, and corresponding trigger markers.

[0121] Preferably, the temperature step discrimination is specifically as follows:

[0122] a. Extract the axial gradient peak segments from the axial thermal gradient data to obtain gradient peak segment data;

[0123] In one embodiment, the system scans the axial thermal gradient data frame by frame along the height zone boundary in chronological order, extracting peak segments using a combination of dual thresholds and continuity. When the gradient intensity of a certain boundary exceeds a preset high standard, that location is designated as the starting point of the peak segment. In subsequent scans, as long as the gradient of adjacent boundaries remains higher than the lower standard, the range of the peak segment continues to expand until several consecutive boundaries fall below the lower standard, at which point the peak segment ends. For each peak segment, the boundary where the maximum gradient is located, the number of boundaries it covers, the corresponding center height position, and the time window in which it appears are recorded. When the width of a peak segment is less than the minimum requirement or appears only in a single time window, it is marked as a suspected peak segment. The system outputs gradient peak segment data containing the above information.

[0124] b. Perform upper-side rapid cooling feature discrimination and lower-side stagnation feature discrimination on the gradient peak segment data to obtain upper-side rapid cooling feature data and lower-side stagnation feature data respectively;

[0125] In one embodiment, analysis windows are constructed upwards and downwards, using the center boundary of the axial gradient peak segment as the dividing line. The upper window covers several height bands above the peak segment, and the lower window covers several height bands below the peak segment. The window range can be set according to the plate thickness or the width of the heat-affected zone. Rapid cooling is detected in the upper window. When an increase in cooling rate and a rapid decrease in temperature occur within a short time range, and this change reaches a set standard and persists for a certain period, the upper rapid cooling characteristic data is output. Heating is detected in the lower window. When the temperature decay trend plateaus, the cooling rate remains in a low range for a relatively long time, and the average temperature remains at a high level, the lower heating characteristic data is output, and the plateau duration and heating intensity are recorded.

[0126] c. Based on the upper rapid cooling characteristic data and the lower heat stagnation characteristic data, the axial step structure is inspected to obtain temperature step data.

[0127] In one embodiment, the system performs a structural consistency check on the upper rapid cooling characteristic and the lower stagnation characteristic. The system performs a directional consistency check, confirming that the average temperature of the lower window is significantly higher than that of the upper window, and the corresponding gradient change direction shows a higher temperature at the bottom and a lower temperature at the top. The system performs a spatial consistency check, confirming that the main influence height bands of rapid cooling and stagnation are located within the upper and lower analysis windows of the peak segment, respectively, and the gradient peak segment is located in the boundary region between the two. The system performs a temporal persistence check, confirming that the above relationship exists stably within multiple consecutive time windows. Only when all three consistency conditions are simultaneously met is a temperature step confirmed to have formed, and the system outputs temperature step data including step location, intensity level (a high-intensity step is determined when the rapid cooling change is large, the stagnation duration is long, and the gradient peak is obvious, or at least two of these conditions are met; a medium-intensity step is determined when the above indicators are at a medium level or at least two of the conditions are at a medium level; a low-intensity step is determined when only the minimum determination condition is met or at least two of the conditions are at the minimum determination condition), duration, and triggering cause identifier; otherwise, it is reverted to general gradient peak processing.

[0128] Preferably, step S2 specifically includes:

[0129] Step S21: Encode the thermal field input based on the thermal field characteristic data to obtain the thermal field input data;

[0130] In one embodiment, the system organizes thermal field feature data according to a partition (combination of circumferential sectors and height bands) – time window structure to construct the input data structure. Continuous features such as temperature, cooling rate, circumferential and axial gradients, thermal non-uniformity, and step or circumferential band intensity are uniformly scaled to ensure they are within a comparable range. For partitions with low confidence data quality indicators, missing measurement mask markers are added to indicate subsequent processing steps. The system calculates statistical features over a short time range and trend features over a longer time range, and then concatenates the two types of statistical results to form the thermal field input data.

[0131] Step S22: Obtain thermal setpoint data, and perform dual-path deviation calculation based on thermal field input data and thermal setpoint data to obtain dual-path deviation data, wherein the dual-path deviation calculation includes steady-state deviation calculation and transient deviation calculation.

[0132] In one embodiment, the system acquires thermal setpoint data, including the target temperature range for each zone, the upper limit of the allowable cooling rate, the upper limit of the gradient change, and the upper limit of the overheating duration. Steady-state deviation calculation measures the degree of deviation of the current state from the control target range: when the temperature is within the target range, it is considered to have no deviation; when it is below the lower limit or above the upper limit, the deviation is recorded according to the degree of exceeding the range; the cooling rate and gradient change are measured based on the magnitude of exceeding the allowable upper limit. Transient deviation calculation represents short-term abrupt changes. By comparing short-term statistical results with long-term baseline trends, a rapid cooling deviation is identified when the short-term cooling rate is significantly higher than the long-term level, and a tearing deviation is identified when the short-term gradient peak rapidly increases. The system outputs dual-path deviation data containing a steady-state deviation vector, a transient deviation vector, and the corresponding time window index.

[0133] Step S23: Competitively fuse the dual-path bias data to obtain competitive situation data;

[0134] In one embodiment, the system maps steady-state and transient deviations to two types of risk channels: a cold crack risk channel and an overheating or deformation risk channel. The cold crack channel mainly gathers deviation information such as temperatures below the lower limit, excessive cooling rates, excessive gradient changes, and abnormal temperature steps or ring strength. The overheating channel mainly gathers deviation information such as temperatures above the upper limit, excessively long residual heat accumulation duration, excessively slow cooling, and excessively high continuous heat input. The system uses competitive fusion for judgment. When the overall evidence strength of the cold crack channel is significantly higher than that of the overheating channel and the difference reaches a set standard, it is determined to be cold crack dominant; otherwise, it is determined to be overheat dominant. If the difference between the two is small, it is determined to be a mixed state, and the strength values ​​of both channels are retained. The system outputs competitive situation data including the dominant type, the strength of the two channels, and a mixed indicator.

[0135] Step S24: Propagate the heat stress risk potential based on the competitive situation data to obtain heat stress risk data.

[0136] In one embodiment, the system constructs a risk propagation topology based on partition adjacency relationships, including connections between circumferentially adjacent sectors, connections between axially adjacent height bands, and circumferential opposite-side pairings to represent symmetrical thermal coupling. An initial risk potential value is assigned to each partition based on competitive situation data (using the strengths of the two risk channels output during the competitive fusion phase as the base quantity, and mapping them in conjunction with the dominant type identifier: if it is a single dominant type, the corresponding channel strength is used as the initial risk potential base value for that partition; if it is a mixed state, the initial risk potential value is generated based on the weighted result / weighted average of the two channel strengths). The propagation direction is selected according to the dominant risk type: when cold cracking is dominant, it preferentially expands along the windward side region and the gradient peak neighborhood; when overheating is dominant, it preferentially expands along the welded stagnation zone and the continuous advancement direction; when it is a mixed state, it combines the two types of expansion paths for propagation. During propagation, the diffusion range is limited by distance attenuation, and the propagation weight is reduced for partitions with lower data quality. The system outputs thermal stress risk data containing the risk potential value, dominant type, and propagation path summary / propagation chain for each partition.

[0137] Preferably, step S3 specifically includes:

[0138] Step S31: Decompose the risk situation based on the heat stress risk data to obtain the control domain data;

[0139] In one embodiment, the system aggregates heat stress risk data by partition into data containing risk potential, dominant type, and propagation chain information. The system classifies risk potential into risk zones: zones exceeding a preset high range are marked as high-risk core zones, those in the middle range are marked as transition zones, and the rest are marked as stable zones. Combining propagation chain relationships, the system selects adjacent partitions as buffer control domains around each high-risk core zone to form a hierarchical control structure, avoiding risk transfer to the boundary due to controlling only a single partition. For partitions with a mixed dominant type, both cold crack weight and overheating weight dual-label information are retained. The system outputs control domain data containing the partitioning results of core domains, buffer domains, and stable domains, along with corresponding type labels.

[0140] Step S32: Generate the partitioned thermal control boundary based on the control domain data to obtain the thermal target zone data;

[0141] In one embodiment, based on the control domain partitioning results, the system generates a set of thermal control parameters for each partition, including the target temperature range, the upper limit of the allowable cooling rate, and the upper limit of the allowable gradient. For core domain partitions dominated by cold cracking, the minimum control temperature requirement is increased, the upper limit of the cooling rate is tightened, and the allowable gradient variation range is correspondingly reduced. For core domain partitions dominated by overheating, the maximum temperature range is reduced, the duration of heat accumulation is limited, and the continuous welding length is constrained. Buffer domain partitions use a relatively relaxed temperature range but retain gradient control requirements to prevent the risk from spreading outward. For mixed-state partitions, two types of control boundaries, namely anti-cold cracking and anti-overheating, are generated simultaneously, and their priorities are marked according to risk weights. The system outputs thermal target band data containing the above control parameters and type labels.

[0142] Step S33: Perform target budget allocation on the thermal target zone data to obtain control target data.

[0143] In one embodiment, the system converts thermal target zone data into an executable control resource budget. This budget includes at least a preheating budget, a heat preservation budget, a waiting budget, and a scheduling budget. The conversion rules are derived from welding heat conduction mechanism analysis, construction process specifications, and historical operation data statistics. These rules are parameterized and solidified into a rule base, and calibrated through historical risk backtracking. Specifically, the preheating budget corresponds to the additional preheating time or power level; the heat preservation budget corresponds to the coverage area or heat preservation duration; the waiting budget corresponds to the allowed recovery waiting time; and the scheduling budget corresponds to the allowed number of skip welding switches or the maximum continuous welding segment length. During allocation, high-risk core areas are prioritized, and the central nodes of the propagation chain are prioritized over edge nodes. Preheating and heat preservation resources are prioritized for cold-cracking-dominant zones, while waiting and segment length restriction resources are prioritized for overheating-dominant zones. When on-site resources are limited, the system appropriately relaxes the target temperature range and lowers the budget allocation for low-risk zones. The system outputs control target data including control boundary parameters, resource budgets, priorities, and rollback flags.

[0144] Preferably, the initial welding action is generated as follows:

[0145] Step S41: Perform target analysis on the control target data to obtain motion constraint data;

[0146] In one embodiment, the system structurally decomposes the control target data, converting the thermal target band parameters and resource budget fields corresponding to each partition into executable welding action constraints. The rule base corresponding to this conversion is derived from welding process standards, historical construction data statistics, and risk feedback calibration results, and is formalized and solidified into conditional triggering rules. The constraints include at least the maximum continuous welding angle, the maximum allowable advance length per height band, the allowed waiting time window, whether preheating measures are mandatory, and gradient upper limit triggering conditions. During the parsing process, differentiated constraints are generated based on the partition's dominant type: for core partitions dominated by cold cracking, constraints are generated requiring preheating and prohibiting continuous welding beyond a set angle range; for core partitions dominated by overheating, constraints are generated limiting the length of continuous segments and forcibly inserting waiting intervals; for mixed-state partitions, both types of constraints are retained and priority identifiers are added. The system outputs action constraint data containing partition identifiers, constraint types, and corresponding parameter values.

[0147] Step S42: Parameterize the welding action based on the action constraint data to obtain the welding action data;

[0148] In one embodiment, the system discretizes the welding operation into several parameterized action units based on action constraint data. Each action unit includes at least the following fields: starting sector angle, ending sector angle, corresponding height zone number, expected duration, and heat input level. When generating an action, the continuous welding angle must not exceed the maximum continuous range specified by the constraints; when the height zone is in an axial high-risk area, the system automatically divides the original advance length into multiple segments to reduce the risk of heat accumulation. The heat input level is dynamically matched according to the target temperature range of the zone. When the current temperature is low, the heat input level is appropriately increased; when in a high-temperature risk zone, the level is limited. The heat input level is obtained by discretizing and classifying welding process parameters such as welding current, voltage, and welding travel speed. The system outputs a list of welding action data containing action parameters and the zone identifier.

[0149] Step S43: Perform mirror-symmetric pairing on the welding action data to generate symmetrical pairing data;

[0150] In one embodiment, the system performs mirror matching based on the circumferential angle range of the welding action. For any welding action, its corresponding symmetrical region is a sector that differs from the angle by approximately 180 degrees and is within a certain tolerance range. The system searches for matching sectors within this range, generates corresponding symmetrical actions, and establishes action pairing relationships. When the opposite side is a high-risk area for cold cracking, symmetrical actions are preferentially generated and scheduled for execution within adjacent time windows; when the opposite side is an overheating risk area, pairing execution can be appropriately delayed, but must be completed within a preset time window. The system outputs symmetrical pairing data containing action pair numbers, corresponding angle intervals, and time window constraint information.

[0151] Step S44: Perform thermal constraints based on the symmetric pairing data to obtain thermal constraint data;

[0152] In one embodiment, the system predicts the heat propagation trend of a proposed symmetrical pairing action based on the circumferential and axial thermal gradient characteristics within the most recent time window, and analyzes the impact of the action on the temperature and cooling status of adjacent zones. When the prediction result indicates that a zone may exceed the target temperature range or the upper limit of the cooling rate, a constraint flag is added to the corresponding action, such as requiring the insertion of a waiting time, reducing the heat input level, or adjusting the execution order. The system combines the axial temperature step flag for verification, and restricts continuous upward or downward advancement behavior when there is a step risk in the current height zone. The system outputs thermal constraint data containing the action identifier, the added constraint content, and the adjustment type.

[0153] Step S45: Assemble the action sequence based on the thermal constraint data and symmetrical pairing data to obtain the initial welding action data.

[0154] In one embodiment, the system constructs a welding action sequence while satisfying symmetrical pairing relationships and various thermal constraints. During assembly, symmetrical actions corresponding to high-risk zones are prioritized, followed by buffer zone actions. Within the same time window, adjacent high-risk zones are prohibited from welding simultaneously to avoid heat accumulation. When an action carries constraints such as requiring waiting or preheating, a corresponding waiting or preheating preemptive action is inserted into the sequence. The overall assembly adopts a strategy combining local priority and global balance; that is, after completing the pairing of a circumferential zone, the system switches to execute adjacent axial zones to avoid continuous advancement along a single direction, forming a heat accumulation path. The system outputs initial welding action data containing an ordered action list, time window allocation, and constraint satisfaction flags.

[0155] Preferably, the thermal stress welding sequence reconstruction specifically includes:

[0156] S46: Perform thermal stress activation positioning on the initial welding action data to obtain conflict action data;

[0157] In one embodiment, the system scans the initial welding action data sequentially over time and compares the corresponding partition for each action with the current thermal stress risk data and the target thermal zone. If the risk potential value of a partition containing an action exceeds a set high-risk range within the corresponding time window, or if, based on heat propagation predictions, its execution causes the cooling rate, temperature gradient, or target temperature range to exceed the control range, the action is marked as a conflict action. Simultaneously, if the action involves the leading edge height zone of an axial temperature step and the direction of advancement aligns with the step's direction, its conflict level is increased. The system outputs conflict action data including action identifier, conflict type, intensity level, and triggering basis description.

[0158] S47: Subsequence splitting of conflict action data to obtain subsequence data;

[0159] In one embodiment, the system segments the sequence according to the position of conflicting actions within the original action sequence, processing them chronologically. Using the conflicting action as a dividing point, normal actions before and after it are divided into independent subsequences. When multiple conflicting actions occur consecutively in time, they are merged into the same high-risk subsequence block for centralized processing. For each subsequence, an attribute label is assigned based on the corresponding partition risk level and propagation information, such as high-risk, buffered, or stable, and the symmetrical pairing relationships and axial advancement status within the subsequence are recorded simultaneously. The system generates subsequence data for reconstruction optimization of local conflict areas without requiring a complete rearrangement of the entire welding sequence.

[0160] S48: Perform heat propagation coupling feedforward deduction based on subsequence data to obtain action partition impact data;

[0161] In one embodiment, the system performs feedforward heat propagation simulation for each subsequence according to the original time sequence. For each action in the subsequence, the system estimates its impact on the direction of temperature change and the trend of cooling rate change in adjacent partitions after execution, based on the current circumferential and axial thermal gradient states. For example, for each action in the subsequence, the system reads the partition location, heat input level, and expected duration corresponding to the action, and obtains the current circumferential and axial thermal gradient distribution. A local influence range is constructed, with the partition where the action is located as the center, extending a neighborhood layer to the circumferential adjacent sectors, the axial adjacent height zone, and the opposite paired partition. The system judges the heat propagation trend based on the current gradient direction: if the gradient points to the neighboring area, the system predicts that the temperature rise trend in the neighboring area will be enhanced; if the neighboring area is in a strong wind cold zone, the temperature rise trend will be attenuated and corrected. At the same time, based on the current cooling rate level, the system judges whether the cooling rate may slow down or intensify after the action is executed. The above judgment forms a qualitative prediction result in the form of an increasing trend, a decreasing trend, or a stable trend. When there are multiple consecutive high-risk actions in the subsequence, the thermal impact they produce in the same partition is cumulatively analyzed to reflect the superposition effect. The simulation results are summarized by region, forming region-specific impact information that includes temperature shift trends, gradient change trends, and risk potential increase / decrease trends. The system outputs action region impact data.

[0162] S49: Reorder the constraints based on the action partition impact data to obtain the welding reconstruction data.

[0163] In one embodiment, the system locally reorders conflicting subsequences based on action partition impact data. During reordering, subsequences that mitigate risk are prioritized, while subsequences that increase risk are delayed or preceded by waiting or preheating adjustments. Simultaneously, two subsequences that have a cumulative effect on the same high-risk partition are prohibited from being executed within the same time window. When an axial temperature step risk is detected, a skip-layer or regression action is forcibly inserted to interrupt the continuous thermal accumulation path. The sorting employs a block-by-block replacement approach, adjusting the order only within conflicting subsequences, while other stable subsequences remain unchanged. The system outputs welding reconstruction data containing the new action sequence, inserted action information, and constraint satisfaction flags, along with a reconstruction reason identifier for traceability.

[0164] Preferably, this application also provides an adaptive optimization system for tower assembly welding sequence, used to execute the adaptive optimization method for tower assembly welding sequence as described above, the adaptive optimization system for tower assembly welding sequence includes:

[0165] Industrial computing nodes, data acquisition interfaces, and industrial communication interfaces;

[0166] The industrial computing node includes a processor and a memory. The memory stores an executable program, and the processor calls the executable program to implement the following functional modules:

[0167] The thermal field feature extraction module is used to acquire cooling status data through the data acquisition interface; and to extract thermal field features based on the cooling status data to obtain thermal field feature data.

[0168] The heat stress risk center mapping processing module is used to obtain the heat field feature data from the heat field feature extraction module, and perform heat stress risk center mapping processing based on the heat field feature data to obtain heat stress risk data.

[0169] The thermal balance control target generation module is used to obtain the thermal stress risk data from the thermal stress risk central mapping processing module, and generate thermal balance control targets based on the thermal stress risk data to obtain control target data;

[0170] The welding sequence generation and reconstruction module is used to obtain the control target data from the thermal balance control target generation module, generate initial welding actions based on the control target data to obtain initial welding action data, and reconstruct the initial welding action data by thermal stress welding sequence to obtain welding reconstruction data.

[0171] The welding reconstruction data is sent to the welding equipment control system via the industrial communication interface to execute the corresponding welding sequence.

Claims

1. An adaptive optimization method for the welding sequence of tower sections, characterized in that, Includes the following steps: Step S1: Obtain cooling status data; extract thermal field features based on the cooling status data to obtain thermal field feature data; Step S2: Perform thermal stress risk center mapping processing based on thermal field characteristic data to obtain thermal stress risk data; Step S3: Generate thermal balance control targets based on thermal stress risk data to obtain control target data; Step S4: Generate initial welding actions based on control target data to obtain initial welding action data; reconstruct the welding sequence using thermal stress on the initial welding action data to obtain welding reconstruction data; Step S2 is as follows: S21: Perform thermal field input encoding based on thermal field characteristic data to obtain thermal field input data; S22: Obtain thermal setpoint data, and perform dual-path deviation calculation based on thermal field input data and thermal setpoint data to obtain dual-path deviation data, wherein the dual-path deviation calculation includes steady-state deviation calculation and transient deviation calculation. S23: Competitively fuse the dual-path bias data to obtain competitive situation data; S24: Based on the competitive situation data, the heat stress risk potential is propagated to obtain heat stress risk data.

2. The adaptive optimization method for tower assembly welding sequence according to claim 1, characterized in that, Step S1 is as follows: S11: Obtain cooling status data; S12: Calculate the cooling rate and thermal non-uniformity based on the cooling status data, and obtain the cooling rate data and thermal non-uniformity data respectively. S13: Acquire welding process data, and perform thermal characteristic calculations based on cooling status data and welding process data to obtain thermal characteristic data, among which... Thermal characteristic calculations include circumferential thermal calculations and axial thermal calculations, and thermal characteristic data includes circumferential thermal characteristic data and axial thermal characteristic data; S14: Integrate cooling rate data, thermal characteristic data, and thermal non-uniformity data to obtain thermal field characteristic data.

3. The adaptive optimization method for tower assembly welding sequence according to claim 2, characterized in that, The circumferential thermal calculation is as follows: S131: Construct the circumferential sector temperature field based on the cooling status data to obtain circumferential temperature field data; S132: Perform air-cooled coupling partitioning on the circumferential temperature field data to obtain air-cooled partitioning data; S133: Perform circumferential thermal gradient localization on the air-cooled partition data to obtain circumferential thermal gradient data; S134: Based on the welding process data, extract the heat input from the opposite side and detect the sudden change in cooling rate to obtain the heat input data from the opposite side and the sudden change in cooling rate data, respectively. S135: Based on the circumferential thermal gradient data, the thermal input data on the opposite side and the abrupt change data of the cooling rate are used to determine the formation of thermal tearing rings, and circumferential thermal characteristic data are obtained.

4. The adaptive optimization method for tower assembly welding sequence according to claim 2, characterized in that, The axial thermal calculation is as follows: S136: Construct the axial height zone temperature field based on the cooling status data to obtain axial temperature field data; S137: Perform axial thermal gradient calculation on the axial temperature field data to obtain axial thermal gradient data; S138: Perform temperature step discrimination on the axial thermal gradient data to obtain temperature step data; S139: Extract residual heat accumulation and gust cooling data based on axial temperature field data to obtain residual heat accumulation data and gust cooling data respectively. S1310: Based on welding process data and temperature step data, advance alignment is performed to obtain advance alignment data; S1311: Axial thermal characteristic data are obtained by quantifying axial thermal accumulation based on residual thermal accumulation data, gust cooling data, and propulsion alignment data.

5. The adaptive optimization method for tower assembly welding sequence according to claim 4, characterized in that, The temperature step determination is specifically as follows: a. Extract the axial gradient peak segments from the axial thermal gradient data to obtain gradient peak segment data; b. Perform upper-side rapid cooling feature discrimination and lower-side stagnation feature discrimination on the gradient peak segment data to obtain upper-side rapid cooling feature data and lower-side stagnation feature data respectively; c. Based on the upper rapid cooling characteristic data and the lower heat stagnation characteristic data, the axial step structure is inspected to obtain temperature step data.

6. The adaptive optimization method for tower assembly welding sequence according to claim 1, characterized in that, Step S3 is as follows: S31: Decompose the risk situation based on the heat stress risk data to obtain the control domain data; S32: Generate the zoned thermal control boundary based on the control domain data to obtain the thermal target zone data; S33: Perform target budget allocation on the thermal target zone data to obtain control target data.

7. The adaptive optimization method for tower assembly welding sequence according to claim 1, characterized in that, The initial welding action is generated as follows: S41: Perform target analysis on the control target data to obtain motion constraint data; S42: Parameterize the welding action based on the action constraint data to obtain the welding action data; S43: Generate symmetrically paired welding action data by mirroring the data; S44: Perform thermal constraints based on the symmetric pairing data to obtain thermal constraint data; S45: Assemble the action sequence based on thermal constraint data and symmetrical pairing data to obtain initial welding action data.

8. The adaptive optimization method for tower assembly welding sequence according to claim 1, characterized in that, The specific steps of reconstructing the thermal stress welding sequence are as follows: S46: Perform thermal stress activation positioning on the initial welding action data to obtain conflict action data; S47: Subsequence splitting of conflict action data to obtain subsequence data; S48: Perform heat propagation coupling feedforward deduction based on subsequence data to obtain action partition impact data; S49: Reorder the constraints based on the action partition impact data to obtain the welding reconstruction data.

9. An adaptive optimization system for tower assembly welding sequence, characterized in that, For executing the adaptive optimization method for tower assembly welding sequence as described in claim 1, the adaptive optimization system for tower assembly welding sequence includes: Industrial computing nodes, data acquisition interfaces, and industrial communication interfaces; The industrial computing node includes a processor and a memory. The memory stores an executable program, and the processor calls the executable program to implement the following functional modules: The thermal field feature extraction module is used to acquire cooling status data through the data acquisition interface; and to extract thermal field features based on the cooling status data to obtain thermal field feature data. The heat stress risk central mapping processing module is used to obtain the heat field feature data from the heat field feature extraction module, and perform heat stress risk central mapping processing based on the heat field feature data to obtain heat stress risk data. Specifically, the heat stress risk central mapping processing includes: encoding the heat field input based on the heat field feature data to obtain heat field input data; obtaining heat setpoint data, and performing dual-path deviation calculation based on the heat field input data and the heat setpoint data to obtain dual-path deviation data, wherein the dual-path deviation calculation includes steady-state deviation calculation and transient deviation calculation; performing competitive fusion on the dual-path deviation data to obtain competitive situation data; and performing heat stress risk potential propagation based on the competitive situation data to obtain heat stress risk data. The thermal balance control target generation module is used to obtain the thermal stress risk data from the thermal stress risk central mapping processing module, and generate thermal balance control targets based on the thermal stress risk data to obtain control target data; The welding sequence generation and reconstruction module is used to obtain the control target data from the thermal balance control target generation module, generate initial welding actions based on the control target data to obtain initial welding action data, and reconstruct the initial welding action data by thermal stress welding sequence to obtain welding reconstruction data. The welding reconstruction data is sent to the welding equipment control system via the industrial communication interface to execute the corresponding welding sequence.

Images