A battery production whole-process data monitoring system and method

Abstract

This invention discloses a data monitoring system and method for the entire battery production process, relating to the field of battery production technology. The system includes: acquiring production process data, process flow records, and quality inspection data corresponding to each process step; determining the starting and ending positions of a production object in each process step based on the flow time periods in the process flow records and the entry and exit time periods in equipment events; and generating the action range of the production object in the corresponding process step. For abnormal production objects, the system extracts the abnormal starting position, abnormal expansion path, and abnormal change intensity according to the production sequence of continuous quality inspection results. This invention first generates action ranges around the starting and ending positions of the production object in each process step, and in scenarios with parallel equipment, parallel workstations, or parallel modules, it confirms the allocation of candidate starting positions, candidate ending positions, candidate action ranges, and candidate occupied ranges through progressive screening and assignment.

Description

Technical Field

[0001] This invention relates to the field of battery manufacturing technology, specifically to a data monitoring system and method for the entire battery manufacturing process. Background Technology

[0002] The battery production process typically includes multiple steps such as coating, rolling, slitting, winding, assembly, formation, capacity testing, and inspection. Different steps correspond to different production equipment, workstations, modules, and quality inspection links. With the improvement of automation and informatization levels in battery production lines, production process data, process flow records, and quality inspection data can be collected and stored to varying degrees, thus providing a data foundation for production monitoring, anomaly analysis, and quality traceability. In existing technologies, the status of a single step or a part of the production process can usually be monitored and analyzed through equipment operation data, process parameter data, and quality inspection results. In multi-process continuous production scenarios, the same production object needs to be processed through multiple processes in sequence. There are usually multiple sources of information between different processes, such as process flow, equipment entry and exit, parallel processing, and subsequent quality inspection. Due to the fast cycle time of battery production, the prevalence of parallel equipment and multi-station structures, the actual starting and ending positions of the production object in different processes cannot always be determined directly based on a single time record. At the same time, abnormal results in the later stages often correspond to the comprehensive results of the combined effects of multiple processes in the earlier or middle stages. Judging solely based on local process records or a single data source is difficult to fully reflect the continuous interaction of the production object throughout the entire process. In existing technologies, there is generally a certain foundation for implementing process-level monitoring, quality analysis, anomaly identification, and responsibility location. For example, production status can be analyzed based on process flow relationships, abnormal trends can be identified based on continuous detection results, traceability analysis can be carried out based on the correlation between preceding and subsequent processes, and the source of anomalies can be located within a certain range. However, in the context of the entire battery production process, there is still room for further refinement and improvement in how to construct cross-process continuous action intervals around the same production object, how to organize the abnormal results of the later stages into an abnormal response structure that can be used for reverse constraints, how to confirm the target action chain among multiple cross-process candidate action interval groups, and how to identify action transmission breakpoints and further converge and locate the target responsibility unit. Therefore, this invention proposes a data monitoring system and method for the entire battery production process. Summary of the Invention

[0003] The purpose of this invention is to provide a data monitoring system and method for the entire battery production process to solve the problems mentioned in the background art.

[0004] This invention can be achieved through the following technical solution: a method for monitoring data throughout the entire battery production process, comprising: S1. Obtain the production process data, process flow records and quality inspection data corresponding to each process, and based on the flow time period in the process flow record and the entry time period and exit time period in the equipment event, determine the start position and end position of the production object in each process, and generate the range of action of the production object in the corresponding process. S2. For abnormal production objects, extract the abnormal starting position, abnormal expansion path and abnormal change intensity according to the production sequence of continuous quality inspection results, and generate the subsequent abnormal response structure based on the sequential connection relationship between the abnormal starting position, abnormal expansion path and abnormal change intensity. S3. Based on the post-stage abnormal response structure, extract the cross-process candidate action interval group corresponding to the abnormal production object from the action interval, and perform continuous correspondence verification according to the process sequence. Retain the cross-process candidate action interval group that satisfies the continuous correspondence relationship between the abnormal start position, abnormal expansion path and abnormal change intensity, and eliminate the remaining cross-process candidate action interval group to determine the target action chain. S4. Based on the target action chain, extract the output state of the previous action interval and the input state of the next action interval between adjacent processes. Construct cross-process action transmission relationship according to the offset amplitude, change direction and continuity between the output state and the input state, and determine the position corresponding to the abrupt change in offset amplitude, reversal of change direction or interruption of continuity as the action transmission breakpoint. S5. Based on the action transmission breakpoint, extract the local action change characteristics of each responsible unit in the corresponding process before and after the action transmission breakpoint, and verify the consistency of the local action change characteristics with the abnormal start position, abnormal expansion path and abnormal change intensity. Gradually eliminate the responsible units that do not meet the consistency conditions to determine the target responsible unit.

[0005] A further technical improvement of the present invention is that: generating the corresponding effective range in step S1 includes: S11. Based on the previous handover period, the current process acceptance period, and the entry and exit periods in the equipment events in the process flow record, extract multiple candidate start positions and multiple candidate end positions of the production object in the corresponding process to form a set of candidate start positions and a set of candidate end positions. S12. Combine each candidate starting position in the candidate starting position set with each candidate ending position in the candidate ending position set to form multiple candidate action intervals, and eliminate candidate action intervals that do not meet the conditions of candidate starting position preceding candidate ending position, preceding continuous handover, and subsequent continuous transfer. S13. For the candidate action intervals that have been screened and retained, compare the consistency between the candidate starting position and the handover sequence of the preceding process, and the correspondence between the candidate ending position and the production sequence of the continuous quality inspection results, and eliminate the candidate action intervals that have inconsistent starting sequences or do not correspond to the ending sequences. S14. In the presence of parallel devices, parallel workstations or parallel modules, compare the interval overlap and occupancy conflict of the remaining candidate action intervals on each parallel processing unit, and eliminate the candidate action intervals that cannot be uniquely attributed to a single parallel processing unit. S15. The candidate action ranges that have been screened and retained are determined as the action ranges of the production object in the corresponding process.

[0006] A further technical improvement of the present invention is that step S14 includes: For each parallel processing unit, the starting position, ending position and duration of the candidate action interval within the corresponding parallel processing unit are extracted, and the occupancy representation of the candidate action interval within the corresponding parallel processing unit is defined as the candidate occupancy interval, forming a set of candidate occupancy intervals for each parallel processing unit. In each set of candidate occupancy intervals, the occupancy overlap relationship between different candidate occupancy intervals within the same occupancy duration segment is identified, and based on the occupancy exclusivity constraint that only a single production object is allowed to continuously occupy the same occupancy duration segment within the same parallel processing unit, candidate occupancy intervals with occupancy overlap conflicts are eliminated. Identify the processing switching positions between adjacent parallel processing units, and construct switching boundary isolation segments with each processing switching position as the center. Compare the cross relationship between the remaining candidate occupancy intervals and the switching boundary isolation segments, and eliminate candidate occupancy intervals that cross the switching boundary isolation segments and cannot maintain occupancy continuity. For the candidate occupancy intervals retained after occupancy exclusion constraints and switching boundary isolation screening, compare their forward closure relationship with the handover sequence of the preceding process, their intermediate closure relationship with the current process acceptance sequence, and their backward closure relationship with the production sequence of continuous quality inspection results. Eliminate candidate occupancy intervals that cannot simultaneously satisfy the forward closure relationship, intermediate closure relationship, and backward closure relationship. Candidate occupancy intervals that have been screened and retained and can form a unique occupancy relationship and a unique closure relationship within a single parallel processing unit are determined as the unique ownership intervals of the production object in the corresponding process, and the corresponding parallel processing unit is determined as the unique ownership processing unit of the production object in the corresponding process.

[0007] A further technical improvement of the present invention is that: the generation of the subsequent abnormal response structure in step S2 includes: According to the production sequence of continuous quality inspection results, the abnormal change sequence of abnormal production objects in the detection sequence is extracted, and the abnormal change intensity sequence and abnormal cluster density sequence corresponding to the production sequence are formed based on the abnormal change sequence. Identify abrupt changes in the gradient of abrupt changes in intensity in the abrupt change sequence and abrupt changes in abrupt cluster density in the abrupt cluster density sequence. Determine the overlapping or adjacent positions of the abrupt changes in intensity gradient and abrupt changes in abrupt cluster density as abrupt evolution inflection points. The detection section between adjacent abnormal evolution inflection points is used as the abnormal response segment. The continuous quality detection results are segmented to form multiple abnormal response segments connected sequentially according to the production order. For each abnormal response segment, the local abnormality start position, local abnormality expansion path, and local abnormality change intensity are extracted respectively. The local abnormality expansion path continuation relationship and local abnormality change intensity change relationship between adjacent abnormal response segments are compared, and abnormal response segments that do not meet the path continuation condition and intensity change condition are eliminated. The selected and retained abnormal response segments are connected in production order, and the local abnormality start position, local abnormality expansion path and local abnormality change intensity in each abnormal response segment are combined to generate the subsequent abnormal response structure.

[0008] A further technical improvement of the present invention is that: step S3 involves performing continuous correspondence verification according to the process sequence, including: For each cross-process candidate action interval group corresponding to the abnormal production object, the output state of the previous candidate action interval and the input state of the next candidate action interval are extracted according to the process sequence to form the state transmission sequence corresponding to each cross-process candidate action interval group. Compare the state continuity, state offset, and state succession relationships between adjacent output and input states in each state transmission sequence, and eliminate cross-process candidate action interval groups that have state transmission interruption, state offset reversal, or state succession mismatch. The cross-process candidate action interval groups retained after state transmission screening are mapped to the anomaly start position, anomaly propagation path and anomaly change intensity in the subsequent anomaly response structure according to the process sequence, forming the anomaly interpretation sequence corresponding to each cross-process candidate action interval group. Compare the continuous correspondence between each anomaly interpretation sequence and the subsequent anomaly response structure in terms of anomaly initiation position, anomaly propagation path and anomaly change intensity, and eliminate cross-process candidate action interval groups that cannot form a continuous explanation for the subsequent anomaly response structure. For the cross-process candidate action interval groups retained after anomaly interpretation screening, their closure relationships at the three levels of previous input, current acceptance, and subsequent output are further compared, and cross-process candidate action interval groups that cannot form a unique closure relationship are eliminated. The cross-process candidate action interval group that has been screened and retained and simultaneously satisfies the closed state transmission, continuous anomaly interpretation, and unique closed relationship is determined as the target action chain corresponding to the abnormal production object.

[0009] A further technical improvement of the present invention is that: determining the target action chain in step S3 includes: The candidate action intervals corresponding to the abnormal production object are constructed into a cross-process candidate transfer graph according to the process sequence. Each candidate action interval is used as a graph node, and graph edges are established between candidate action intervals that meet the acceptance conditions between adjacent processes. For each cross-process candidate action interval group, the corresponding state transmission constraint value is calculated based on the output state of the previous candidate action interval and the input state of the next candidate action interval between adjacent candidate action intervals. Map the anomaly in the subsequent anomaly response structure to each cross-process candidate action interval group, and calculate the corresponding anomaly interpretation constraint value. Based on the parallel processing unit occupancy conflict relationship, interval overlap relationship and attribution conflict relationship, calculate the conflict suppression constraint value of each cross-process candidate action interval group; Based on the state transmission constraint value, anomaly interpretation constraint value, and conflict suppression constraint value, calculate the closed transmission value of each cross-process candidate action interval group, and remove cross-process candidate action interval groups whose closed transmission value is lower than the preset closed threshold. The cross-process candidate action interval group retained after screening by closed transitive values ​​is normalized, and the initial uncertainty value of the candidate set is calculated based on the normalization result. For each selected and retained cross-process candidate action interval group, other cross-process candidate action interval groups that conflict with them in terms of parallel processing unit affiliation and process succession relationship are removed. The uncertainty value of the remaining candidate set is recalculated, and the corresponding uncertainty collapse degree is calculated based on the difference between the initial uncertainty value and the recalculated uncertainty value. The cross-process candidate action interval group that simultaneously satisfies the condition that the closed transmission value is not lower than the preset closed threshold and has the largest uncertainty collapse degree is determined as the target action chain corresponding to the abnormal production object.

[0010] A further technical improvement of the present invention is that: determining the action transmission breakpoint in step S4 includes: For the output state of the previous action interval and the input state of the next action interval between adjacent processes in the target action chain, extract the state offset, state direction change and state continuation interruption at the corresponding positions to form the corresponding coupling mismatch feature group for each adjacent process pair. By comparing the co-occurrence relationships of state offset, state direction change, and state continuation interruption in each coupling mismatch feature group at the same transmission position, candidate breakpoint positions that simultaneously satisfy offset mismatch, direction reversal mismatch, and continuation interruption mismatch are identified. For each candidate breakpoint, the state recovery amount, state direction recovery amount and state continuation recovery amount of the corresponding position in the subsequent process are extracted. The recovery relationship of each candidate breakpoint in the subsequent process is compared, and candidate breakpoints that can recover state offset, state direction and state continuation within the preset subsequent process range are eliminated. For the candidate breakpoint positions retained after the recovery relationship screening, compare their closure status in the preceding output relationship, current input relationship and subsequent succession relationship respectively, and remove the candidate breakpoint positions that can still form the preceding output closure, current input closure and subsequent succession closure; Candidate breakpoints that have been screened and retained and simultaneously satisfy the conditions of coupling mismatch, recovery failure, and unique closure failure are determined as action propagation breakpoints.

[0011] A further technical improvement of the present invention is that: the determination of the target responsibility unit in step S5 includes: Based on the abnormal propagation path in the subsequent abnormal response structure, the corresponding local action change section is pointed back in the reverse direction according to the production sequence to form a candidate set of responsibility units. For each responsible unit in the candidate set of responsible units, the initial consistency relationship between its local action change characteristics and the anomaly start position is compared. Responsible units that do not meet the anomaly start position consistency are eliminated to obtain the first candidate set of responsible units. For each responsibility unit in the first responsibility unit candidate set, compare the propagation consistency relationship between its local role change characteristics and the abnormal propagation path, and eliminate responsibility units that do not meet the consistency of the abnormal propagation path to obtain the second responsibility unit candidate set. For each responsibility unit in the candidate set of the second responsibility unit, the intensity consistency relationship between its local action change characteristics and the intensity of abnormal change is compared. Responsibility units that do not meet the intensity consistency of abnormal change are eliminated to obtain the candidate set of the third responsibility unit. After each round of responsibility unit screening, the uncertainty value of the corresponding candidate set of responsibility units is calculated, and the degree of uncertainty collapse is determined based on the difference in uncertainty values ​​between the candidate sets of responsibility units in two adjacent rounds. The responsible unit that simultaneously satisfies the consistency of the anomaly's starting location, the anomaly's expansion path, and the anomaly's change intensity, and corresponds to the largest degree of uncertainty collapse, is identified as the target responsible unit.

[0012] On the other hand, the present invention also discloses a data monitoring system for the entire battery production process, which employs any of the above-mentioned data monitoring methods.

[0013] Compared with the prior art, the present invention has the following beneficial effects: This invention first generates an action range based on the starting and ending positions of the production object in each process. In scenarios with parallel equipment, parallel workstations, or parallel modules, it improves the accuracy of determining the action range of the production object in the corresponding process by progressively screening and confirming candidate starting positions, candidate ending positions, candidate action ranges, and candidate occupied ranges. This provides clearer data boundaries for subsequent cross-process data organization and continuous action relationship analysis, and helps improve the consistency of the monitoring objects in the production process. Furthermore, this invention generates a subsequent abnormal response structure based on continuous quality inspection results. On this basis, it extracts cross-process candidate action interval groups corresponding to abnormal production objects from the action interval. Through state transmission verification, abnormal interpretation verification, and target action chain confirmation, it constructs the continuous action relationship of production objects in multiple processes. This enables a closer correspondence between subsequent abnormal results and the continuous production action process from the preceding to the middle stage, so that abnormal identification is not limited to a single process or a single parameter level, but can reflect the characteristics of continuous transmission across processes. On the other hand, this invention also analyzes the action transmission relationship between adjacent processes based on the target action chain, identifies action transmission breakpoints, and combines the consistency of local action change characteristics with the anomaly initiation position, anomaly propagation path, and anomaly change intensity to gradually converge and locate the responsible unit. Through this technical solution, anomaly identification, action chain confirmation, breakpoint identification, and responsibility location can be correlated in the entire battery production process, which helps to improve the pertinence of anomaly tracing and responsibility location, and provides richer technical support for battery production process monitoring, quality analysis, and production management. Attached Figure Description

[0014] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0015] Figure 1 This is a schematic diagram of the data monitoring method in this invention. Detailed Implementation

[0016] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0017] Please see Figure 1 As shown, this invention provides a method for monitoring data throughout the entire battery production process, including: S1. Obtain the production process data, process flow records and quality inspection data corresponding to each process, and based on the flow time period in the process flow record and the entry time period and exit time period in the equipment event, determine the start position and end position of the production object in each process, and generate the range of action of the production object in the corresponding process. Specifically, in the coating process, the production process data includes at least tension, speed, and temperature values, with a collection interval set to 1 second; the process flow record includes at least the preceding handover period, the current process acceptance period, and the subsequent transfer-out period; equipment events include at least the entry period and exit period; quality inspection data includes at least the production sequence and corresponding inspection values ​​of continuous quality inspection results, wherein the production sequence of continuous quality inspection results is represented by inspection numbers in this embodiment. Taking production object numbered 202603100008 as an example, the preceding handover period is recorded as 08:15:12 to 08:15:18, the current process acceptance period is recorded as 08:15:14 to 08:15:20, the entry period is recorded as 08:15:15, the exit period is recorded as 08:19:46, and the subsequent transfer-out period is recorded as 08:19:44 to 08:19:50. During the aforementioned period, the corresponding production process data showed the following: tension values ​​continuously varied between 18.2 N and 18.6 N; velocity values ​​continuously varied between 42.0 m / min and 42.5 m / min; and temperature values ​​continuously varied between 78.4 °C and 79.1 °C. Based on these continuously changing production process data, it can be further verified that the production object was under continuous action during this period, thus providing process support for subsequently determining the start and end points of the action.

[0018] Secondly, based on the preceding handover period, the current process acceptance period, and the entry and exit periods in the equipment events in the process flow record, multiple candidate start positions and multiple candidate end positions of the production object in the corresponding process are extracted to form a candidate start position set and a candidate end position set, where the candidate start position and candidate end position are corresponding time positions. Then, each candidate start position in the candidate start position set is combined with each candidate end position in the candidate end position set to form multiple candidate action intervals, and candidate action intervals that do not meet the conditions of candidate start position preceding candidate end position, preceding handover continuity, and subsequent transfer-out continuity are eliminated. In this embodiment, the rule for determining whether a candidate starting position precedes a candidate ending position is: the starting time position is at least 10 seconds earlier than the ending time position; the rule for determining the continuity of the preceding handover condition is: the time deviation between the candidate starting position and the preceding handover period does not exceed 6 seconds, and the tension value fluctuation amplitude does not exceed 0.8N, the velocity value fluctuation amplitude does not exceed 1.0m / min, and the temperature value fluctuation amplitude does not exceed 1.0℃ within 3 consecutive seconds from the candidate starting position; the rule for determining the continuity of the subsequent turnout condition is: the time deviation between the candidate ending position and the subsequent turnout period does not exceed 6 seconds, and the tension value, velocity value, and temperature value remain monotonically continuous without any abrupt changes within 3 consecutive seconds before the candidate ending position. According to the above rules, candidate action intervals that start at 08:15:20 and end at 08:19:44 but do not meet the continuity of the preceding handover condition are eliminated, and candidate action intervals that meet the above conditions are retained for subsequent verification.

[0019] After obtaining the candidate action intervals through the above screening, the candidate action intervals are subjected to sequence constraint verification with the handover sequence of the preceding process and the production sequence of continuous quality inspection results to further complete S13. In this embodiment, the handover sequence of the preceding process is represented by the actual handover number sequence. Let the handover sequence number of the production object in the preceding process be 127, and the handover sequence numbers of adjacent production objects in the same batch be 126 and 128, respectively. Then, the consistency judgment method for the candidate starting position is: the order of the candidate starting position on the time axis should be consistent with the order of the handover sequence number of the preceding process; otherwise, it is considered that the starting order is inconsistent. In this embodiment, the production sequence of continuous quality inspection results is represented by the inspection number. Let the inspection number of the production object in the subsequent inspection be 309. Then, the correspondence judgment method for the candidate ending position is: project the candidate ending position onto the production sequence of continuous quality inspection results. If the projected position deviates from the actual inspection number by more than 2 positions, it is considered that the ending order is not corresponding. According to this determination method, the candidate action interval corresponding to 08:15:18 to 08:19:50 is eliminated because its termination position is projected onto detection number 313, which deviates from the actual detection number by 4 positions. The candidate action interval corresponding to 08:15:15 to 08:19:46 is retained because its candidate starting position order is consistent with the handover order of the preceding process, and its candidate termination position is projected onto detection number 309. Therefore, in the absence of interference from parallel processing units, 08:15:15 to 08:19:46 can be determined as the action interval of the production object in the corresponding process.

[0020] When parallel equipment, parallel workstations, or parallel modules exist within the same process, the overlap and conflict relationships of the remaining candidate action intervals in each parallel processing unit are compared, and candidate action intervals that cannot be uniquely attributed to a single parallel processing unit are eliminated. In this embodiment, it is assumed that there are 3 parallel processing units in the rolling process, corresponding to the 1st rolling position, the 2nd rolling position, and the 3rd rolling position, respectively. For each parallel processing unit, the starting position, ending position, and duration of the occupation of the candidate action interval within the corresponding parallel processing unit are extracted, and the occupation representation of the candidate action interval within the corresponding parallel processing unit is defined as the candidate occupation interval, forming a set of candidate occupation intervals corresponding to each parallel processing unit. For example, the candidate occupation interval corresponding to the 1st rolling position is from 09:02:11 to 09:05:40, and the candidate occupation interval corresponding to the 2nd rolling position overlaps and conflicts with the adjacent production object numbered 202603100009 in the interval from 09:02:18 to 09:05:39, so it is eliminated. The rule for determining overlapping occupancy conflicts is as follows: if two candidate occupancy intervals overlap for more than 5 seconds on the same parallel processing unit and correspond to different production objects, they are considered to be in an overlapping occupancy conflict. Next, the processing switchover positions between adjacent parallel processing units are identified, and switchover boundary isolation sections are constructed centered on each switchover position. If a candidate occupancy interval crosses a switchover boundary isolation section and its occupancy duration cannot form continuous time coverage on both sides of the switchover boundary isolation section, it is considered that occupancy continuity cannot be maintained and is eliminated. Subsequently, for the candidate occupancy intervals retained after occupancy exclusivity constraints and switchover boundary isolation screening, their forward closure relationship with the handover sequence of the preceding process, their intermediate closure relationship with the current process's acceptance sequence, and their backward closure relationship with the production sequence of continuous quality inspection results are compared. The rules for determining forward closure are as follows: the deviation between the starting position of the candidate occupied interval and the handover sequence of the preceding process does not exceed one sequence. The rules for determining intermediate closure are as follows: the proportion of the candidate occupied interval's continuous occupied section completely covering the corresponding processing section in the current process is not less than 90%. The rules for determining backward closure are as follows: the deviation between the ending position of the candidate occupied interval and the production sequence of the continuous quality inspection results does not exceed two sequences. After comparison, the candidate occupied interval corresponding to the first roller pressing position simultaneously satisfies forward, intermediate, and backward closure relationships. While the candidate occupied interval corresponding to the third roller pressing position satisfies both forward and intermediate closure relationships, its backward closure deviation reaches three sequences, therefore it is eliminated. Finally, the candidate occupied intervals that have been screened and retained and can form a unique occupied relationship and a unique closure relationship within a single parallel processing unit are determined as the unique assigned function intervals of the production object in the corresponding process, and the corresponding parallel processing unit is determined as the unique assigned processing unit of the production object in the corresponding process.In this embodiment, the final determined unique attribution range is from 09:02:11 to 09:05:40, and the corresponding unique attribution processing unit is the first roller pressing position.

[0021] S2. For abnormal production objects, extract the abnormal starting position, abnormal expansion path and abnormal change intensity according to the production sequence of continuous quality inspection results, and generate the subsequent abnormal response structure based on the sequential connection relationship between the abnormal starting position, abnormal expansion path and abnormal change intensity. Specifically, in this embodiment, an abnormal production object refers to a production object whose corresponding quality inspection data meets the preset abnormality judgment condition in the continuous quality inspection results; the preset abnormality judgment condition may be that the abnormal change value at the corresponding inspection number is not lower than the preset abnormality threshold, or that the abnormal change values ​​at multiple consecutive inspection numbers are not lower than the preset abnormality threshold.

[0022] The production sequence of continuous quality inspection results is represented by inspection numbers. Let's assume the inspection number range for a given abnormal production item is 301 to 340. Each inspection number corresponds to a capacity deviation value, an internal resistance deviation value, and an appearance defect level value. To form an abnormal change sequence, the capacity anomaly value, internal resistance anomaly value, and normalized appearance defect level value are first calculated separately. The capacity anomaly value is defined as the absolute value of the difference between the actual capacity and the nominal capacity divided by the nominal capacity; the internal resistance anomaly value is defined as the absolute value of the difference between the actual internal resistance and the reference internal resistance divided by the reference internal resistance; and the normalized appearance defect level value is defined as the appearance defect level value divided by the highest allowable defect level value for that process. Then, the capacity anomaly value, internal resistance anomaly value, and normalized appearance defect level value are weighted and summed with weights of 0.4, 0.4, and 0.2, respectively, to obtain the abnormal change value at the corresponding inspection number, thus forming the abnormal change sequence. For example, the capacity anomaly value corresponding to inspection number 309 is 0.038, the internal resistance anomaly value is 0.041, the normalized appearance defect level value is 0.20, and the corresponding abnormal change value is 0.0716. After calculating the abnormal change sequences for detection numbers 301 to 340 sequentially, an abnormal change sequence was obtained. Following the production order of continuous quality inspection results, the abnormal change sequences of abnormal production objects within the detection sequence were extracted. Based on these abnormal change sequences, an abnormal change intensity sequence and an abnormal cluster density sequence corresponding to the production order were formed. The abnormal change intensity sequence directly takes the abnormal change value corresponding to each detection number in the abnormal change sequence. The abnormal cluster density sequence is generated using a sliding window counting method, with a window length of 5 detection numbers and a step size of 1 detection number. Within each window, the number of detection numbers with an abnormal change value not less than 0.060 is counted, and this number is divided by the window length to obtain the abnormal cluster density.

[0023] After obtaining the abnormal change intensity sequence and the abnormal cluster density sequence, the abnormal change intensity gradient abrupt change locations are identified in the abnormal change intensity sequence, and the abnormal cluster density change locations are identified in the abnormal cluster density sequence. The overlapping or adjacent locations of the abnormal change intensity gradient abrupt change locations and the abnormal cluster density change locations are determined as abnormal evolution inflection points. In this embodiment, the identification method for abnormal change intensity gradient abrupt change locations is as follows: the difference between the abnormal change values ​​of adjacent detection numbers in the abnormal change intensity sequence is used to obtain the gradient sequence; if the absolute value of the gradient corresponding to a certain detection number is not less than 0.018, and the absolute value of the gradient is greater than 1.5 times the average of the absolute values ​​of the gradients corresponding to the two detection numbers before and after it, then the detection number is determined as the abnormal change intensity gradient abrupt change location. The identification method for abnormal cluster density change locations is as follows: if the absolute value of the difference between the abnormal cluster density of the window corresponding to a certain detection number and the abnormal cluster density of the previous window is not less than 0.20, then the detection number is determined as the abnormal cluster density change location. To standardize the rules for identifying inflection points in abnormal evolution, this embodiment stipulates that: when the location of a sudden change in the intensity gradient of abnormal changes coincides with the location of a change in abnormal aggregation density, the overlapping detection number is determined as the inflection point; when the location of a sudden change in the intensity gradient of abnormal changes is adjacent to the location of a change in abnormal aggregation density, the smaller number is determined as the inflection point. Using the above method, detection numbers 307, 313, and 327 can be identified as inflection points in abnormal evolution.

[0024] After identifying the inflection points of abnormal evolution, the detection segments between adjacent inflection points are used as abnormal response segments. Continuous quality detection results are then segmented to form multiple abnormal response segments connected sequentially according to production order. In this embodiment, based on the three inflection points of abnormal evolution (detection numbers 307, 313, and 327), detection numbers 301 to 307 are divided into the first abnormal response segment, detection numbers 308 to 313 into the second abnormal response segment, detection numbers 314 to 327 into the third abnormal response segment, and detection numbers 328 to 340 into the fourth abnormal response segment. Subsequently, for each abnormal response segment, the local abnormality start position, local abnormality expansion path, and local abnormality change intensity are extracted. The continuity relationship of the local abnormality expansion path and the variation relationship of the local abnormality change intensity between adjacent abnormal response segments are compared, and abnormal response segments that do not meet the path continuity condition and intensity variation condition are eliminated. In this embodiment, to unify the statistical boundaries of paths and intensities, the starting position of a local anomaly is defined as the detection number of the first anomaly change value in the corresponding anomaly response segment that is not less than 0.050; the local anomaly expansion path is defined as the sequence of consecutive detection numbers within the corresponding anomaly response segment, starting from the starting position of the local anomaly and ending at the detection number where the anomaly change value in that segment reaches its maximum value; and the local anomaly change intensity is defined as the average value of the anomaly change values ​​of the detection numbers corresponding to the local anomaly expansion path. For example, the second anomaly response segment corresponds to detection numbers 308 to 313, its starting position of the local anomaly is 309, its local anomaly expansion path is 309 to 313, and its local anomaly change intensity is 0.071. Similarly, the local anomaly start position of the first anomaly response segment is 304, the local anomaly expansion path is from 304 to 307, and the local anomaly change intensity is 0.055; the local anomaly start position of the third anomaly response segment is 315, the local anomaly expansion path is from 315 to 327, and the local anomaly change intensity is 0.067; the local anomaly start position of the fourth anomaly response segment is 330, the local anomaly expansion path is from 330 to 334, and the local anomaly change intensity is 0.028. The rules for determining the continuation relationship of local anomaly propagation paths between adjacent anomaly response segments are as follows: the number interval between the starting position of the local anomaly in the subsequent anomaly response segment and the ending number of the local anomaly propagation path in the preceding anomaly response segment does not exceed two detection numbers, and the absolute value of the difference between the anomaly change value at the end of the local anomaly propagation path in the preceding anomaly response segment and the anomaly change value at the starting position of the local anomaly in the subsequent anomaly response segment does not exceed 0.020; the rules for determining the changing relationship of local anomaly change intensity are as follows: the ratio of the local anomaly change intensity of the subsequent anomaly response segment to the local anomaly change intensity of the preceding anomaly response segment is within the range of 0.70 to 1.40. If any of the above conditions are not met, the corresponding anomaly response segment is removed.In this embodiment, the fourth abnormal response segment is excluded because the ratio of its local abnormal change intensity of 0.028 to that of the third abnormal response segment of 0.067 is 0.42, which is lower than 0.70.

[0025] After selecting and retaining multiple abnormal response segments, the selected and retained abnormal response segments are connected in production order, and the local abnormality start position, local abnormality expansion path, and local abnormality change intensity of each abnormal response segment are combined to generate the subsequent abnormal response structure. In this embodiment, the retained first, second, and third abnormal response segments are connected sequentially according to the detection number to form a complete subsequent abnormal response structure. Specifically, the abnormality start position in the subsequent abnormal response structure is formed by combining the local abnormality start positions of each abnormal response segment in production order, which is 304-309-315; the abnormality expansion path in the subsequent abnormal response structure is formed by combining the local abnormality expansion paths of each abnormal response segment in production order, which is 304 to 307, 309 to 313, and 315 to 327; and the abnormality change intensity in the subsequent abnormal response structure is formed by combining the local abnormality change intensity of each abnormal response segment in production order, which is 0.055-0.071-0.067. The above method can transform the discrete detection values ​​in the original continuous quality inspection results into a subsequent abnormal response structure with the sequential connection relationship of abnormal starting position, abnormal expansion path and abnormal change intensity, providing a structured basis for the continuous corresponding verification of subsequent cross-process candidate action interval groups.

[0026] S3. Based on the post-stage abnormal response structure, extract the cross-process candidate action interval group corresponding to the abnormal production object from the action interval, and perform continuous correspondence verification according to the process sequence. Retain the cross-process candidate action interval group that satisfies the continuous correspondence relationship between the abnormal start position, abnormal expansion path and abnormal change intensity, and eliminate the remaining cross-process candidate action interval group to determine the target action chain. Specifically, in this embodiment, three sets of cross-process candidate action intervals corresponding to the abnormal production object are first extracted from the action intervals. The first set of cross-process candidate action intervals includes: the action intervals for the coating process (08:15:15 to 08:19:46), the action intervals for the rolling process (09:02:11 to 09:05:40), the action intervals for the slitting process (09:28:04 to 09:31:18), and the action intervals for the winding process (10:06:20 to 10:10:02). The second and third sets of cross-process candidate action intervals are the remaining two sets of candidate groups to be verified. Subsequently, continuous corresponding verification is performed according to the process sequence, including: for each cross-process candidate action interval group corresponding to the abnormal production object, extracting the output state of the previous candidate action interval and the input state of the next candidate action interval between adjacent candidate action intervals according to the process sequence, forming a state transmission sequence corresponding to each cross-process candidate action interval group. Both the output and input states consist of three state components: thickness deviation, tension deviation, and temperature deviation. The thickness deviation is defined as the difference between the actual thickness and the reference thickness divided by the reference thickness. The tension deviation is defined as the difference between the actual tension and the reference tension divided by the reference tension. The temperature deviation is defined as the difference between the actual temperature and the reference temperature divided by the reference temperature. Taking the first group of cross-process candidate action intervals as an example, the average output state of the last 5 seconds of the coating process action interval is (0.036, 0.028, 0.011), the average input state of the first 5 seconds of the rolling process action interval is (0.040, 0.031, 0.010); the average output state of the last 5 seconds of the rolling process action interval is (0.048, 0.034, 0.012), the average input state of the first 5 seconds of the slitting process action interval is (0.051, 0.036, 0.013); the average output state of the last 5 seconds of the slitting process action interval is (0.057, 0.041, 0.014), and the average input state of the first 5 seconds of the winding process action interval is (0.061, 0.043, 0.015), thus forming the state transmission sequence corresponding to the first group of cross-process candidate action intervals.

[0027] After the state transmission sequences corresponding to each cross-process candidate action interval group have been formed, the state continuity, state offset, and state succession relationships between adjacent output and input states in each state transmission sequence are compared. Cross-process candidate action interval groups with state transmission interruptions, state offset reversals, or state succession mismatches are eliminated. Specifically, the state continuity relationship is determined by averaging the absolute values ​​of the differences between the corresponding state components of the output and input states in the same adjacent process pair. If the average difference does not exceed 0.010, the state continuity relationship is considered valid. The state offset relationship is determined by comparing the offset direction of the output state of the previous candidate action interval relative to the reference state of that process with the offset direction of the input state of the next candidate action interval relative to the reference state of the next process. If at least two of the three state components maintain the same offset direction, the state offset relationship is considered valid. The state succession relationship is determined by the following: if the average value of the input state of the next candidate action interval is continuous with the average value of the output state of the previous candidate action interval in both time sequence and numerical trend, and the interval between them does not exceed 120 seconds, the state succession relationship is considered valid. Taking the first group of cross-process candidate action intervals as an example, the average difference between the output state of the coating process and the input state of the rolling process is 0.0027, the offset direction is consistent, and the action interval is 42s. Therefore, the state continuation relationship, state offset relationship, and state succession relationship are all valid. The rolling process and the slitting process, and the slitting process and the winding process also satisfy the above conditions respectively. After comparison, the second group of cross-process candidate action intervals was eliminated due to the existence of state transmission interruption and state offset reversal, and the third group of cross-process candidate action intervals was eliminated due to the existence of state succession mismatch.

[0028] Subsequently, for the first group of cross-process candidate action intervals retained after state transmission screening, the anomaly starting position, anomaly propagation path, and anomaly change intensity are mapped to the subsequent anomaly response structure according to the process sequence, forming a corresponding anomaly interpretation sequence. Mapping is performed using a one-to-one correspondence between the carrier number formed at the end of the winding process for the same production object and the carrier number at the entrance of the subsequent inspection process. If the carrier number at the end of the winding process is 5209, and the corresponding inspection number at the entrance of the subsequent inspection process is 305, then inspection number 305 is used as the base point for the anomaly starting position of the first group of cross-process candidate action intervals. The rule for extracting the anomaly propagation path is as follows: first, the thickness deviation value, tension deviation value, and temperature deviation value in the output state of each process are judged to be enhanced in the same direction. If at least two of the three state components in the same continuous segment monotonically increase with the process sequence, then the segment is identified as a continuously enhanced segment; if the interval between continuously enhanced segments does not exceed one process node, then they are merged into the same anomaly propagation path. The extraction rule for the intensity of abnormal changes is as follows: For each continuously reinforced section, calculate the average absolute value of the deviation of the output state of each process within that section from its respective reference state. Then, sum the values ​​by weighting the thickness deviation, tension deviation, and temperature deviation by 0.4, 0.4, and 0.2, respectively, to obtain the intensity of abnormal changes corresponding to that continuously reinforced section. If multiple continuously reinforced sections exist, each section will form an abnormal change intensity corresponding to an abnormal expansion path. According to this rule, the abnormal interpretation sequence formed by the first group of cross-process candidate action intervals is: abnormal starting position 305-309-314, abnormal expansion path 305 to 307, 309 to 313, 314 to 327, and abnormal change intensity 0.057-0.070-0.068, where the abnormal change intensity here is the interpretation value of the subsequent abnormal response structure.

[0029] After an anomaly interpretation sequence has been formed, the continuous correspondence between each anomaly interpretation sequence and the subsequent anomaly response structure in terms of anomaly initiation position, anomaly propagation path, and anomaly change intensity is compared. Cross-process candidate action interval groups that cannot form a continuous explanation for the subsequent anomaly response structure are eliminated. For the cross-process candidate action interval groups retained after anomaly interpretation screening, their closure relationships at the three levels of preceding input, current acceptance, and subsequent output are further compared, and cross-process candidate action interval groups that cannot form a unique closure relationship are eliminated. Specifically, the rules for determining the continuous correspondence of anomaly initiation position are: the numbering deviation between the anomaly initiation position in the anomaly interpretation sequence and the anomaly initiation position in the subsequent anomaly response structure does not exceed one detection number; the rules for determining the continuous correspondence of anomaly propagation path are: the overlap ratio between the anomaly propagation path segment in the anomaly interpretation sequence and the anomaly propagation path segment in the subsequent anomaly response structure is not less than 80%; and the rules for determining the continuous correspondence of anomaly change intensity are: the absolute value of the difference between the anomaly change intensity in the anomaly interpretation sequence and the anomaly change intensity in the subsequent anomaly response structure does not exceed 0.010. According to the above rules, in the first group of cross-process candidate action intervals, the deviations of the abnormal starting position are 1, 0, and 1, respectively; the overlap ratios of the abnormal expansion path segments are 75%, 100%, and 92.3%, respectively. If the overlap ratio of any single segment is less than 80%, the overall overlap ratio is further calculated. As long as the overall overlap ratio is not less than 85%, the continuous correspondence of the abnormal expansion path is still considered to be established. In this embodiment, the overall overlap ratio is 88.2%; the abnormal change intensity difference is 0.002, 0.001, and 0.001, respectively, all of which meet the conditions. The rules for determining the closure relationship at the three levels of preceding input, current acceptance, and subsequent output are as follows: the closure relationship of preceding input requires that the input order of the coating process deviates from the previous material handover order by no more than 1 order; the closure relationship of current acceptance requires that the time interval between the action intervals of each intermediate process does not exceed 120s and the state acceptance relationship is established; the closure relationship of subsequent output requires that the deviation of the production order after mapping the end carrier number of the winding process to the continuous quality inspection results does not exceed 2 inspection numbers. The first group of cross-process candidate action intervals is valid at all three levels mentioned above, and is therefore identified as the target action chain retention candidate group.

[0030] After the above continuous correspondence verification, the candidate action interval groups corresponding to the abnormal production objects are constructed into a cross-process candidate transfer graph according to the process sequence. Each candidate action interval serves as a graph node, and graph edges are established between candidate action intervals that meet the acceptance conditions between adjacent processes. For each cross-process candidate action interval group, the corresponding state transfer constraint value is calculated based on the output state of the previous candidate action interval and the input state of the next candidate action interval between adjacent candidate action intervals. The abnormal start position, abnormal expansion path, and abnormal change intensity in the subsequent abnormal response structure are mapped to each cross-process candidate action interval group, and the corresponding abnormal interpretation constraint value is calculated. Based on the parallel processing unit occupancy conflict relationship, interval overlap relationship, and attribution conflict relationship, the conflict suppression constraint value of each cross-process candidate action interval group is calculated. Based on the state transfer constraint value, abnormal interpretation constraint value, and conflict suppression constraint value... The process involves calculating the closed-loop transfer value of each cross-process candidate action interval group and eliminating cross-process candidate action interval groups whose closed-loop transfer values ​​are lower than a preset closure threshold. The remaining cross-process candidate action interval groups after closure transfer value screening are normalized, and the initial uncertainty value of the candidate set is calculated based on the normalization result. For each screened and retained cross-process candidate action interval group, other cross-process candidate action interval groups that conflict with it in terms of parallel processing unit occupancy conflict, interval overlap, attribution conflict, and process acceptance conflict are eliminated. The uncertainty value of the remaining candidate set is recalculated, and the corresponding uncertainty collapse degree is calculated based on the difference between the initial uncertainty value and the recalculated uncertainty value. The cross-process candidate action interval group that simultaneously satisfies a closed-loop transfer value not lower than the preset closure threshold and has the largest uncertainty collapse degree is determined as the target action chain corresponding to the abnormal production object. In this embodiment, the cross-process candidate transfer graph has 12 graph nodes, and 9 graph edges are established between each graph node. The state propagation constraint values ​​for groups 1, 2, and 3 are 0.864, 0.721, and 0.688, respectively; the anomaly interpretation constraint values ​​are 0.903, 0.664, and 0.592, respectively; and the conflict suppression constraint values ​​are 0.112, 0.247, and 0.331, respectively. Based on these values, the closure propagation values ​​are 1.655, 1.138, and 0.949, respectively. In this embodiment, the closure threshold is specifically set to 1.000, therefore group 3 is removed. The remaining groups 1 and 2 are normalized, with normalization results of 0.593 and 0.407, respectively. Based on this, the initial uncertainty value of the candidate set is calculated to be 0.676.Subsequently, for the first group of cross-process candidate action intervals, the second group of cross-process candidate action intervals that conflicted with it was removed, and the uncertainty value of the remaining candidate set was recalculated to be 0, so the uncertainty collapse degree was 0.676. If the second group was retained and the first group that conflicted with it was removed, the uncertainty value obtained by recalculation was also 0. However, since the closed-loop transit value of the second group (1.138) was lower than that of the first group (1.655), the first group of cross-process candidate action intervals was finally determined as the target action chain corresponding to the abnormal production object. S4. Based on the target action chain, the output state of the previous action interval and the input state of the next action interval between adjacent processes were extracted. The cross-process action transmission relationship was constructed according to the offset amplitude, change direction, and continuity between the output state and the input state, and the positions corresponding to the abrupt change in offset amplitude, reversal of change direction, or interruption of continuity were determined as action transmission breakpoints. Specifically, in this embodiment, based on the aforementioned determined target action chain, the target action chain sequentially includes the action intervals of the coating process (08:15:15 to 08:19:46), the action intervals of the rolling process (09:02:11 to 09:05:40), the action intervals of the slitting process (09:28:04 to 09:31:18), and the action intervals of the winding process (10:06:20 to 10:10:02). Furthermore, the action chain continues to be associated with the action intervals of the formation process (10:42:10 to 10:46:28), the capacity grading process (11:18:04 to 11:22:16), and the action intervals of the inspection process (11:43:21 to 11:46:30), for comparison of subsequent recovery and subsequent continuation relationships. Within each operating range, both the output and input states are composed of thickness deviation, tension deviation, and temperature deviation values. The thickness deviation is defined as the difference between the actual thickness and the reference thickness divided by the reference thickness; the tension deviation is defined as the difference between the actual tension and the reference tension divided by the reference tension; and the temperature deviation is defined as the difference between the actual temperature and the reference temperature divided by the reference temperature. To ensure consistent state extraction boundaries, the output state of the preceding operating range is the average of the state components within a consecutive 5 seconds from the end of that range, and the input state of the following operating range is the average of the state components within a consecutive 5 seconds from the beginning of that range. Taking the coating process and the rolling process as an example, the output status of the coating process is (0.036, 0.028, 0.011), and the input status of the rolling process is (0.040, 0.031, 0.010); between the rolling process and the slitting process, the output status of the rolling process is (0.048, 0.034, 0.012), and the input status of the slitting process is (0.051, 0.036, 0.013); between the slitting process and the winding process, the output status of the slitting process is (0.057, 0.041, 0.014), and the input status of the winding process is (0.083, 0.020, 0.026). Subsequently, the action transmission breakpoints are determined, including: for the output state of the previous action interval and the input state of the next action interval between adjacent actions in the target action chain, extracting the state offset, state direction change, and state continuation interruption at the corresponding positions, forming a coupling mismatch feature group for each pair of adjacent actions. Here, the state offset is defined as the average of the absolute values ​​of the differences between the output and input state components in the same pair of adjacent actions; the state direction change is defined as the proportion of the number of state components whose offset signs change in the opposite direction between adjacent actions to the total number of state components; and the state continuation interruption is defined as the ratio of the actual interval between adjacent actions divided by the upper limit of the allowed continuation interval, which is set to 120s in this embodiment.

[0031] In this embodiment, the actual interval between adjacent processes is not directly taken as the surface time difference between the two working intervals, but is calculated by subtracting the final dwell time on the intermediate vehicle after the previous interval leaves the station from the time the latter interval enters the station for the same production object. The final dwell time on the intermediate vehicle is determined by the contact triggering time in the process flow record and the departure time in the equipment event. If the contact triggering time is later than the departure time, the contact triggering time is taken as the final dwell time on the intermediate vehicle. For example, between the rolling process and the slitting process, although the time difference between the surface exit time of the rolling process and the surface entry time of the slitting process is 1444s, the contact activation time of the production object on the intermediate carrier is 09:27:26, and the entry time of the slitting process is 09:28:04, so the effective contact interval time is 38s; between the slitting process and the winding process, the contact activation time on the intermediate carrier is 10:05:29, and the entry time of the winding process is 10:06:20, so the effective contact interval time is 51s. Accordingly, the state offset between the coating process and the rolling process is 0.0027, the state direction change is 0, and the state continuity interruption is 42÷120=0.35; the state offset between the rolling process and the slitting process is 0.0020, the state direction change is 0, and the state continuity interruption is 38÷120=0.32; the state offset between the slitting process and the winding process is 0.0197. The thickness deviation, tension deviation, and temperature deviation values ​​corresponding to the output state of the slitting process all have positive signs relative to their respective reference states; in the input state of the winding process, the thickness deviation and temperature deviation values ​​have positive signs relative to their respective reference states, while the tension deviation value has a negative sign relative to the reference tension state of the winding process. Therefore, only the tension deviation value experiences a direction reversal, with a state direction change of 1÷3=0.33; the state continuity interruption is 51÷120=0.43, thus forming three sets of coupling mismatch characteristic groups.

[0032] After forming the corresponding coupling mismatch feature groups for each adjacent process pair, the co-occurrence relationship of state offset, state direction change, and state continuation interruption in each coupling mismatch feature group at the same transmission position is compared to identify candidate breakpoint positions that simultaneously satisfy offset mismatch, direction reversal mismatch, and continuation interruption mismatch. In this embodiment, the determination rule for offset mismatch is that the state offset is not less than 0.015; the determination rule for direction reversal mismatch is that the state direction change is not less than 0.30; and the determination rule for continuation interruption mismatch is that the state continuation interruption is not less than 0.40. According to these rules, the coating process and the rolling process do not constitute candidate breakpoint positions; the rolling process and the slitting process do not constitute candidate breakpoint positions; the state offset of 0.0197, the state direction change of 0.33, and the state continuation interruption of 0.43 between the slitting process and the winding process all meet the conditions, therefore, the position between the slitting process and the winding process is identified as a candidate breakpoint position. Subsequently, for each candidate breakpoint location, the state recovery amount, state direction recovery amount, and state continuation recovery amount for the corresponding position in its subsequent processes are extracted. The recovery relationship of each candidate breakpoint location in the subsequent processes is compared, and candidate breakpoint locations that can recover the state offset, state direction, and state continuation within a preset range of subsequent processes are eliminated. In this embodiment, the preset range of subsequent processes is set as the connection position of the two subsequent processes after the candidate breakpoint location, namely the position from the winding process to the formation process and the position from the formation process to the capacity separation process. The state recovery amount is defined as the decrease ratio of the state offset of the corresponding position in the subsequent process relative to the state offset of the candidate breakpoint location. The state direction recovery amount is defined as the decrease ratio of the change in state direction of the corresponding position in the subsequent process relative to the change in state direction of the candidate breakpoint location. The state continuation recovery amount is defined as the decrease ratio of the interruption of state continuation at the corresponding position in the subsequent process relative to the interruption of state continuation at the candidate breakpoint location. If the state recovery amount, state direction recovery amount, and state continuation recovery amount are all not less than 0.50, the recovery is considered successful and the candidate breakpoint location is eliminated. In this embodiment, the state offset from the winding process to the formation process is 0.0170, the state recovery amount is 0.14, the state direction recovery amount is 0, and the state continuation recovery amount is 0.09; the state offset from the formation process to the capacity separation process is 0.0162, the state recovery amount is 0.18, the state direction recovery amount is 0, and the state continuation recovery amount is 0.14. Since all of these recovery amounts are below 0.50, this candidate breakpoint position cannot be recovered within the preset subsequent process range and is therefore retained.

[0033] If candidate breakpoints are retained after being filtered by the recovery relationship, the closure status of the retained candidate breakpoints in terms of the preceding output relationship, the current input relationship, and the subsequent succession relationship is compared. Candidate breakpoints that can still form the preceding output closure, the current input closure, and the subsequent succession closure are eliminated. In this embodiment, the rule for determining the preceding output relationship is: if the average absolute value of the difference between the output state of the previous action interval before the candidate breakpoint position and the historical stable state of the output of this process does not exceed 0.010, then the preceding output relationship is considered closed. The rule for determining the current input relationship is: if the average absolute value of the difference between the input state of the next action interval after the candidate breakpoint position and the input reference state of this process does not exceed 0.010, then the current input relationship is considered closed. The rule for determining the subsequent succession relationship is: the input state of the next action interval after the candidate breakpoint position and the output state of the next process are mapped one-to-one according to the carrier number and production sequence of the same production object in the process flow record. If the average absolute value of the difference between the corresponding state components after mapping does not exceed 0.012, then the subsequent succession relationship is considered closed. Taking the candidate breakpoint position between the slitting and winding processes as an example, the average difference between the output state of the slitting process and the historical stable output state of the slitting process (0.041, 0.036, 0.013) is 0.0073, therefore the preceding output relationship is closed. The average difference between the input state of the winding process and the input reference state of the winding process (0.061, 0.029, 0.015) is 0.0140, which exceeds 0.010, therefore the current input relationship is not closed. The average difference between the input state of the winding process and the output state of the subsequent formation process after mapping by carrier number and production sequence is 0.016, which exceeds 0.012, therefore the subsequent continuation relationship is not closed. If at least two of the preceding output relationship, current input relationship, and subsequent continuation relationship corresponding to the candidate breakpoint position are not closed, and there are no other candidate breakpoint positions in the same target action chain that satisfy the same coupling mismatch and recovery failure conditions, then the candidate breakpoint position is considered to satisfy the unique closure failure condition. Since in this embodiment only the position between the slitting process and the winding process satisfies the conditions for coupling mismatch and recovery failure, and both the current input relationship and the subsequent connection relationship are not closed, the candidate breakpoint position satisfies the condition for unique closure failure.

[0034] After the above screening, the candidate breakpoints that are selected and simultaneously satisfy the conditions of coupling mismatch, recovery failure, and unique closure failure are determined as action transmission breakpoints. In this embodiment, after screening based on the co-occurrence relationship of state offset, state direction change, and state continuation interruption, only the candidate breakpoint between the slitting and winding processes is retained. After comparing the recovery relationships of subsequent processes, this position fails to meet the requirements of state offset recovery, state direction recovery, and state continuation recovery. After comparing the closure conditions of the preceding output relationship, the current input relationship, and the subsequent successor relationship, this position satisfies the condition of unique closure failure. Therefore, the position between the slitting and winding processes is finally determined as the action transmission breakpoint. In this embodiment, the time range corresponding to this action transmission breakpoint is from 09:31:18 to 10:06:20, and the corresponding process connection relationship is from the slitting process to the winding process. Through this step, based on the already determined target action chain, the key mismatch positions in the cross-process action transmission relationship can be further identified, providing accurate breakpoint basis for subsequent responsibility unit positioning based on action transmission breakpoints.

[0035] S5. Based on the action transmission breakpoint, extract the local action change characteristics of each responsible unit in the corresponding process before and after the action transmission breakpoint, and verify the consistency of the local action change characteristics with the abnormal start position, abnormal expansion path and abnormal change intensity. Gradually eliminate the responsible units that do not meet the consistency conditions to determine the target responsible unit. Specifically, firstly, taking the established action transmission breakpoint located at the corresponding position in the winding process as an example, in this embodiment, winding stations 1, 2, and 3, and needle modules 1, 2, and 3 within the winding process are designated as responsibility units. Subsequently, the process of "based on the abnormal expansion path in the subsequent abnormal response structure, reversing the corresponding local action change segment according to the production sequence to form a candidate set of responsibility units" is executed. The specific rule is: if, after the winding completion sequence corresponding to a responsibility unit is mapped to the production sequence of continuous quality inspection results, its mapped segment overlaps with at least three inspection numbers in any segment of abnormal expansion paths 304 to 307, 309 to 313, or 315 to 327, then that responsibility unit is included in the candidate set of responsibility units. According to this rule, winding stations 1, 2, and 3, as well as needle modules 1, 2, and 3, all enter the candidate set of responsibility units.

[0036] For each responsible unit in the candidate set of responsible units, the initial consistency relationship between its local action change characteristics and the abnormal starting position is compared. Responsible units that do not meet the consistency of the abnormal starting position are eliminated, resulting in the first candidate set of responsible units. In this embodiment, the local action change characteristics consist of thickness deviation change value, tension fluctuation value, and temperature rise deviation value. If the position of a responsible unit that first exceeds the initial judgment threshold is mapped to the production sequence of continuous quality inspection results, and the number deviation from any position among the abnormal starting positions 304-309-315 does not exceed one inspection number, then the consistency of the abnormal starting position is considered to be satisfied. Among them, the threshold for thickness deviation change value is 0.012, the threshold for tension fluctuation value is 0.010, and the threshold for temperature rise deviation value is 0.008. After comparison, winding stations 1 and 2 and needle winding modules 1 and 2 are retained, while winding station 3 and needle winding module 3 are eliminated, resulting in the first candidate set of responsible units. Subsequently, for each responsible unit in the first set of candidate responsible units, the propagation consistency relationship between its local action change characteristics and the abnormal propagation path was compared. Responsible units that did not meet the abnormal propagation path consistency requirement were eliminated, resulting in the second set of candidate responsible units. The rule for determining the propagation consistency relationship was: the overlap ratio between the propagation segment corresponding to the responsible unit and the abnormal propagation path was not less than 80%, and the longest continuous overlap length was not less than 5 detection numbers. After comparison, winding stations 1 and 2, and winding needle module 2 were retained, while winding needle module 1 was eliminated, resulting in the second set of candidate responsible units.

[0037] For each responsibility unit in the second set of candidate responsibility units, the intensity consistency relationship between its local action change characteristics and abnormal change intensity is compared. Responsibility units that do not meet the abnormal change intensity consistency requirement are eliminated, resulting in the third set of candidate responsibility units. In this embodiment, the intensity value corresponding to the local action change characteristics is obtained by weighting the thickness deviation change value, tension fluctuation value, and temperature rise deviation value with weights of 0.4, 0.4, and 0.2, and then averaging the values ​​of the overlapping sections. The intensity value of winding station 1 is 0.057, the intensity value of winding station 2 is 0.068, and the intensity value of needle module 2 is 0.049. If the absolute value of the difference between the intensity value corresponding to the responsibility unit and any corresponding segment of the abnormal change intensity (0.055-0.071-0.067) does not exceed 0.010, and the change direction is consistent, then the abnormal change intensity consistency requirement is met. After comparison, winding stations 1 and 2 are retained, and needle module 2 is eliminated, resulting in the third set of candidate responsibility units.

[0038] After each round of responsibility unit screening, the uncertainty value of the corresponding candidate set of responsibility units is calculated, and the degree of uncertainty collapse is determined based on the difference in uncertainty values ​​between adjacent rounds of candidate sets. The responsibility unit that simultaneously satisfies the consistency of the anomaly initiation position, the consistency of the anomaly propagation path, and the consistency of the anomaly change intensity, and has the largest degree of uncertainty collapse, is determined as the target responsibility unit. In this embodiment, the comprehensive explanatory value is obtained by weighted summation of the initial consistency relationship score, the propagation consistency relationship score, and the intensity consistency relationship score, then normalized to obtain the explanatory probability, and the uncertainty value is calculated. The uncertainty value of the initial candidate set of responsibility units is 1.694, the uncertainty value of the first candidate set of responsibility units is 1.385, the uncertainty value of the second candidate set of responsibility units is 1.098, and the uncertainty value of the third candidate set of responsibility units is 0.693, with corresponding degrees of uncertainty collapse of 0.309, 0.287, and 0.405, respectively. In the third candidate set of responsibility units, winding station 1 and winding station 2 both satisfy the three types of consistency conditions; if the degree of uncertainty collapse corresponding to the two is the same, the one with the larger comprehensive explanatory value is selected as the target responsibility unit. Since the comprehensive interpretation value of winding station 2 (0.84) is higher than that of winding station 1 (0.81), winding station 2 is ultimately determined as the target responsibility unit.

[0039] On the other hand, this embodiment also discloses a data monitoring system for the entire battery production process, including: The data acquisition module is used to acquire production process data, process flow records and quality inspection data corresponding to each process, and based on the flow time period in the process flow record and the entry time period and exit time period in the equipment event, determine the start position and end position of the production object in each process, and generate the range of action of the production object in the corresponding process. The abnormal response generation module is used to extract the abnormal starting position, abnormal expansion path and abnormal change intensity for abnormal production objects according to the production sequence of continuous quality inspection results, and generate the subsequent abnormal response structure based on the sequential connection relationship between the abnormal starting position, abnormal expansion path and abnormal change intensity. The target action chain determination module is used to extract cross-process candidate action interval groups corresponding to abnormal production objects from the action interval based on the downstream abnormal response structure, and perform continuous correspondence verification according to the process sequence. It retains cross-process candidate action interval groups that satisfy the continuous correspondence relationship of abnormal start position, abnormal expansion path and abnormal change intensity, and removes the remaining cross-process candidate action interval groups to determine the target action chain. The action transmission breakpoint determination module is used to extract the output state of the previous action interval and the input state of the next action interval between adjacent processes based on the target action chain. It constructs the cross-process action transmission relationship based on the offset magnitude, change direction and continuity between the output state and the input state, and determines the position corresponding to the sudden change in offset magnitude, reversal of change direction or interruption of continuity as the action transmission breakpoint. The target responsibility unit determination module is used to extract the local action change characteristics of each responsibility unit in the corresponding process before and after the action transmission breakpoint, based on the action transmission breakpoint, and to verify the consistency of the local action change characteristics with the anomaly start position, anomaly propagation path and anomaly change intensity, and gradually eliminate responsibility units that do not meet the consistency conditions in order to determine the target responsibility unit.

[0040] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0041] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for monitoring data throughout the entire battery production process, characterized in that, include: S1. Obtain the production process data, process flow records and quality inspection data corresponding to each process, and based on the flow time period in the process flow record and the entry time period and exit time period in the equipment event, determine the start position and end position of the production object in each process, and generate the range of action of the production object in the corresponding process. S2. For abnormal production objects, extract the abnormal starting position, abnormal expansion path and abnormal change intensity according to the production sequence of continuous quality inspection results, and generate the subsequent abnormal response structure based on the sequential connection relationship between the abnormal starting position, abnormal expansion path and abnormal change intensity. S3. Based on the post-stage abnormal response structure, extract the cross-process candidate action interval group corresponding to the abnormal production object from the action interval, and perform continuous correspondence verification according to the process sequence. Retain the cross-process candidate action interval group that satisfies the continuous correspondence relationship between the abnormal start position, abnormal expansion path and abnormal change intensity, and eliminate the remaining cross-process candidate action interval group to determine the target action chain. S4. Based on the target action chain, extract the output state of the previous action interval and the input state of the next action interval between adjacent processes. Construct cross-process action transmission relationship according to the offset amplitude, change direction and continuity between the output state and the input state, and determine the position corresponding to the abrupt change in offset amplitude, reversal of change direction or interruption of continuity as the action transmission breakpoint. S5. Based on the action transmission breakpoint, extract the local action change characteristics of each responsible unit in the corresponding process before and after the action transmission breakpoint, and verify the consistency of the local action change characteristics with the abnormal start position, abnormal expansion path and abnormal change intensity. Gradually eliminate the responsible units that do not meet the consistency conditions to determine the target responsible unit.

2. The method for monitoring data throughout the entire battery production process according to claim 1, characterized in that, Step S1 generates the corresponding effective range, including: S11. Based on the previous handover period, the current process acceptance period, and the entry and exit periods in the equipment events in the process flow record, extract multiple candidate start positions and multiple candidate end positions of the production object in the corresponding process to form a set of candidate start positions and a set of candidate end positions. S12. Combine each candidate starting position in the candidate starting position set with each candidate ending position in the candidate ending position set to form multiple candidate action intervals, and eliminate candidate action intervals that do not meet the conditions of candidate starting position preceding candidate ending position, preceding continuous handover, and subsequent continuous transfer. S13. For the candidate action intervals that have been screened and retained, compare the consistency between the candidate starting position and the handover sequence of the preceding process and the correspondence between the candidate ending position and the production sequence of the continuous quality inspection results, and eliminate the candidate action intervals that have inconsistent starting sequences or do not correspond to the ending sequences. S14. In the presence of parallel devices, parallel workstations or parallel modules, compare the interval overlap and occupancy conflict of the remaining candidate action intervals on each parallel processing unit, and eliminate the candidate action intervals that cannot be uniquely attributed to a single parallel processing unit. S15. The candidate action ranges that have been screened and retained are determined as the action ranges of the production object in the corresponding process.

3. The method for monitoring data throughout the entire battery production process according to claim 2, characterized in that, Step S14 includes: For each parallel processing unit, the starting position, ending position and duration of the candidate action interval within the corresponding parallel processing unit are extracted, and the occupancy representation of the candidate action interval within the corresponding parallel processing unit is defined as the candidate occupancy interval, forming a set of candidate occupancy intervals for each parallel processing unit. In each set of candidate occupancy intervals, the occupancy overlap relationship between different candidate occupancy intervals within the same occupancy duration segment is identified, and based on the occupancy exclusivity constraint that only a single production object is allowed to continuously occupy the same occupancy duration segment within the same parallel processing unit, candidate occupancy intervals with occupancy overlap conflicts are eliminated. Identify the processing switching positions between adjacent parallel processing units, and construct switching boundary isolation segments with each processing switching position as the center. Compare the cross relationship between the remaining candidate occupancy intervals and the switching boundary isolation segments, and eliminate candidate occupancy intervals that cross the switching boundary isolation segments and cannot maintain occupancy continuity. For the candidate occupancy intervals retained after occupancy exclusion constraints and switching boundary isolation screening, compare their forward closure relationship with the handover sequence of the preceding process, their intermediate closure relationship with the current process acceptance sequence, and their backward closure relationship with the production sequence of continuous quality inspection results, and eliminate candidate occupancy intervals that cannot simultaneously satisfy the forward closure relationship, intermediate closure relationship, and backward closure relationship. Candidate occupancy intervals that have been screened and retained and can form a unique occupancy relationship and a unique closure relationship within a single parallel processing unit are determined as the unique ownership intervals of the production object in the corresponding process, and the corresponding parallel processing unit is determined as the unique ownership processing unit of the production object in the corresponding process.

4. The method for monitoring data throughout the entire battery production process according to claim 1, characterized in that, Step S2 generates the subsequent abnormal response structure, including: According to the production sequence of continuous quality inspection results, the abnormal change sequence of abnormal production objects in the detection sequence is extracted, and the abnormal change intensity sequence and abnormal cluster density sequence corresponding to the production sequence are formed based on the abnormal change sequence. Identify abrupt changes in the gradient of abrupt changes in intensity in the abrupt change sequence and abrupt changes in abrupt cluster density in the abrupt cluster density sequence. Determine the overlapping or adjacent positions of the abrupt changes in intensity gradient and abrupt changes in abrupt cluster density as abrupt evolution inflection points. The detection section between adjacent abnormal evolution inflection points is used as the abnormal response segment. The continuous quality detection results are segmented to form multiple abnormal response segments connected sequentially according to the production order. For each abnormal response segment, the local abnormality start position, local abnormality expansion path, and local abnormality change intensity are extracted. The local abnormality expansion path continuation relationship and local abnormality change intensity change relationship between adjacent abnormal response segments are compared, and abnormal response segments that do not meet the path continuation condition and intensity change condition are eliminated. The selected and retained abnormal response segments are connected in production order, and the local abnormality start position, local abnormality expansion path and local abnormality change intensity in each abnormal response segment are combined to generate the subsequent abnormal response structure.

5. The method for monitoring data throughout the entire battery production process according to claim 1, characterized in that, Step S3 involves performing continuous corresponding verification according to the process sequence, including: For each cross-process candidate action interval group corresponding to the abnormal production object, the output state of the previous candidate action interval and the input state of the next candidate action interval are extracted according to the process sequence to form the state transmission sequence corresponding to each cross-process candidate action interval group. Compare the state continuity, state offset, and state succession relationships between adjacent output and input states in each state transmission sequence, and eliminate cross-process candidate action interval groups that have state transmission interruption, state offset reversal, or state succession mismatch. The cross-process candidate action interval groups retained after state transmission screening are mapped to the anomaly start position, anomaly propagation path and anomaly change intensity in the subsequent anomaly response structure according to the process sequence, forming the anomaly interpretation sequence corresponding to each cross-process candidate action interval group. Compare the continuous correspondence between each anomaly interpretation sequence and the subsequent anomaly response structure in terms of anomaly initiation position, anomaly propagation path and anomaly change intensity, and eliminate cross-process candidate action interval groups that cannot form a continuous explanation for the subsequent anomaly response structure. For the cross-process candidate action interval groups retained after anomaly interpretation screening, their closure relationships at the three levels of previous input, current acceptance, and subsequent output are further compared, and cross-process candidate action interval groups that cannot form a unique closure relationship are eliminated. The cross-process candidate action interval group that has been screened and retained and simultaneously satisfies the closed state transmission, continuous anomaly interpretation, and unique closed relationship is determined as the target action chain corresponding to the abnormal production object.

6. The method for monitoring data throughout the entire battery production process according to claim 5, characterized in that, Step S3 involves determining the target action chain, including: The candidate action intervals corresponding to the abnormal production object are constructed into a cross-process candidate transfer graph according to the process sequence. Each candidate action interval is used as a graph node, and graph edges are established between candidate action intervals that meet the acceptance conditions between adjacent processes. For each cross-process candidate action interval group, the corresponding state transmission constraint value is calculated based on the output state of the previous candidate action interval and the input state of the next candidate action interval between adjacent candidate action intervals. Map the anomaly in the subsequent anomaly response structure to each cross-process candidate action interval group, and calculate the corresponding anomaly interpretation constraint value. Based on the parallel processing unit occupancy conflict relationship, interval overlap relationship and attribution conflict relationship, calculate the conflict suppression constraint value of each cross-process candidate action interval group; Based on the state transmission constraint value, anomaly interpretation constraint value, and conflict suppression constraint value, calculate the closed transmission value of each cross-process candidate action interval group, and remove cross-process candidate action interval groups whose closed transmission value is lower than the preset closed threshold. The cross-process candidate action interval group retained after screening by closed transitive values ​​is normalized, and the initial uncertainty value of the candidate set is calculated based on the normalization result. For each selected and retained cross-process candidate action interval group, other cross-process candidate action interval groups that conflict with them in terms of parallel processing unit affiliation and process succession relationship are removed. The uncertainty value of the remaining candidate set is recalculated, and the corresponding uncertainty collapse degree is calculated based on the difference between the initial uncertainty value and the recalculated uncertainty value. The cross-process candidate action interval group that simultaneously satisfies the condition that the closed transmission value is not lower than the preset closed threshold and has the largest uncertainty collapse degree is determined as the target action chain corresponding to the abnormal production object.

7. The method for monitoring data throughout the entire battery production process according to claim 1, characterized in that, Step S4 involves determining the action propagation breakpoint, including: For the output state of the previous action interval and the input state of the next action interval between adjacent processes in the target action chain, extract the state offset, state direction change and state continuation interruption at the corresponding positions to form the corresponding coupling mismatch feature group for each adjacent process pair. By comparing the co-occurrence relationships of state offset, state direction change, and state continuation interruption in each coupling mismatch feature group at the same transmission position, candidate breakpoint positions that simultaneously satisfy offset mismatch, direction reversal mismatch, and continuation interruption mismatch are identified. For each candidate breakpoint, the state recovery amount, state direction recovery amount and state continuation recovery amount of the corresponding position in the subsequent process are extracted. The recovery relationship of each candidate breakpoint in the subsequent process is compared, and candidate breakpoints that can recover state offset, state direction and state continuation within the preset subsequent process range are eliminated. For the candidate breakpoints retained after the relationship restoration screening, compare their closure status in the preceding output relationship, current input relationship and subsequent succession relationship, and eliminate the candidate breakpoints that can still form the preceding output closure, current input closure and subsequent succession closure; Candidate breakpoints that have been screened and retained and simultaneously satisfy the conditions of coupling mismatch, recovery failure, and unique closure failure are determined as action propagation breakpoints.

8. The method for monitoring data throughout the entire battery production process according to claim 1, characterized in that, Step S5 involves identifying the target responsibility unit, including: Based on the abnormal propagation path in the subsequent abnormal response structure, the corresponding local action change section is pointed back in the reverse direction according to the production sequence to form a candidate set of responsibility units. For each responsible unit in the candidate set of responsible units, compare the initial consistency relationship between its local action change characteristics and the anomaly start position, and eliminate the responsible units that do not meet the anomaly start position consistency to obtain the first candidate set of responsible units; For each responsibility unit in the first responsibility unit candidate set, compare the propagation consistency relationship between its local role change characteristics and the abnormal propagation path, and eliminate responsibility units that do not meet the abnormal propagation path consistency to obtain the second responsibility unit candidate set. For each responsibility unit in the candidate set of the second responsibility unit, compare the intensity consistency relationship between its local action change characteristics and the intensity of abnormal change, and eliminate responsibility units that do not meet the intensity consistency of abnormal change to obtain the candidate set of the third responsibility unit. After each round of responsibility unit screening, the uncertainty value of the corresponding candidate set of responsibility units is calculated, and the degree of uncertainty collapse is determined based on the difference in uncertainty values ​​between the candidate sets of responsibility units in two adjacent rounds. The responsible unit that simultaneously satisfies the consistency of the anomaly's starting location, the anomaly's expansion path, and the anomaly's change intensity, and corresponds to the largest degree of uncertainty collapse, is identified as the target responsible unit.

9. A data monitoring system for the entire battery production process, characterized in that, The data monitoring system employs the data monitoring method described in any one of claims 1-8.

Images