Thermal insulation welding optimization method and system based on AI

Abstract

The invention relates to the technical field of intelligent control, in particular to an AI-based heat preservation welding optimization method and system, and the method comprises the following steps: obtaining welding node temperature data, recognizing an abnormal generation log, dividing partitions to generate control parameters, evaluating the thermal stability, adjusting the heat preservation time, and generating a heat preservation welding optimization scheme. According to the invention, through identification of a temperature curve derivative extreme value, enhancement of dynamic change perception capability, construction of a node identification mechanism based on coordinates and time progress, region division precision and pertinence are improved, a gradient difference value and a trend frequency index are introduced, a heat flow consistency judgment dimension is expanded, and a heat conduction stability evaluation effect is enhanced. And the heat preservation time is dynamically adjusted in combination with the deviation coefficient and the adjustment coefficient, a linkage path between static parameters and adjustment response is opened, the control scale is unified through time standardization processing, a dynamic closed-loop adjustment system of the multi-node heat preservation time is formed, and the adaptivity, stability and collaboration of welding path heat control are enhanced.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to an AI-based thermal insulation welding optimization method and system. Background Art

[0002] The field of intelligent control technology encompasses automated control methods that monitor, analyze, and adjust the dynamic parameters of various industrial and manufacturing processes in real time. Its core approach is to utilize control theory, information processing, and machine learning methods to perceive the state of complex systems, make decisions, and execute control. It is widely used in a variety of fields, including process industries, intelligent manufacturing, and energy scheduling, with the goal of improving system adaptability and control accuracy. Intelligent control systems typically optimize control strategies in real time based on behavioral models of the controlled object, combined with external environmental changes and historical data, thereby improving the responsiveness and stability of the overall system.

[0003] The AI-based insulation welding optimization method utilizes supervised learning mechanisms within artificial intelligence to perform multidimensional data modeling of physical parameters such as heat input, insulation duration, and welding sequence during the welding process. This method then constructs an optimization path based on the laws of heat conduction and analysis of material thermal performance parameters. To address the issue of fluctuating weld quality caused by unstable heat-affected zones in multi-layer welding processes, this method optimizes process parameters through temperature sensor data acquisition, numerical modeling of the thermal process, and welding path planning. This method utilizes three approaches: physical law modeling, historical data regression fitting, and control logic deduction to intelligently adjust welding heat control and insulation time strategies.

[0004] Existing insulation welding processes often use preset alarm modes based on temperature thresholds to address heat conduction issues at multiple welding nodes. These systems lack the ability to identify key temperature fluctuations, making it difficult to detect sudden local anomalies in the initial stages, leading to delayed and ineffective control strategies. Node identification and region division are often based on fixed topology or uniform distribution assumptions, failing to consider temporal and spatial coupling. This results in a single regional response and a disconnect between control objectives and actual thermal conditions. Thermal field analysis focuses on determining static gradient values, failing to analyze temperature trends from a multi-directional and multi-frequency perspective. This makes it difficult to deconstruct multi-path heat conduction relationships within complex structures. The time control mechanism relies on static time models or fixed compensation settings, lacking responsive control logic. This leads to frequent imbalances in node thermal behavior during multi-layer welding. Regarding unified parameter management, a parameter integration mechanism centered on standardized strategies has yet to be established. This results in error accumulation when control commands are transferred between different regions, weakening the effectiveness of collaborative control. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and propose an AI-based insulation welding optimization method and system.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solution: an AI-based thermal insulation welding optimization method, comprising the following steps: S1: Obtain continuous temperature records of each welding node during the insulation process, temperature change rate curve, detect the second-order derivative extreme point in the curve, compare it with the temperature gradient threshold, record temperature anomaly events, and generate a node temperature anomaly log; S2: Read the event coordinate information in the node temperature anomaly log, sort the nodes according to the node number and the time progress in the welding path, set the nodes where continuous abnormal events exceed the time window threshold as key control nodes, divide the welding area into several process zones according to the node spatial topology relationship, and generate regional collaborative control parameters; S3: Based on the regional collaborative control parameters, extract the difference between the horizontal and vertical temperature gradients of each node, calculate the temperature trend consistency index and trend reversal frequency in each direction, determine the degree of heat conduction consistency, and generate a regional thermal stability assessment report; S4: Based on the regional thermal stability assessment report, extract the current holding time record of the corresponding node, determine the time deviation coefficient of the node in the partition, set the PID adjustment coefficient and calculate the adjustment time based on the deviation, and superimpose it with the original time to obtain the process optimization time, and generate a process parameter optimization data set; S5: Based on the process parameter optimization data set and in combination with whether each partition has performed parameter adjustment operations, the partition nodes where the same-direction adjustment has occurred are extracted, the accumulated time deviations are counted and archived by partition, and an insulation welding optimization plan is generated.

[0007] As a further solution of the present invention, the node temperature anomaly log includes the temperature anomaly level, the frequency of anomaly occurrence, and the location information of the anomaly period; the regional collaborative control parameters include the coordinates of key control nodes, the partition topology relationship matrix, and the partition temperature fluctuation weight coefficient; the regional thermal stability assessment report includes the temperature consistency score, the thermal diffusion direction map, and the trend reversal density index; the process parameter optimization data set includes the standardized insulation time vector, the node adjustment response coefficient matrix, and the parameter correction recommendation list; the insulation welding optimization scheme includes the partition-level adjustment strategy and the cumulative deviation archiving report.

[0008] As a further solution of the present invention, the specific steps for obtaining the node temperature anomaly log are: S111: Obtain continuous temperature records of each welding node during the insulation process, calculate the temperature difference of each adjacent welding node in chronological order, record the temperature change corresponding to each time point, classify the nodes according to their numbers, and generate a temperature change rate trend curve; S112: Based on the temperature change rate trend curve, extract the continuous change rate data on each trend curve, calculate the second-order derivative sequence of the temperature change rate, identify the time points corresponding to the local maximum and minimum values, and generate a temperature change rate mutation value set; S113: Based on the mutation value of each node in the temperature change rate mutation set, compare it with the temperature gradient threshold one by one, screen the welding nodes whose mutation value is greater than the temperature gradient threshold, extract the corresponding time point and node number, and generate a node temperature anomaly log.

[0009] As a further solution of the present invention, the specific steps for obtaining the regional collaborative control parameters are: S211: Read the event coordinate information and node number recorded in the node temperature anomaly log, extract the time progress sequence in the corresponding welding path according to the node number, calculate the time interval of adjacent events in the same node, mark the nodes whose time interval of consecutive events is less than the set event association time range threshold as continuous abnormal nodes, obtain the corresponding abnormal duration, and generate the abnormal duration distribution; S212: Based on the continuous abnormal nodes in the abnormal duration distribution, screening the spatial adjacency relationship of each welding node according to the distance range, constructing node groups with spatial connectivity, numbering and marking all node groups, and generating a spatial node group sequence; S213: Based on the spatial node group sequence, the node numbers and welding time progress data in each group are extracted, the group synchronization deviation value is counted, and the number of abnormal events corresponding to the node numbers in the group is summarized and the abnormal event density is obtained. The dynamic synchronization control weight of each partition is calculated according to the synchronization deviation value and the abnormal event density, and the regional collaborative control parameters are generated.

[0010] As a further solution of the present invention, the dynamic synchronization control weight of each partition adopts the formula: ; Calculate, where Representative Group Dynamic synchronization control weight of the partition to which it belongs, represents the synchronization deviation density eigenvalue of group g, represents the average value of the synchronization deviation density eigenvalues ​​of all groups, Representative Group The average absolute deviation of the welding time of all nodes in represents the absolute deviation of the welding time of the mth node in group g, Representative Group The number of nodes in Represents the total number of groups.

[0011] As a further solution of the present invention, the specific steps for obtaining the regional thermal stability assessment report are: S311: Based on the regional collaborative control parameters and the temperature data of the corresponding nodes in each time period, extract the temperature values ​​of adjacent nodes in the same area in the horizontal and vertical directions, calculate the temperature difference between the nodes in each direction, record the horizontal and vertical differences in each time period, and establish a directional gradient difference; S312: Based on the horizontal and vertical temperature differences recorded in the directional gradient difference, calculate the temperature change direction in adjacent time periods, count the time proportion of continuous changes in the same direction, calculate the trend consistency degree in the horizontal and vertical directions respectively, and take the average value to obtain the temperature trend consistency index; S313: Based on the temperature trend consistency index, according to the temperature change direction sequence of each node, identify the moment when the change direction of adjacent time periods is reversed, count the number of positive and negative alternations of direction changes per unit time, combine the trend consistency index and the reversal frequency, judge the degree of continuity of heat conduction changes in each region, and obtain a regional thermal stability assessment report.

[0012] As a further solution of the present invention, the temperature trend consistency index adopts the formula: ; Calculate, where Representative direction The degree of trend consistency, represents the total number of analysis time periods in direction d, Representative direction Middle The incremental change of the temperature difference in a period relative to the previous period, Indicates that from the second time period to the The cumulative sum operation of the time period, Indicates the The absolute value of the temperature difference during the period, The item is used to determine whether two adjacent time periods change in the same direction. The term is used to introduce the similarity adjustment coefficient of the change amplitude between two adjacent time periods. When the change direction is consistent and the amplitude is similar, the value of this term is close to 1, otherwise it decreases, thereby improving the sensitivity of trend consistency judgment.

[0013] As a further solution of the present invention, the specific steps for obtaining the process parameter optimization data set are: S411: Based on the regional thermal stability assessment report, extract the current holding time records of all nodes in each partition, calculate the mean and standard deviation of the holding time of the nodes in each partition, divide the difference between the holding time of a single node and the partition mean by the standard deviation to obtain the time deviation coefficient of each node, summarize the node numbers of each partition and the corresponding deviation coefficients, and establish a time deviation coefficient matrix; S412: Based on the time deviation coefficient matrix, the proportional coefficient, integral coefficient, and differential coefficient in the PID adjustment coefficient are set. The proportional term is multiplied by the deviation value, the integral term is multiplied by the cumulative sum of the current deviation and the previous deviation, and the differential term is multiplied by the difference between the current deviation and the previous deviation. The three terms are added together to obtain the adjustment time value corresponding to each node. The updated time is obtained by adding the three terms to the original holding time of the node, and the time adjustment increment data is obtained by summing them up. S413: Based on the updated holding time records of each node in the time adjustment incremental data, classification and extraction are performed according to the partition number, and the holding time of all nodes in each partition is normalized respectively to form a process parameter optimization data set.

[0014] As a further solution of the present invention, the specific steps for obtaining the thermal insulation welding optimization solution are: S511: Based on the process parameter optimization data set and the parameter adjustment status identified in the partition control record, the partitions that have been subjected to parameter adjustment operations are screened, the holding time change direction identifiers of all nodes in each partition are extracted, and whether the adjustment is in the same direction is determined based on the consistency of the direction. The node numbers and partition numbers that meet the same direction adjustment condition are recorded to obtain a same direction adjustment node sequence; S512: Based on each node number in the same-direction adjustment node sequence, the corresponding holding time increment value sequence is retrieved, the area of ​​the time period of continuous same-direction change within a unit time is calculated as the corresponding node cumulative offset value, the cumulative offset values ​​of all nodes in the same partition are accumulated, and the accumulated offset values ​​are classified by partition number to generate a cumulative time deviation; S513: According to each partition number and the corresponding adjustment value in the accumulated time deviation, combined with the node distribution map and the current optimization state, the standardized time adjustment results of each area are matched according to each partition number to obtain the insulation welding optimization plan.

[0015] AI-based thermal insulation welding optimization system, the system includes: The node temperature analysis module obtains continuous temperature records of each welding node during the insulation process, detects extreme points in the curve, records temperature anomalies, and generates node temperature anomaly logs; The area division module is based on the node temperature anomaly log, sorts the nodes according to the node number and the time progress in the welding path, sets the key control nodes to divide the welding area into several process zones, and generates regional collaborative control parameters; The stability assessment module extracts the node temperature gradient difference based on the regional collaborative control parameters, calculates the temperature trend consistency index, determines the degree of heat conduction consistency, and generates a regional thermal stability assessment report; The parameter optimization module determines the node time deviation coefficient within the partition based on the regional thermal stability assessment report, sets the PID adjustment coefficient, calculates the process optimization time based on the deviation, and generates a process parameter optimization data set; The insulation optimization module extracts the partition nodes where the same-direction adjustment occurs based on the process parameter optimization data set, calculates the accumulated time deviation and archives it by partition, and generates an insulation welding optimization plan.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by identifying the extreme values ​​of the derivative of the temperature curve, the rapid capture of sudden thermal anomalies is achieved, the ability to perceive dynamic changes is enhanced, a node identification mechanism is constructed based on coordinates and time progress, the accuracy and pertinence of regional division are improved, the gradient difference and trend frequency indicators are introduced, the heat flow consistency judgment dimension is expanded, the thermal conduction stability evaluation effect is enhanced, the holding time is dynamically adjusted in combination with the deviation coefficient and the adjustment coefficient, the linkage path between the static parameters and the adjustment response is opened, the time standardization processing is introduced to unify the control scale, the cross-region control compatibility and coordination are improved, a dynamic closed-loop adjustment system for multi-node holding time is formed, and the adaptability, stability and coordination of the thermal control of the welding path are enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Schematic diagram of the steps of the present invention; Figure 2 A flowchart for obtaining a node temperature anomaly log according to the present invention; Figure 3 A flow chart for obtaining regional collaborative control parameters of the present invention; Figure 4 Obtain a flow chart for the regional thermal stability assessment report of the present invention; Figure 5 Obtaining a flow chart for the process parameter optimization data set of the present invention; Figure 6 A flow chart is obtained for the thermal insulation welding optimization scheme of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0019] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0020] See also Figure 1 , the AI-based thermal insulation welding optimization method includes the following steps: S1: Obtain continuous temperature records of each welding node during the insulation process, perform differential processing based on the temperature values ​​of adjacent time periods to construct a temperature change rate curve, detect the second-order derivative extreme point in the curve, and compare it with the temperature gradient threshold (the allowable temperature change rate standard (GB / T19869.1) set according to the material phase transition point). If the change is greater than the temperature gradient threshold, it is recorded as a temperature anomaly event and a node temperature anomaly log is generated; S2: Read the event coordinate information in the node temperature anomaly log, sort the nodes according to the node number and the time progress in the welding path, set the nodes where continuous abnormal events exceed the time window threshold (the event correlation time range determined by the welding process specification) as key control nodes, divide the welding area into several process zones according to the node spatial topology relationship (the adjacency relationship model established according to the spatial position of the welding nodes), and generate regional collaborative control parameters; S3: Based on the regional collaborative control parameters, the difference between the horizontal and vertical temperature gradients of each node is extracted, and the temperature trend consistency index (a quantitative indicator reflecting the synchronization of regional temperature changes) and the trend reversal frequency in each direction are calculated to determine the degree of heat conduction consistency and generate a regional thermal stability assessment report. S4: Based on the regional thermal stability assessment report, extract the current holding time records of the corresponding nodes, determine the time deviation coefficient of the nodes in the partition, set the PID adjustment coefficient (a standard parameter combination of the proportional-integral-differential control algorithm), and calculate the adjustment time based on the deviation. The time is then added to the original time to obtain the process optimization time. The updated partition time is then standardized to generate a process parameter optimization data set. S5: Based on the process parameter optimization data set and considering whether each partition has performed parameter adjustment operations, the partition nodes where the same-direction adjustment has occurred are extracted, the accumulated time deviation (the time dimension integral value of the process parameter adjustment amount) is counted and archived by partition to generate the insulation welding optimization plan.

[0021] The node temperature anomaly log includes the temperature anomaly level, anomaly frequency, and anomaly period location information. The regional collaborative control parameters include the coordinates of key control nodes, partition topology relationship matrix, and partition temperature fluctuation weight coefficient. The regional thermal stability assessment report includes temperature consistency score, thermal diffusion direction map, and trend reversal density index. The process parameter optimization data set includes standardized insulation time vector, node adjustment response coefficient matrix, and parameter correction recommendation list. The insulation welding optimization plan includes partition-level adjustment strategy and cumulative deviation archiving report.

[0022] See also Figure 2 , the specific steps of S1 are: S111: Obtain continuous temperature records of each welding node during the insulation process, calculate the temperature difference of each adjacent welding node in chronological order, record the temperature change corresponding to each time point, classify the nodes according to their numbers, and generate a temperature change rate trend curve; To obtain continuous temperature records of each welding node during the insulation process, it is necessary to first arrange thermocouples or infrared temperature measuring instruments at all nodes to ensure that the measurement point positions correspond to the welding node numbers. For example, in the pipeline section, welding nodes numbered N1 to N20 are arranged every 0.5 meters. At the same time, a temperature acquisition system is configured to record the temperature value every 1 second to generate a temperature time series for each node. , in the acquisition process, the sampling time is set to 3600 seconds, and 3600 temperature data are obtained for each node. Then the temperature difference of each adjacent welding node is calculated according to the time point, that is, the temperature difference of each adjacent welding node is calculated. Next adjacent node and implement Operations, such as time When the temperature of nodes N5 and N6 is 88.3℃ and 86.7℃ respectively, the temperature difference is 1.6℃. This type of calculation is performed at each time point to form a node difference time series matrix. After the difference calculation is completed, all difference sequences are reclassified according to the node number, that is, the data of each pair of adjacent nodes are grouped and stored in the form of numbers. For example, for node N5, its attribution item is and Then, the time variables are sorted and extracted according to the temperature difference sequence of each node, and the temperature difference change trend in the current time segment is combined to obtain the temperature difference change rate using a single difference method. , for example, at node pair N5-N6, if , , then the rate of change is ,Finally, the change rates of all node pairs over time constitute a ,temperature change rate trend curve, which is convenient for subsequent ,analysis.

[0023] S112: Based on the temperature change rate trend curve, extract the continuous change rate data on each trend curve, calculate the second-order derivative sequence of the temperature change rate, identify the time points corresponding to the local maximum and minimum values, and generate a temperature change rate mutation value set; Based on the temperature change rate trend curve, a continuous sequence of change rate data points needs to be extracted from each curve. For example, for node pair N5-N6, the change rate data sequence is , calculate the second-order derivative of the sequence, using the difference method: , such as at time t=1202s, if , , then the second-order derivative is , identify local extreme points through the positive and negative changes and their sizes of the sequence: when the second-order derivative at a certain time point is positive and the previous and next derivatives are negative to positive, it is identified as a minimum value; if it is negative and the previous and next derivatives are positive to negative, it is identified as a maximum value. In this way, all node pairs of trend curves are processed in sequence to extract the time of the extreme point, that is, the mutation time set , and record the mutation value of the time point with the node pair number. For example, in the node pair N5-N6, three mutation points are identified at 1202s, 1755s, and 2890s, and their mutation values ​​are 2.1℃ / s, -1.8℃ / s, and 3.2℃ / s, respectively. Finally, the temperature change rate mutation value set corresponding to all node pairs at each time point is generated.

[0024] S113: Based on the mutation value of each node in the temperature change rate mutation set, the mutation value is compared with the temperature gradient threshold item by item, and the welding nodes whose mutation value is greater than the temperature gradient threshold are selected. The corresponding time point and node number are extracted, and a node temperature anomaly log is generated; Based on the mutation value set generated above, each mutation value needs to be compared with the preset temperature gradient threshold one by one. For example, the temperature change rate threshold is set to ±2.0℃ / s. This threshold is determined by the process standard. The threshold source is the average change rate of non-abnormal nodes in the insulation process of the past 200 welding records of ±1.2℃ / s. The standard deviation of ±1.5 times is set to 2.0℃ / s. The specific calculation is as follows: Assuming that the standard deviation of the change rate sample is 0.53℃ / s, the threshold is When the mutation rate of a node is greater than 2.0℃ / s or less than -2.0℃ / s, it is judged as abnormal. For example, in the node pair N5-N6 mentioned above, the mutation rate at 1202s is 2.1℃ / s, which meets the abnormal condition. The mutation rate at 2890s is 3.2℃ / s, which is also abnormal. However, the mutation rate at 1755s is -1.8℃ / s, which does not meet the threshold condition and is therefore not included in the abnormal range. After screening, the time and node number of the abnormal point are retained to form an abnormal log, as shown in Table 1.

[0025] Table 1 Temperature mutation abnormal node log table As shown in Table 1, the log clearly indicates that at 1202 seconds and 2890 seconds, the mutation values ​​of nodes N5-N6 reached 2.1°C / s and 3.2°C / s, respectively, both exceeding the preset threshold of 2.0°C / s and thus being classified as abnormal behavior records for subsequent analysis.

[0026] See also Figure 3 , the specific steps of S2 are: S211: Read the event coordinate information and node number recorded in the node temperature anomaly log, extract the time progress sequence in the corresponding welding path according to the node number, calculate the time interval of adjacent events in the same node, mark the nodes with consecutive event time intervals less than the set event correlation time range threshold as continuous abnormal nodes, obtain the corresponding abnormal duration, and generate the abnormal duration distribution; Read the event coordinate information and node number recorded in the node temperature anomaly log. First, extract the corresponding node number (such as N5, N6) and the time point when the anomaly occurred (such as 1202s, 2890s) for each anomaly record in the log. In the stored welding process information library, index the corresponding welding path and time progress sequence according to the node number. For example, the welding process of node N5 starts at 1000s and ends at 3600s. There is a temperature record per second. This information can be used to locate the specific position of the abnormal event in the welding process. Then, sort the multiple abnormal events with the same number by node and calculate the time interval between two adjacent events. For example, node N5 has an abnormal event at 1202s and 1230s, with an interval of 28s. If the preset event association time range threshold is 30s, the event interval in this node is less than the threshold, and it is marked as a continuous abnormal node. The total time period of the continuous abnormal events in this node is further extracted. If the continuous abnormal sequence is 1202s, 1230s, 1255s, and 1270s, the interval between each two is calculated to be less than 30s. If it is satisfied in sequence, the final duration is 1270s-1202s=68s. Repeat this operation for all nodes to form a set of abnormal duration data series. ,For example , the sequence is statistically analyzed to generate the duration distribution frequency, which is divided into time intervals (e.g. 0–30s, 31–60s, 61–90s). The number of events falling into each time interval is calculated, and the duration distribution is formed as shown in Table 2.

[0027] Table 2 Statistics of abnormal duration distribution As shown in Table 2, the distribution of the number of nodes in different duration intervals provides a basis for subsequent partitioning of spatial association relationships.

[0028] S212: Based on the continuous abnormal nodes in the abnormal duration distribution, the spatial adjacency relationship of each welding node is screened by distance range, node groups with spatial connectivity are constructed, all node groups are numbered and marked, and a spatial node group sequence is generated; Based on the continuous abnormal nodes in the abnormal duration distribution, the spatial coordinates of each node are first extracted. Assuming that the node coordinate format is , based on the preset spatial adjacent distance range threshold , compare the distances between all consecutive abnormal nodes, and perform the following operations: if any two nodes and satisfy , then it is determined that the spatial adjacency relationship is established. For example, the coordinates of node N5 are (1.2, 3.5, 0.0), and N6 are (1.3, 3.6, 0.0), and the distance is If the threshold is 0.5 meters, the spatial adjacency between the two is established. According to this rule, the adjacency matrix is ​​constructed and the spatial connectivity grouping operation is performed. That is, there is an adjacency chain between the node pairs in each group to form a spatial node group. The group number is marked in the order of processing. For example, group G1 contains nodes N3, N4, and N5, and group G2 contains N8, N9, N10, and N11, forming a spatial node group sequence. , each group contains at least two nodes and the spatial distance between nodes satisfies the adjacency rule.

[0029] S213: Based on the spatial node group sequence, extract the node number and welding time progress data in each group, calculate the group synchronization deviation value, summarize the number of abnormal events corresponding to the node number in the group, and obtain the abnormal event density. Calculate the dynamic synchronization control weight of each partition based on the synchronization deviation value and the abnormal event density, and generate the regional collaborative control parameters; The dynamic synchronization control weight of each partition is calculated using the formula: ; Calculate, where Representative Group Dynamic synchronization control weight of the partition to which it belongs, represents the synchronization deviation density eigenvalue of group g, represents the average value of the synchronization deviation density eigenvalues ​​of all groups, Representative Group The average absolute deviation of the welding time of all nodes in represents the absolute deviation of the welding time of the mth node in group g, Representative Group The number of nodes in Represents the total number of groups.

[0030] A welding section has five measuring nodes, numbered N1 to N5. The monitoring system collects abnormal events and records welding time at each node, obtaining the following data: The number of abnormal events N1 to N5 are: [6, 9, 7, 5, 8], obtained through the thermal imaging recognition and recording module; The corresponding welding time schedules of N1 to N5 are: [1020s, 1012s, 1016s, 1030s, 1025s], which are collected through the sampling interface of the industrial computer; The node time deviation value is calculated for each node based on the process reference point 1030s, in seconds; The group g to which each node belongs is G2. The other groups G1 and G3 are composed of different nodes. Their synchronization deviation density characteristic values ​​are as follows: =0.85, =1.04, =0.93.

[0031] Average number of anomalies ; The absolute value of the difference between each node and the average value: [1, 2, 0, 2, 1]; The average welding time schedule is: ; Welding time deviation value: [0.6, 8.6, 4.6, 9.4, 4.4], calculated , and then we get: ; Mean absolute deviation term: ; Compute the global average density eigenvalue: ; Substituting into the formula, the numerator is: ; The denominator is: ; The final control weight result is: ; The results show that the dynamic synchronization control weight of group G2 in the current region is 0.04225. This weight is derived from the deviation between the group's eigenvalue and the global average density and its internal temporal volatility. A larger value indicates greater synchronization fluctuations and density differences within the group, indicating that synchronization control efforts should be strengthened in this region.

[0032] The dynamic synchronization control weight of each partition is used to quantify the relative control priority of different areas in the overall welding synchronization control system. The weight comprehensively considers the fluctuation of abnormal event density and welding time deviation of nodes in each area, and reflects the degree of node synchronization deviation and internal collaborative stability as the measurement basis. When system resources are limited or control parameters need to be dynamically allocated, partitions with high weight values ​​represent larger synchronization risks and significant abnormal interference, and feedback adjustment or precise intervention should be given priority. The calculation results of the weight values ​​are directly involved in the dynamic configuration of regional collaborative parameters, realizing differentiated control of the synchronization response capabilities of different spatial areas under multi-source interference, thereby enhancing the coordination and beat stability of the overall manufacturing process.

[0033] The dynamic synchronization control weight of each partition is used to quantify the relative control priority of different areas in the overall welding synchronization control system. The weight comprehensively considers the fluctuation of abnormal event density and welding time deviation of nodes in each area, and reflects the degree of node synchronization deviation and internal collaborative stability as the measurement basis. When system resources are limited or control parameters need to be dynamically allocated, partitions with high weight values ​​represent larger synchronization risks and significant abnormal interference, and feedback adjustment or precise intervention should be given priority. The calculation results of the weight values ​​are directly involved in the dynamic configuration of regional collaborative parameters, realizing differentiated control of the synchronization response capabilities of different spatial areas under multi-source interference, thereby enhancing the coordination and beat stability of the overall manufacturing process.

[0034] See also Figure 4 , the specific steps of S3 are: S311: Based on the regional collaborative control parameters and the temperature data of the corresponding nodes in each time period, the temperature values ​​of adjacent nodes in the same area are extracted in the horizontal and vertical directions. The temperature difference between the nodes in each direction is calculated. The horizontal and vertical differences in each time period are recorded to establish the directional gradient difference. According to the regional collaborative control parameters, first determine the numbers of all nodes contained in the target area node group and extract their corresponding temperature monitoring data. In the specific implementation, take the spatial node group G1 (such as nodes N3, N4, and N5) as an example, extract its temperature data sequence in the time period from t=1000s to t=1200s, set the analysis time period to be a group of 10 seconds, and a total of 5 groups of time periods. Identify the adjacency relationship between nodes in the horizontal and vertical directions, where the horizontal adjacency relationship is such as N3-N4 and the vertical adjacency relationship is such as N3-N5. Extract the temperature values ​​of adjacent nodes in the same time period and perform a subtraction operation to obtain the temperature difference. For example, in time period T1, the temperature of N3 is 85.2℃, and that of N4 is 86.4℃. The horizontal temperature difference is 1.2℃, and the vertical temperature difference between N3 and N5 is 1.0℃. Record these differences and sort them according to the time period number. After processing all time periods in sequence, establish the horizontal temperature difference sequence respectively. With the longitudinal temperature difference series , which is the directional gradient difference, as shown in Table 3.

[0035] Table 3 Directional gradient difference scale As shown in Table 3, the temperature difference data in the horizontal and vertical directions at different time periods provide a basic basis for the calculation of subsequent trend changes and reversal frequencies.

[0036] S312: Based on the horizontal and vertical temperature differences recorded in the directional gradient difference, calculate the temperature change direction in adjacent time periods, count the time proportion of continuous changes in the same direction, calculate the trend consistency degree in the horizontal and vertical directions respectively, and take the average value to obtain the temperature trend consistency index; Temperature trend consistency index, using the formula: ; Calculate, where Representative direction The degree of trend consistency, represents the total number of analysis time periods in direction d, Representative direction Middle The incremental change of the temperature difference in a period relative to the previous period, Indicates that from the second time period to the The cumulative sum operation of the time period, Indicates the The absolute value of the temperature difference during the period, The item is used to determine whether two adjacent time periods change in the same direction. The term is used to introduce the similarity adjustment coefficient of the change amplitude between two adjacent time periods. When the change direction is consistent and the amplitude is similar, the value of this term is close to 1, otherwise it decreases, thereby improving the sensitivity of trend consistency judgment.

[0037] In the transverse direction of a welding area In the process, temperature sensors are placed and a set of horizontal temperature difference data is collected every 20 seconds. The temperature difference sequence of the time period is obtained through the on-site monitoring system: ; According to the definition, the total number of directional analysis periods is: ; Substitute the formula and calculate item by item: Item 2: ; Item 3: ; Item 4: ; Item 5: ; Item 6: ; Add the sum of the items and divide by the number of periods: ; This result shows that the direction The change in temperature difference during this time period showed a trend direction consistency of 0.7814, which is used to reflect the continuity and stability of temperature fluctuations in the current spatial direction, and provides data support for subsequent trend identification modules and reversal assessments.

[0038] The temperature trend consistency index is used to measure the degree of consistency between the direction and amplitude of temperature changes in a certain spatial direction within a continuous time period. Its value ranges from 0 to 1. When the index is close to 1, it means that the temperature change maintains the same trend in multiple time periods and the adjacent fluctuation amplitudes are similar, reflecting strong thermal conduction stability and thermal field continuity. When the index approaches 0, it means that the temperature change frequently reverses direction or fluctuates violently in amplitude, reflecting disordered heat distribution or discontinuous control. This indicator is used to describe the coordination characteristics of the temperature evolution state during thermal processing such as welding. It is the basic numerical support in temperature trend judgment, anomaly detection and process control.

[0039] The calculation logic of the formula is based on the consistency measurement of the direction and amplitude of the temperature difference change. In the overall structure, the denominator Normalization is achieved to ensure that the output indicators are within a comparable range. The sum symbol represents the accumulation across time periods, reflecting the global trend. In the fractional structure, the numerator part and the denominator The ratio is used to determine whether the temperature change direction in adjacent time periods is consistent. When the direction is consistent, the ratio is 1, and when the direction is opposite, the value is close to 0; the correction term multiplied by it is It is used to measure the similarity of amplitude changes. When the amplitudes of the two are similar, the value is close to 1. Otherwise, the value approaches 0 after the deviation increases. The product of the two is superimposed as the trend consistency weight to reflect the performance of the trend continuity during this period, thereby forming a comprehensive evaluation factor for the temperature fluctuation trend in the continuous time series.

[0040] S313: Based on the temperature trend consistency index and the temperature change direction sequence of each node, the moment when the change direction reverses in adjacent time periods is identified. The number of positive and negative direction changes per unit time is counted. Combining the trend consistency index and the reversal frequency, the degree of continuity of heat conduction changes in each region is determined to obtain a regional thermal stability assessment report. Based on the trend consistency index, the aforementioned direction change sequence is further analyzed to identify whether the direction of change in adjacent time periods is reversed. The positive and negative signs are used to alternately judge, such as [+1, -1] is a reversal, and [-1, -1] is no reversal. The number of reversals in the entire sequence is counted. In the horizontal sequence [+1, -1, -1, +1], T1-T2 is a reversal, T2-T3 is no reversal, and T3-T4 is a reversal, for a total of 2 times. The average reversal frequency per unit time is 2 / 4=0.5 times / period. Similarly, there are 2 reversals in the vertical direction, and the frequency is also 0.5. The reversal frequency is combined with the trend consistency index. If the trend consistency index is in the range of [0.6, 1.0] and the reversal frequency is less than 0.3 times / period, it is defined as high continuity. If it is in the range of [0.4, 0.6] and the frequency is between 0.3-0.7, it is medium continuity. If it is less than 0.4 or the frequency is greater than 0.7, it is low continuity. The consistency index of this area is 0.5 and the frequency is 0.5, which belongs to the medium continuity area. Based on this, the thermal stability assessment results of the group area are formed and included in the regional thermal stability assessment report to determine whether there are structural abnormalities in the welding heat conduction process.

[0041] See also Figure 5 , the specific steps of S4 are: S411: Based on the regional thermal stability assessment report, extract the current holding time records of all nodes in each partition, calculate the mean and standard deviation of the holding time of the nodes in each partition, divide the difference between the holding time of a single node and the partition mean by the standard deviation to obtain the time deviation coefficient of each node, summarize the node numbers of each partition and the corresponding deviation coefficients, and establish a time deviation coefficient matrix; Based on the regional thermal stability assessment report, first read the classified node set information in each partition and extract the current insulation time record data of each node. For example, in partition A, it contains nodes N1 to N5, whose insulation times are 370s, 390s, 410s, 385s, and 395s respectively. First calculate the mean insulation time of the nodes in the partition seconds, and then calculate the standard deviation seconds, and then calculate the holding time deviation coefficient for each node, which is defined as the holding time of the node minus the partition mean and divided by the standard deviation. For example, the deviation coefficient of node N1 is , node N3 is The rest of the nodes are deduced in the same way, and finally the following deviation coefficient matrix is ​​formed: Partition A→N1: -1.58, N2: 0.00, N3: +1.58, N4: -0.39, N5: +0.39. Each node number is bound to its corresponding time deviation coefficient, and they are classified and summarized by partition to form a time deviation coefficient matrix, which serves as the basis for subsequent adjustment time calculations.

[0042] S412: Based on the time deviation coefficient matrix, the proportional coefficient, integral coefficient, and differential coefficient in the PID adjustment coefficient are set. The proportional term is multiplied by the deviation value, the integral term is multiplied by the cumulative sum of the current deviation and the previous deviation, and the differential term is multiplied by the difference between the current deviation and the previous deviation. The three terms are added together to obtain the adjustment time value corresponding to each node. The updated time is obtained by adding the three terms to the original holding time of the node, and the time adjustment increment data is obtained by summing them up. Based on the time deviation coefficient matrix, set the three adjustment coefficients and proportional coefficient in PID control , integral coefficient , differential coefficient , the coefficient value is selected based on the control response sensitivity standard and is set to 、 、 , where the proportional term is calculated as the current deviation value and For example, if the current deviation of node N1 is -1.58, the proportional term is The integral term is calculated as the sum of the current deviation and the previous cumulative deviation multiplied by , assuming the current cumulative deviation is -3.0, the integral term is The differential term is the difference between the current deviation and the previous deviation multiplied by , if the last deviation is -1.0, then the differential term is The sum of the three items gives the adjusted time value of the node: This value is added to the original holding time of the node, 370 seconds, and the updated holding time is 356.44 seconds. All nodes are processed in this way to generate a complete time adjustment incremental data sequence, which constitutes the adjustment timetable corresponding to each node.

[0043] S413: Based on the updated holding time records of each node in the time-adjusted incremental data, the data is classified and extracted by partition number, and the holding time of all nodes in each partition is normalized to form a process parameter optimization data set. Based on the adjusted incremental data obtained above, read the original and updated holding times of each node, and sort and extract the data according to the partition number. For example, the updated holding times of the nodes in partition A are N1: 356.44s, N2: 390s, N3: 406.5s, N4: 382.3s, and N5: 398.8s. Normalize the sequence by first determining the minimum and maximum values ​​within the partition, which are 356.44s and 406.5s, respectively. Normalization uses linear mapping: , the normalized time of node N1 is , node N3 is , N5 is , and the rest are calculated in sequence, eventually forming a normalized holding time data set of all nodes under each partition. Combined with the original trend index, thermal stability level and other indicator parameters, a process parameter optimization data set for thermal control optimization is formed.

[0044] See also Figure 6 , the specific steps of S5 are: S511: Based on the process parameter optimization data set and the parameter adjustment status identified in the partition control record, the partitions that have been subjected to parameter adjustment operations are screened, and the holding time change direction identifiers of all nodes in each partition are extracted. Based on the direction consistency, whether the adjustment is in the same direction is determined. The node numbers and partition numbers that meet the same direction adjustment condition are recorded to obtain a same direction adjustment node sequence; Based on the process parameter optimization data set, first extract the partition number and locate its adjustment status in the parameter adjustment record table, filter out all partitions recorded as "adjusted", for example, partitions A and C are set to "adjusted", retain their numbers and exclude unadjusted items, and extract the direction of change of the holding time before and after adjustment for each node number in each adjusted partition. The judgment method is to calculate the difference between the updated holding time and the original holding time. If the difference is positive, it is defined as "upward adjustment", and if it is negative, it is defined as "downward adjustment", for example, node N1 is increased from 370s to 378s The direction is "upward adjustment", and the direction of node N2 is "downward adjustment" from 390s to 385s. Then, the change directions of all nodes in each partition are traversed to determine whether they are all in the same direction. If all nodes in a partition change in the same direction, they are marked as same-direction adjustment. For example, in partition A, if the directions of N1, N3, and N4 are all "upward adjustment", it is considered to be same-direction adjustment. All node numbers that meet the conditions are extracted and combined with partition numbers to form a data sequence, such as (A, N1), (A, N3), and (A, N4). Finally, a same-direction adjustment node sequence is constructed for subsequent offset accumulation.

[0045] S512: Based on each node number in the same-direction adjustment node sequence, the corresponding holding time increment value sequence is retrieved, and the area of ​​the time period with continuous same-direction changes per unit time is calculated as the corresponding node cumulative offset value. The cumulative offset values ​​of all nodes in the same partition are accumulated and classified by partition number to generate a cumulative time deviation value. Based on the number of each node in the above-mentioned same-direction adjustment node sequence, the time increment sequence experienced by each node in the time adjustment record table is called. Assume that the sequence is the time change recorded at 1-second intervals. For example, the node N1 adjustment process lasts for 10 seconds, and its increment sequence is [+0.8, +0.9, +1.0, +1.1, +1.0, +1.2, +1.1, +1.0, +0.9, +0.8] seconds. Determine whether it is continuous in the same direction, that is, all values ​​in the sequence have the same sign (positive or negative). After meeting the "same-direction change" condition, the area calculation method is used to accumulate the time increments within this period as the cumulative offset value of the node. The specific accumulation method is to sum the values ​​of the sequence to obtain the total offset. The N1 offset is After repeating the operation to obtain the cumulative offset values ​​of all nodes in the same direction, the accumulated values ​​are classified and added according to the partition number. If partition A contains three nodes N1, N3, and N4, and the offset values ​​are 10.8s, 12.5s, and 11.2s respectively, the accumulated time deviation of partition A is All partition results constitute a cumulative time deviation data set classified by number, providing a basis for time adjustment amplitude for further optimization schemes.

[0046] S513: Based on each partition number and corresponding adjustment value in the accumulated time deviation, combined with the node distribution map and the current optimization state, the standardized time adjustment results of each area are matched according to each partition number to obtain the insulation welding optimization plan; According to each partition number and its corresponding adjustment value in the cumulative time deviation, read the topological structure information in the spatial layout diagram of each partition node, locate the relative layout position of each node in the spatial grid, and adjust the result sequence in combination with the original regional standardized time. For example, the standard value interval is the normalized holding time value , determine whether each node falls within the standard interval after adjustment, and match the adjustment direction and adjustment amount in the cumulative time deviation of the nodes that do not fall within the interval according to their partitions, and remap the adjustment results. For example, after adjustment, the holding time of node N1 is 0.75 (exceeding the upper limit of 0.7), and the corresponding deviation is +10.8 seconds. In this case, the time value needs to be lowered to the boundary of the standard interval, and the adjustment range is recalculated as The normalized unit is multiplied by the original time range of 50 seconds to obtain a corrected time value of 2.5 seconds. The final corrected node insulation time is the original time minus 2.5 seconds. All such nodes are standardized and corrected once according to the adjustment direction and deviation value. The updated result is compared with the original standard value and the final adjustment parameter is recorded. The overall insulation welding optimization plan is generated by summarizing the partition number to ensure that the plan meets the thermal stability control requirements in the region in the time dimension.

[0047] AI-based thermal insulation welding optimization system, the system includes: The node temperature analysis module obtains continuous temperature records of each welding node during the insulation process, detects extreme points in the curve, records temperature anomalies, and generates node temperature anomaly logs; The area division module is based on the node temperature anomaly log, sorts the nodes according to the node number and the time progress in the welding path, sets the key control nodes to divide the welding area into several process zones, and generates regional collaborative control parameters; The stability assessment module extracts the node temperature gradient difference based on the regional collaborative control parameters, calculates the temperature trend consistency index, determines the degree of heat conduction consistency, and generates a regional thermal stability assessment report; The parameter optimization module determines the node time deviation coefficient within the partition based on the regional thermal stability assessment report, sets the PID adjustment coefficient, calculates the process optimization time based on the deviation, and generates a process parameter optimization data set; The insulation optimization module extracts the partition nodes where the same-direction adjustment occurs based on the process parameter optimization data set, calculates the accumulated time deviation and archives it by partition, and generates an insulation welding optimization plan.

[0048] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. AI-based thermal insulation welding optimization method, characterized in that: The following steps are involved: S1: Obtain continuous temperature records of each welding node during the insulation process, draw a temperature change rate curve, detect the second-order derivative extreme point in the curve, compare it with the temperature gradient threshold, record temperature anomalies, and generate a node temperature anomaly log; S2: Read the event coordinate information in the node temperature anomaly log, sort the nodes according to the node number and the time progress in the welding path, set the nodes where continuous abnormal events exceed the time window threshold as key control nodes, divide the welding area into several process zones according to the node spatial topology relationship, and generate regional collaborative control parameters; S3: Based on the regional collaborative control parameters, extract the difference between the horizontal and vertical temperature gradients of each node, calculate the temperature trend consistency index and trend reversal frequency in each direction, determine the degree of heat conduction consistency, and generate a regional thermal stability assessment report; S4: Based on the regional thermal stability assessment report, extract the current holding time record of the corresponding node, determine the time deviation coefficient of the node in the partition, set the PID adjustment coefficient and calculate the adjustment time based on the deviation, and superimpose it with the original time to obtain the process optimization time, and generate a process parameter optimization data set; S5: Based on the process parameter optimization data set and in combination with whether each partition has performed parameter adjustment operations, the partition nodes where the same-direction adjustment has occurred are extracted, the accumulated time deviations are counted and archived by partition, and an insulation welding optimization plan is generated.

2. The AI-based thermal insulation welding optimization method according to claim 1, characterized in that: The node temperature anomaly log includes the temperature anomaly level, anomaly frequency, and anomaly period location information; the regional collaborative control parameters include key control node coordinates, partition topology relationship matrix, and partition temperature fluctuation weight coefficient; the regional thermal stability assessment report includes temperature consistency score, thermal diffusion direction map, and trend reversal density index; the process parameter optimization data set includes standardized insulation time vector, node adjustment response coefficient matrix, and parameter correction recommendation list; the insulation welding optimization plan includes partition-level adjustment strategy and cumulative deviation archiving report.

3. The AI-based thermal insulation welding optimization method according to claim 1, characterized in that: The specific steps for obtaining the node temperature anomaly log are: S111: Obtain continuous temperature records of each welding node during the insulation process, calculate the temperature difference of each adjacent welding node in chronological order, record the temperature change corresponding to each time point, classify the nodes according to their numbers, and generate a temperature change rate trend curve; S112: Based on the temperature change rate trend curve, extract the continuous change rate data on each trend curve, calculate the second-order derivative sequence of the temperature change rate, identify the time points corresponding to the local maximum and minimum values, and generate a temperature change rate mutation value set; S113: Based on the mutation value of each node in the temperature change rate mutation set, compare it with the temperature gradient threshold one by one, screen the welding nodes whose mutation value is greater than the temperature gradient threshold, extract the corresponding time point and node number, and generate a node temperature anomaly log.

4. The AI-based thermal insulation welding optimization method according to claim 1, characterized in that: The specific steps for obtaining the regional collaborative control parameters are: S211: Read the event coordinate information and node number recorded in the node temperature anomaly log, extract the time progress sequence in the corresponding welding path according to the node number, calculate the time interval of adjacent events in the same node, mark the nodes whose time interval of consecutive events is less than the set event association time range threshold as continuous abnormal nodes, obtain the corresponding abnormal duration, and generate the abnormal duration distribution; S212: Based on the continuous abnormal nodes in the abnormal duration distribution, screening the spatial adjacency relationship of each welding node according to the distance range, constructing node groups with spatial connectivity, numbering and marking all node groups, and generating a spatial node group sequence; S213: Based on the spatial node group sequence, the node numbers and welding time progress data in each group are extracted, the group synchronization deviation value is counted, and the number of abnormal events corresponding to the node numbers in the group is summarized and the abnormal event density is obtained. The dynamic synchronization control weight of each partition is calculated according to the synchronization deviation value and the abnormal event density, and the regional collaborative control parameters are generated.

5. The AI-based thermal insulation welding optimization method according to claim 4, characterized in that: The dynamic synchronization control weight of each partition is calculated using the formula: ； Calculate, where Representative Group Dynamic synchronization control weight of the partition to which it belongs, represents the synchronization deviation density eigenvalue of group g, represents the average value of the synchronization deviation density eigenvalues ​​of all groups, Representative Group The average absolute deviation of the welding time of all nodes in represents the absolute deviation of the welding time of the mth node in group g, Representative Group The number of nodes in Represents the total number of groups.

6. The AI-based thermal insulation welding optimization method according to claim 1, characterized in that: The specific steps for obtaining the regional thermal stability assessment report are as follows: S311: Based on the regional collaborative control parameters and the temperature data of the corresponding nodes in each time period, extract the temperature values ​​of adjacent nodes in the same area in the horizontal and vertical directions, calculate the temperature difference between the nodes in each direction, record the horizontal and vertical differences in each time period, and establish a directional gradient difference; S312: Based on the horizontal and vertical temperature differences recorded in the directional gradient difference, calculate the temperature change direction in adjacent time periods, count the time proportion of continuous changes in the same direction, calculate the trend consistency degree in the horizontal and vertical directions respectively, and take the average value to obtain the temperature trend consistency index; S313: Based on the temperature trend consistency index, according to the temperature change direction sequence of each node, identify the moment when the change direction of adjacent time periods is reversed, count the number of positive and negative alternations of direction changes per unit time, combine the trend consistency index and the reversal frequency, judge the degree of continuity of heat conduction changes in each region, and obtain a regional thermal stability assessment report.

7. The AI-based thermal insulation welding optimization method according to claim 6, characterized in that: The temperature trend consistency index adopts the formula: ； Calculate, where Representative direction The degree of trend consistency, represents the total number of analysis time periods in direction d, Representative direction Middle The incremental change of the temperature difference in a period relative to the previous period, Indicates that from the second time period to the The cumulative sum operation of the time period, Indicates the The absolute value of the temperature difference during the period, The item is used to determine whether two adjacent time periods change in the same direction. The term is used to introduce the similarity adjustment coefficient of the change amplitude between two adjacent time periods.

8. The AI-based thermal insulation welding optimization method according to claim 1, characterized in that: The specific steps for obtaining the process parameter optimization data set are: S411: Based on the regional thermal stability assessment report, extract the current holding time records of all nodes in each partition, calculate the mean and standard deviation of the holding time of the nodes in each partition, divide the difference between the holding time of a single node and the partition mean by the standard deviation to obtain the time deviation coefficient of each node, summarize the node numbers of each partition and the corresponding deviation coefficients, and establish a time deviation coefficient matrix; S412: Based on the time deviation coefficient matrix, the proportional coefficient, integral coefficient, and differential coefficient in the PID adjustment coefficient are set. The proportional term is multiplied by the deviation value, the integral term is multiplied by the cumulative sum of the current deviation and the previous deviation, and the differential term is multiplied by the difference between the current deviation and the previous deviation. The three terms are added together to obtain the adjustment time value corresponding to each node. The updated time is obtained by adding the three terms to the original holding time of the node, and the time adjustment increment data is obtained by summing them up. S413: Based on the updated holding time records of each node in the time adjustment incremental data, classification and extraction are performed according to the partition number, and the holding time of all nodes in each partition is normalized respectively to form a process parameter optimization data set.

9. The AI-based thermal insulation welding optimization method according to claim 1, characterized in that: The specific steps for obtaining the thermal insulation welding optimization solution are as follows: S511: Based on the process parameter optimization data set and the parameter adjustment status identified in the partition control record, the partitions that have been subjected to parameter adjustment operations are screened, the holding time change direction identifiers of all nodes in each partition are extracted, and whether the adjustment is in the same direction is determined based on the consistency of the direction. The node numbers and partition numbers that meet the same direction adjustment condition are recorded to obtain a same direction adjustment node sequence; S512: Based on each node number in the same-direction adjustment node sequence, the corresponding holding time increment value sequence is retrieved, the area of ​​the time period of continuous same-direction change within a unit time is calculated as the corresponding node cumulative offset value, the cumulative offset values ​​of all nodes in the same partition are accumulated, and the accumulated offset values ​​are classified by partition number to generate a cumulative time deviation; S513: According to each partition number and the corresponding adjustment value in the accumulated time deviation, combined with the node distribution map and the current optimization state, the standardized time adjustment results of each area are matched according to each partition number to obtain the insulation welding optimization plan.

10. AI-based thermal insulation welding optimization system, characterized by: The system is used to implement the AI-based thermal insulation welding optimization method according to any one of claims 1 to 9, and the system includes: The node temperature analysis module obtains continuous temperature records of each welding node during the insulation process, detects extreme points in the curve, records temperature anomalies, and generates node temperature anomaly logs; The area division module is based on the node temperature anomaly log, sorts the nodes according to the node number and the time progress in the welding path, sets the key control nodes to divide the welding area into several process zones, and generates regional collaborative control parameters; The stability assessment module extracts the node temperature gradient difference based on the regional collaborative control parameters, calculates the temperature trend consistency index, determines the degree of heat conduction consistency, and generates a regional thermal stability assessment report; The parameter optimization module determines the node time deviation coefficient within the partition based on the regional thermal stability assessment report, sets the PID adjustment coefficient, calculates the process optimization time based on the deviation, and generates a process parameter optimization data set; The insulation optimization module extracts the partition nodes where the same-direction adjustment occurs based on the process parameter optimization data set, calculates the accumulated time deviation and archives it by partition, and generates an insulation welding optimization plan.

Images