An AGI-based process handling system

Abstract

The application relates to the technical field of enterprise process management, in particular to a process processing system based on AGI, which comprises a semantic density analysis module, a fusion node construction module, a process rhythm analysis module, a device task matching module and a scheduling window regulation module. According to the application, the process steps and resource number fields are extracted, the deviation between the keywords and the number items is calculated, the fusion sorting between the task groups is guided, the original instruction is logically reconstructed in the task relationship, and the accuracy and efficiency of task fusion are improved. The intersection comparison of the resource calling time field and the investigation mode of the device conflict field are used to establish the task parallel execution relationship under the condition that the resources are conflict-free, the scheduling relationship between the multiple tasks is more reasonable, the resource occupation conflict is avoided, and the concurrency and smoothness of execution are enhanced. Through the clustering processing of the material label and the process time, the offset mapping is generated in combination with the actual arrival time, and the adaptation degree between the task rhythm and the process time efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of enterprise process management technology, and in particular to an AGI-based process processing system. Background Technology

[0002] Enterprise process management (ERP) technology encompasses activities such as planning, designing, executing, monitoring, and optimizing internal and cross-departmental workflows. Its core content lies in improving the standardization, automation, and collaboration of business processes through systematic methods, thereby enhancing organizational operational efficiency and resource utilization. This technology covers aspects such as business process modeling, process engine design, rule engine execution, process monitoring, and data analysis, and is widely applied in various enterprise business scenarios, including manufacturing, supply chain management, customer service, financial management, and human resources.

[0003] Among them, the process processing system refers to the information system structure used to execute specific business process tasks and realize the logical advancement of the process. It includes multiple aspects such as the allocation condition judgment of business task nodes, process path control, execution status recording, and abnormal process handling. Specifically, it defines the logical relationship and dependency order between nodes by constructing a task process mapping structure, uses a set of rules to determine path branches, combines task execution interfaces to trigger tasks and update status, and records and tracks process execution information through a unified data structure.

[0004] In existing technologies, task node setting relies on manual configuration and predefined process paths, lacking the ability to automatically integrate task semantics. This results in weak recognition of collaborative relationships between tasks, often leading to redundant scheduling when faced with semantically similar tasks. Process path control is based solely on static judgment and rule combinations, failing to perform high-precision comparison and processing of equipment usage time and resource conflicts. This easily leads to resource preemption during concurrent execution, affecting the overall process continuity. Execution status records are not strongly correlated with equipment operating status, lacking a dynamic comparison mechanism for rhythm deviations and process standards. Task beats often misalign with equipment response times, causing equipment idling or extended waiting periods. Task scheduling lacks dynamic adjustment capabilities. When faced with delays and equipment failures, process activation times cannot be flexibly adjusted according to real-time status, allowing delays to easily propagate through the process. For example, in manufacturing processes, when the same equipment continuously executes multiple process tasks, the failure to determine the semantic similarity between tasks and the time intervals for resource scheduling leads to significantly increased task queuing times and frequent resource conflicts, directly impacting production rhythm and capacity release. These problems limit the adaptive capabilities and resource coordination efficiency of the process system in dynamic environments. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and propose an AGI-based process processing system.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an AGI-based process processing system, the system comprising:

[0007] The semantic density parsing module obtains the task execution instruction log issued by the enterprise scheduler, extracts the process step name and resource station number fields from the task instructions, counts the number of keywords and number items, calculates the semantic density deviation, and merges task groups based on the semantic density deviation to generate a task fusion candidate list.

[0008] The fusion node construction module extracts the fields of conflicting task resource call time and device usage from the task fusion candidate list, performs time segment intersection comparison and conflict field investigation, and generates a set of parallel task nodes.

[0009] The process rhythm analysis module obtains the material tag field corresponding to the task in the parallel task node set, clusters all time data of process number to form standard process time, compares the current arrival time offset with the task, and generates a process rhythm offset list.

[0010] The equipment task matching module extracts the equipment resource pool field of the parallel task nodes in the process rhythm offset list, combines the equipment operating status and response time period, establishes a task-equipment rhythm matching relationship, binds the equipment and sets an executable window, and generates the task flow processing result.

[0011] The present invention improves upon this invention by including the following: the task fusion candidate list includes a semantic density deviation threshold, the semantic correlation between adjacent tasks, and the fusion task group index; the parallel task node set specifically includes resource call time coverage relationship, equipment conflict exclusion marker, and task synchronization start identifier; the process rhythm offset list includes process rhythm offset, standard process time difference, and task combination rhythm comparison relationship; and the task flow processing result includes task node binding structure, equipment rhythm matching relationship, and task executable time window.

[0012] The present invention is improved in that the semantic density parsing module includes:

[0013] The field extraction submodule obtains the task execution instruction log issued by the enterprise scheduler, extracts the process step name and resource station number fields corresponding to each task instruction, identifies the number of keyword items corresponding to each content of the field, identifies the number of equipment number, station number and tool number in the resource station number field, and generates the field structure density based on the extracted field content and quantity distribution.

[0014] The semantic computation submodule constructs the density vector of each task based on the field structure density using the AGI natural language processing model, calculates the difference between the density vectors of two adjacent tasks in the task sequence, obtains the semantic deviation between each group of tasks, and compares the difference with the set semantic fusion judgment threshold to obtain the semantic deviation degree between adjacent tasks.

[0015] The task fusion identification submodule filters task pairs with semantic deviation less than the semantic fusion judgment threshold based on the semantic deviation of adjacent tasks, combines and classifies the task pairs according to their order in the original task sequence, identifies the semantic similarity tags between tasks in each combination, establishes a fusion task group index, and obtains a task fusion candidate list.

[0016] The present invention is improved in that the fusion node construction module includes:

[0017] The time intersection comparison submodule obtains the resource call time field of each group of tasks in the task fusion candidate list, identifies the start and end time segments of each group of tasks, performs an intersection comparison on the time segments between all tasks, filters the intersection segments with continuous coverage relationship, calculates the coverage integrity ratio of each group of intersection segments, and obtains the resource call coverage set.

[0018] The conflict field screening submodule obtains the device usage conflict field of each group of tasks based on the resource call coverage set, determines whether there are overlapping and duplicate items in the device number field between tasks, marks the task combination with the intersection mark in the conflict field, compares the coverage integrity ratio with the ratio threshold, obtains the task combination that meets the coverage requirements and has no intersection mark, and establishes a conflict-free fusion pair list.

[0019] The task node generation submodule recombines the time period content and device number field of each group of tasks according to the list of conflict-free fusion pairs, constructs the node structure in the task scheduling relationship graph, configures the parallel identifier and task path index number for each combination item, and generates a set of parallel task nodes.

[0020] The present invention is improved in that the process rhythm analysis module includes:

[0021] The material identification submodule obtains the material tag field corresponding to each task in the parallel task node set, identifies the association information between the material number and the process number in the material tag field, calls the task index number in the parallel task node, constructs the binding structure between the process number and the task node, establishes a material correspondence table for each task node, and generates task process mapping results.

[0022] The time extraction submodule extracts the process record time field corresponding to the process number in the bound task according to the task process mapping result, gathers the same type of data according to the process number, performs time period classification and clustering on the process number time field, counts the frequency center position in the time period of each type of process, and obtains the standard process time interval.

[0023] The rhythm comparison submodule compares the current arrival time field of each task in the parallel task node set with the standard process time interval, calculates the offset time between the current arrival time of each task and the center of the standard process time interval, and combines them according to the order of task nodes to establish a process rhythm offset list.

[0024] The present invention is improved in that the device task matching module includes:

[0025] The resource status identification submodule extracts the equipment resource pool field corresponding to each parallel task node in the process rhythm offset list, identifies the current running status and response time field corresponding to the equipment number field in the resource pool, establishes a mapping set between equipment number and corresponding status information according to the equipment number order, calls the task node index information for matching and sorting, and generates equipment status operation analysis results.

[0026] The rhythm adaptation mapping submodule compares the rhythm offset of each task with the response time interval of the corresponding device number based on the device status operation analysis results, filters the device numbers that meet the response time correspondence, and binds the filtered device numbers to the task nodes to generate node device adaptation results.

[0027] Based on the node device adaptation results, the task path construction submodule extracts the response time period field corresponding to the device number in each group of binding results, sets the executable time interval for each task node, sorts the paths according to the sequential numbering of the task nodes, combines the device number with the task node and its executable window information, and establishes the task flow processing result.

[0028] The present invention is improved by further including a scheduling window control module. The scheduling window control module calculates the ratio of task delay time to equipment recovery cycle based on the task flow processing result, generates a delay coefficient, adjusts the task activation time field, writes it into the task control table, and obtains the optimized task flow.

[0029] The optimized task flow includes activation time adjustment, delay coefficient mapping relationship, and task control field update record.

[0030] The present invention is improved in that the scheduling window control module includes:

[0031] The delay coefficient calculation submodule obtains the task node sequence number in the task flow processing result, extracts the previous task delay duration field corresponding to each task node and the recovery cycle field corresponding to the current device number, pairs the fields according to the task number order and calculates the corresponding numerical ratio to generate the task delay coefficient.

[0032] The time field adjustment submodule calls the activation time field of each task node in the task flow processing result according to the task delay coefficient, calculates the time offset between the delay coefficient and the original activation time field, updates the field content according to the adjusted time of each task node, and obtains the activation time adjustment result.

[0033] Based on the activation time adjustment result, the scheduling structure update submodule performs field value replacement operation on the activation field in the control table corresponding to each task node, writes the adjusted time to the corresponding activation field position, and reconstructs the scheduling path and resource binding information according to the task number order to establish the optimized task flow.

[0034] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0035] In this invention, by extracting process steps and resource number fields and calculating the deviation between keywords and numbered items, the fusion and sorting of task groups is guided, enabling the original instructions to logically reconstruct task relationships and improving the accuracy and efficiency of task fusion. By utilizing the intersection comparison of resource call time fields and the investigation of equipment conflict fields, a parallel execution relationship of tasks under resource-free conditions is established, making the scheduling relationship between multiple tasks more reasonable, avoiding resource occupation conflicts, and enhancing the concurrency and smoothness of execution. Through clustering of material tags and process times, standard process time intervals are formed, and then offset mapping is generated by combining actual arrival times, enhancing the accurate matching of task timing and significantly improving the adaptability between task rhythm and process timeliness. Dynamic matching of equipment operating status and task time offsets achieves reasonable binding of task nodes and equipment response periods, significantly improving equipment resource utilization, while setting clear executable time windows for tasks, improving the overall scheduling rationality. At the level of dynamic adjustment of execution rhythm, a delay coefficient is established by combining task delay duration and equipment recovery cycle to achieve dynamic adjustment of activation time, further reconstructing the scheduling path and task control fields to achieve adaptive optimization of the process. This processing logic achieves a comprehensive improvement in process scheduling efficiency, equipment utilization, and task timeliness accuracy in complex process scenarios through multi-level and multi-faceted innovations such as semantic quantification recognition, task relationship deconstruction and reconstruction, resource status matching, and task time adjustment. Attached Figure Description

[0036] Figure 1 This is a system module diagram of the present invention;

[0037] Figure 2 This is a system framework diagram of the present invention;

[0038] Figure 3 This is a schematic diagram of the semantic density parsing module of the present invention;

[0039] Figure 4 This is a schematic diagram of the fusion node construction module of the present invention;

[0040] Figure 5 This is a schematic diagram of the process rhythm analysis module of the present invention;

[0041] Figure 6 This is a schematic diagram of the device task matching module of the present invention;

[0042] Figure 7 This is a schematic diagram of the scheduling window control module of the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0044] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0045] Please see Figure 1 This invention provides a technical solution: an AGI-based process processing system, the system comprising:

[0046] The semantic density parsing module obtains the task execution instruction log issued by the enterprise scheduler, extracts the process step name and resource station number fields from the task instructions, counts the number of keywords and number items, calculates the semantic density deviation, and merges task groups based on the semantic density deviation to generate a task fusion candidate list.

[0047] The fusion node construction module extracts the fields of conflicting task resource call time and device usage from the task fusion candidate list, performs time segment intersection comparison and conflict field investigation, and generates a set of parallel task nodes.

[0048] The process rhythm analysis module obtains the material tag field corresponding to the task in the parallel task node set, clusters all time data of process number to form standard process time, compares the current arrival time offset with the task, and generates a process rhythm offset list.

[0049] The equipment task matching module extracts the equipment resource pool field of the parallel task nodes in the process rhythm offset list, combines the equipment running status and response time period to establish the task and equipment rhythm matching relationship, binds the equipment and sets the executable window, and generates the task process processing result.

[0050] The task fusion candidate list includes semantic density deviation threshold, semantic correlation between adjacent tasks, and fusion task group index. The parallel task node set specifically includes resource call time coverage relationship, equipment conflict exclusion mark, and task synchronization start mark. The process rhythm offset list includes process rhythm offset, standard process time difference, and task combination rhythm comparison relationship. The task flow processing result includes task node binding structure, equipment rhythm matching relationship, and task executable time window.

[0051] Please see Figure 2 and Figure 3 The semantic density parsing module includes:

[0052] The field extraction submodule obtains the task execution instruction log issued by the enterprise scheduler, extracts the process step name and resource station number fields corresponding to each task instruction, identifies the number of keyword items corresponding to each content of the field, identifies the number of equipment number, station number and tool number in the resource station number field, and generates the field structure density based on the extracted field content and quantity distribution.

[0053] In the initial processing, all task execution instructions from the enterprise scheduling log are imported into a standard log parser. Following the log structure template, the "Process Step Name" and "Resource Station Number" fields are located in each record. Field content is uniformly separated by " / " or "-". First, a standardized terminology library is loaded for the "Process Step Name" field. This library contains predefined terms such as "stamping," "welding," "heat treatment," "assembly," and "inspection." A regular expression matching algorithm is used to sequentially identify whether the field contains these keywords. If a record contains "stamping / welding / assembly," then three keywords are identified and recorded as the number of keywords. The system first identifies 3 items. Then, it proceeds to parsing the "Resource Workstation Number" field, recognizing it as a three-part combination: Equipment Number, Workstation Number, and Tool Number. For example, the number "EQ31-WK09-TL18" will be decomposed into Equipment Number EQ31, Workstation Number WK09, and Tool Number TL18. The system iterates through all records in the log to count the occurrences of each type of number. For instance, after reading 100 task records, the Equipment Number appears 97 times, the Workstation Number 94 times, and the Tool Number 90 times, which are recorded as the number of items for each of the three resource identifiers. Combining this with the number of process keywords and number items identified in each record, a list of the total number of field items for each task is constructed. For example, if a task extracts 3 keyword items, includes Equipment Number and Workstation Number but no Tool Number, then the total number of field items is 5. After identifying all field items for all tasks, the average number of field items across the 100 records is calculated to obtain the field structure density. This is calculated by dividing the total number of field items by the total number of tasks. For example, if the total number of field items is 551, the field structure density is... This value reflects the complexity of the current task instruction structure.

[0054] The semantic computation submodule calculates the task density of each task based on the field structure density, constructs a task density vector through the AGI natural language processing model, identifies the semantic deviation between each group of tasks based on the difference between two adjacent task density vectors in the task sequence, and compares the difference with the set semantic fusion judgment threshold to obtain the semantic deviation degree between adjacent tasks.

[0055] The number of fields extracted from each task needs to be further categorized and evaluated for their density performance within the task. Different field types represent different dimensions in information expression, so they cannot be directly summed. When calculating task density, the task field structure is first decomposed into three categories: the number of process step keywords (counted by terminology), the number of equipment numbers, the number of workstation numbers, and the number of tool numbers. Each type of field is counted independently. For example, in Task A, 3 process keywords, 1 equipment number, 1 workstation number, and 1 tool number are identified, each labeled as K. i =3, D i =1, W i =1,T i=1, and then calculate the individual density for each field category separately, i.e., keyword density is 3 items, equipment identifier density is 1 item, workstation density is 1 item, and tool density is 1 item. The final task density is listed separately for each of the four field categories, rather than being mixed, thus reflecting a more specific structural composition. For example, if task B is "assembly / inspection / grinding, EQ12-WK07", then the field density composition is keyword density 3, equipment density 1, workstation density 1, and tool density 0. These densities are used for subsequent vector modeling. In constructing the task density vector, the OpenAI API combined with the HuggingFaceTransformers framework is used, and modeling is performed through the AGI natural language processing model. This model structure is based on a multi-layer Transformer encoder and a self-attention mechanism, which can generate a fixed-dimensional density vector based on the combination of input field densities. A vector set is constructed using a Python environment, with each task generating a vector in the form V. i =[K i D i W i ,T i For example, if the vector for task A is [3,1,1,1] and the vector for task B is [3,1,1,0], then all task vectors are arranged sequentially to form a density vector sequence. For any two adjacent tasks in the sequence, the Euclidean distance between their vectors is calculated as the semantic dissimilarity. For example, if the task vectors are [3,1,1,1] and [3,1,1,0], then the dissimilarity is... The obtained values ​​are the density differences between adjacent tasks, forming a complete list of task semantic deviations. The deviation judgment threshold is determined by the statistical range of the density difference sequence. Assuming that the maximum value of all Δρ is 3.2 and the minimum value is 0.1, the fusion threshold can be set as the median or interquartile range. For example, if the interquartile range of 1.5 is selected as the threshold, when the density difference between a task pair is less than 1.5, it is judged to be semantically consistent.

[0056] The task fusion identification submodule filters task pairs with semantic deviation less than the semantic fusion judgment threshold based on the semantic deviation of adjacent tasks, combines and classifies the task pairs according to their order in the original task sequence, identifies the semantic similarity tags between tasks in each combination, establishes a fusion task group index, and obtains a task fusion candidate list.

[0057] Task pairs with semantic deviation less than a set threshold are categorized by execution order. A sliding window logic is used to progressively advance the task number sequence. If consecutive task pairs, such as T2-T3 and T3-T4, both meet the criteria, they are merged into task group T2-T4. The grouping ends with task pairs that no longer meet the criteria, and the next task sequence is redefined. After grouping, to ensure semantic consistency among tasks within a group, similarity calculations are performed on the keyword "process step name" and the task structure density vector separately to avoid interference from mixed field units in the overall similarity assessment. A weighted Jaccard variant formula is used for keyword similarity to improve the distinguishability of repeated keywords and word order. The formula is as follows:

[0058]

[0059] Among them, Sim k (A, B) represents the keyword similarity between task A and task B, where A and B represent the sets of process step keywords in task A and task B, respectively, and x is each keyword in set A∩B, i.e., the common keywords appearing in task A and task B. ∑ x∈A∩B min(f A (x),f B (x)) represents the sum of the smaller values ​​of the occurrence counts of all common keywords in the two tasks, ∑ x∈A∪B max(f A (x),f B (x) represents the sum of the largest number of occurrences of all keywords that appear in any task.

[0060] Keyword Similarity Sim k (A,B) represents the degree of overlap between Task A and Task B in the "Process Step Name" field. Its purpose is to quantify the consistency of content between the two tasks at the linguistic description level. Specifically, by comparing the presence and frequency of identical process operation terms in the two tasks, it determines whether the two tasks are similar in their process flow. In the task fusion process, tasks typically originate from different scheduling instructions. If their process descriptions are similar, such as both containing terms like "assembly," "debugging," or "quality inspection," it indicates that the two tasks may be in the same production stage or have similar objectives. By calculating keyword similarity, potential semantic connections between tasks can be discovered even before structural information is fully matched, improving the accuracy and precision of combined recognition. Furthermore, this indicator can effectively avoid recognition biases caused by different word orders or variations in sentence expression, enhancing the system's ability to recognize tasks with actual semantic similarity.

[0061] The keywords for task T6 are {assembly: 2, welding: 1}, and the keywords for task T7 are {assembly: 1, gluing: 1}. The intersection keyword is "assembly", which appears 2 times and 1 time in the two tasks respectively. Therefore:

[0062]

[0063] Structural density similarity is handled separately to avoid mixing keywords and density units. Each task has a structural density vector in the form [K,D,W,T], where K is the number of keywords in the process steps, and D, W, and T are the number of equipment numbers, workstation numbers, and tool numbers, respectively. The structural difference between the density vectors of any two tasks is calculated using Euclidean distance.

[0064]

[0065] Among them, Dist d Let K1, D1, W1, T1 represent the Euclidean distance between the structure densities of the two tasks, and K2, D2, W2, T2 represent the number of corresponding field items in task 1. For example, consider tasks T6 = [3, 1, 1, 0] and T7 = [2, 1, 1, 1].

[0066]

[0067] Then, the distance values ​​are converted into structural similarity using the following mapping method:

[0068]

[0069] Substituting into the above equation, we get:

[0070]

[0071] Keyword similarity and structure density similarity represent the similarity of linguistic content and the consistency of structural complexity in the semantics of the task, respectively. These two values ​​cannot be mixed and must be processed separately. The final combined similarity is the average of these two values, which serves as the criterion for task fusion.

[0072]

[0073] If the similarity threshold for fusion judgment is set to 0.3, then the task pair can be included in the fusion task group. All combined tasks are similarity calculated and merged by number to form a combination sequence, such as T6-T7-T8. Each group of combined tasks then extracts representative words through keyword intersection frequency. For example, if the high-frequency keyword of the combined group is "assembly", then the label is "assembly stage". The combined group data is finally written into the fusion task index table, which includes fields such as combination number, task number range, average keyword similarity, average structural similarity, and label, providing structured grouping support for task scheduling and aggregation.

[0074] Please see Figure 2 and Figure 4 The fusion node building module includes:

[0075] The time intersection comparison submodule obtains the resource call time field of each group of tasks in the task fusion candidate list, identifies the start and end time segments of each group of tasks, performs intersection comparison on the time segments between all tasks, filters the intersection segments with continuous coverage relationship, calculates the coverage integrity ratio of each group of intersection segments, and obtains the resource call coverage set.

[0076] The time intersection comparison operation is based on task combinations in the task fusion candidate list. First, the resource call time field of each task group is read, and the start and end times of each task are extracted to form time segment pairs. The format is uniformly timestamp-like, such as "2025-04-10 08:30:00" to "2025-04-10 11:00:00". The time segments of each task within a task group are aggregated to form the entire coverage segment of the task group. After completing the time aggregation of a single group, pairwise intersection comparisons are performed between the time segments of all task combinations. That is, for any two combinations A and B, their time segments are extracted. and in This represents the starting time point in combination group A. This is the end time of combination group A. This is the starting time point for combination group B. For the end time of combination group B, if there is a time overlap between these two segments, then the condition is met. At that time, the intersection segment is extracted as the shared resource usage time period, and the segment is marked as having a continuous coverage relationship. Then, the coverage integrity ratio is calculated for the intersection segment formed by each task combination. This ratio is used to quantify the density of resource sharing between two task combinations on the time axis. The calculation formula is: Coverage Integrity Ratio = Intersection Time Period Duration ÷ Minimum Task Segment Total Duration, where the intersection time period duration is the overlap length between the two task combinations, and the minimum task segment total duration is the shorter of the two combined segments, compared in minutes. For example, group A starts at 08:00 and ends at 10:00, lasting 120 minutes. Group B starts at 09:00 and ends at 11:30, lasting 150 minutes. The overlap is from 09:00 to 10:00, a total of 60 minutes. The coverage integrity ratio of this group is 60 ÷ min(120, 150) = 0.5, indicating that the tasks have 50% time overlap in resource calls. This comparison is then performed on all groups in this manner, and the system summarizes the resource call coverage set, which is used to filter task combinations with continuous overlap.

[0077] The conflict field screening submodule obtains the device usage conflict field of each group of tasks based on the resource call coverage set, determines whether there are overlapping and duplicate items in the device number field between tasks, marks the task combination with the intersection mark in the conflict field, compares the coverage integrity ratio with the ratio threshold, obtains the task combination that meets the coverage requirements and has no intersection mark, and establishes a list of conflict-free fusion pairs.

[0078] Obtain resource usage records for each task combination, focusing on extracting the device ID field. Compare and analyze the device IDs of each task within the task combination to screen for duplicate IDs appearing in different tasks. For example, if device ID EQ17 exists in both combination A and combination B, then there is overlap between the two groups. The method for determining this is the intersection operation of sets. If device set A ∩ set B... The items are marked as conflict field intersection items, and all marked items are added to the conflict identifier list. Then, the coverage integrity ratio of each pair of combinations is compared with the preset ratio threshold. The ratio threshold is set to 0.6. The standard is set as "integrity ratio ≥ 0.6" to determine that the coverage requirement is met. The threshold value comes from the production cycle standard. For example, a resource chain task usually has a continuous scheduling time of 10 minutes. If the intersection segment is less than 6 minutes, the overlap is considered not significant. Therefore, the threshold is set to 60%. In the example, if the coverage integrity ratio of combination C and combination D is 0.75 and there are no duplicate device numbers, then combination C and D are determined to be a conflict-free fusion pair. The system retains all combination items that meet the "coverage integrity ratio ≥ threshold" and do not have an intersection of device numbers. Finally, a list of conflict-free fusion pairs is established. This list will be used to build the next step of graph structure generation for the task scheduling structure.

[0079] The task node generation submodule recombines the time period content and device number field of each group of tasks based on the list of conflict-free fusion pairs, constructs the node structure in the task scheduling relationship graph, configures the parallel identifier and task path index number for each combination item, and generates a set of parallel task nodes.

[0080] First, the time period field in each group of merged tasks is read, and the start and end time information of each task is extracted and integrated in chronological order to determine the overall time period interval after combination. Then, combined with the equipment number information involved in the combination, a device item merging operation is performed to deduplicate the equipment numbers involved in each task within the combination to generate a unique device set. For example, if combination T9-T11 contains devices EQ21, EQ22, and EQ21, it is merged into the device set {EQ21, EQ22}. Subsequently, based on the relative position of the combination group in the overall task sequence, node elements of the task scheduling structure graph are constructed. Each combination group corresponds to a node in the graph structure, and this node contains the fields: start time, end time, device set, and task. The system assigns a parallel identifier to each node based on the principle of task parallelism. If there is no time overlap and the device numbers are not repeated between two combined groups, the system sets its parallel flag to "1", indicating that it can be executed in parallel. If there is a time or device conflict, it is marked as "0", indicating that it cannot be executed in parallel. At the same time, the system assigns a task path index number to each node. The number is generated based on the combination and sorting order of the original numbers in the task fusion. For example, the path number of combined group T1-T3 is N001, and that of combined group T4-T5 is N002. Finally, the system forms a set of parallel task nodes. The structure includes the time range, device usage, whether it is executed in parallel, and task path code of each combined node, providing a clear node definition for the subsequent execution plan and scheduling graph construction.

[0081] Please see Figure 2 and Figure 5 The process rhythm analysis module includes:

[0082] The material identification submodule obtains the material tag field corresponding to each task in the parallel task node set, identifies the association information between the material number and the process number in the material tag field, calls the task index number in the parallel task node, constructs the binding structure between the process number and the task node, establishes a material correspondence table for each task node, and generates task process mapping results.

[0083] First, the "Material Label Field" is extracted from the task structure. This field typically contains the material number (e.g., MTR1024) and the corresponding process number (e.g., PRC05). The field structure can be in the form of "Material Number - Process Number", such as "MTR1024-PRC05". The system performs semantic parsing on this field to separate each material from its corresponding process. Then, by matching the task index number field in the parallel task node structure, each task is bound to the parsed process number, forming a "Process Number - Task Node" mapping pair. For example, if the task index number is T004 and the material field is "MTR1205-PRC08", then the record is mapped to "P". RC08→T004”, then the system traverses all node records, organizes the material numbers and associated process numbers contained in each task node, and constructs a material correspondence table. The table structure is: task node index number, corresponding process number, and material number set. The correspondence record method supports multiple materials being bound to the same process. For example, “PRC10” contains materials MTR1011 and MTR1012. When they are executed in task nodes T005 and T006 respectively, the material correspondence table will list the correspondence between the two task index numbers and the same process number. Finally, the mapping list of all tasks is output according to the three-element structure of “task index-process number-material number”.

[0084] The time extraction submodule extracts the process record time field corresponding to the process number in the bound task based on the task process mapping result, gathers the same type of data by process number, performs time period classification and clustering on the process number time field, counts the frequency center position in the time period of each type of process, and obtains the standard process time interval.

[0085] First, the system reads the process numbers from all mapping structures and uses the process number as the primary key to aggregate the "process record time field" related to that number in all tasks. This field records the start and end times of each process execution, such as "2025-04-12 09:10:00" to "2025-04-12 10:00:00". The system then performs time-segment clustering on this time field, grouping all execution time segments under the same process number into a single dataset. For example, under "PRC08", 50 task records are aggregated, forming a time-segment set. Next, time-segment clustering is performed on this set, using the distance between time centroids as the grouping criterion. The time centroid of each record is defined as the average time between the start and end times. For example, a... The task starts at 09:00 and ends at 10:00, so the center point is 09:30. The system divides time periods into clusters with a 30-minute interval and calculates the most frequent position of the time center point in each cluster. The time center point with the most occurrences is used as the standard time interval center of the process. Then, the standard time periods are summarized for each type of process. The output format is "process number-time interval center-time period boundary", such as "PRC08-09:45-[09:30,10:15]". The time period boundary can be set with upper and lower limits by offsetting the time center point to the left or right. The default upper and lower limits are set to ±15 minutes. That is, the standard process time interval is the center point plus or minus a fixed offset value. The resulting time period data is finally used as the standard benchmark for subsequent rhythm offset judgment.

[0086] The rhythm comparison submodule compares the current arrival time field of each task in the parallel task node set with the standard process time interval, calculates the offset time between the current arrival time of each task and the center of the standard process time interval, and combines them according to the order of task nodes to establish a process rhythm offset list.

[0087] The system reads each task record in the set of parallel task nodes and extracts the current arrival time field of the task, which records the actual time when the task begins to enter the corresponding process flow. For example, if the current arrival time of task T021 is "2025-04-12 09:42:00", the system calculates the difference between this time and the center of the standard process time interval of the process number bound to the task. The offset duration is calculated as follows: Offset duration = Current arrival time - Standard time center point. The calculation result is output in minutes. If the center point is 09:30, the offset is +12 minutes. If the arrival time is 09:20, the offset is -10 minutes. A positive value indicates that the task arrives later, and a negative value indicates that the task arrives earlier. The unit of the offset value is minutes. The system organizes the results for each task according to its index order and finally generates a rhythm offset list with a three-field structure of "task index - process number - offset duration", such as "T021-PRC08-12min". The system summarizes this list after traversing all parallel task nodes to further analyze the rhythm fluctuation pattern based on the task path or resource usage characteristics.

[0088] Please see Figure 2 and Figure 6 The device task matching module includes:

[0089] The resource status identification submodule extracts the equipment resource pool field corresponding to each parallel task node in the process rhythm offset list, identifies the current running status and response time field corresponding to the equipment number field in the resource pool, establishes a mapping set between equipment number and corresponding status information according to the equipment number order, calls the task node index information for matching and sorting, and generates equipment status operation analysis results.

[0090] The system extracts the device resource pool field corresponding to each parallel task node. This field is typically a set of structured IDs, such as "EQ101, EQ102, EQ103". The system expands each device ID in this field, querying the database for the current running status and response time period fields for each device. The running status field can be identified as "Running", "Idle", or "Maintenance". The response time period field is the time interval during which the device can respond to jobs, formatted as "08:00-10:30" or "14:00-17:00". The system constructs a status mapping table using the device ID as the primary key, with a record format of "Device ID → Status → Response Time Period", such as EQ101 → Running → 08:00-10:30. The system then performs a status query for all device IDs. After data collection by time period, the above information is matched and integrated according to the task node index number in the process rhythm offset list. For example, if task T014 uses resource pools EQ101 and EQ102, the status records of the two devices are bound to T014. If EQ101 is running and EQ102 is idle, the device status corresponding to T014 is a mixed state. The system will mark the status of all devices corresponding to the task node. If one of the devices is in maintenance status, the entire resource status will be recorded as "partially unavailable". Finally, all task nodes and the number, status, and response time period of the devices they use form a complete mapping. The output content forms a device status operation analysis table with the structure of "task node index → ​​device number → status → response interval", which is used for correlation comparison in the subsequent rhythm adaptation screening stage.

[0091] The rhythm adaptation mapping submodule compares the rhythm offset of each task with the response time interval of the corresponding device number based on the device status operation analysis results, filters the device numbers that meet the response time correspondence, and binds the filtered device numbers to the task nodes to generate node device adaptation results.

[0092] By traversing each task node record, the rhythm offset data item is extracted. For example, the rhythm offset of task T018 is +10 minutes, indicating that the task is delayed by 10 minutes relative to the standard process time. Then, the set of equipment numbers bound to the task node is processed, and the response time field of each device is read. The task rhythm offset time is compared with the device response time to determine whether the actual execution time of the task falls within the device response time interval. The judgment condition is "the center point of the task's standard time + the offset time ∈ the response time interval". If the condition is met, the device number is a matching device. Taking task T018 as an example, its process center time is 10:00, and the rhythm offset is +1. If the execution time is 0 minutes, the actual execution time is 10:10. If the response time of EQ122 is 09:30-11:00, then the device can respond to this task, and EQ122 is marked as a matching device. If the response time of another device is 08:00-09:00, then the condition is not met and it is not included in the binding. The system traverses all devices, filters out the set of device numbers that meet the time adaptation relationship, and rebinds the set with the task node. Finally, it generates the adaptation result structure of "task node → matching device number list". After executing this logical process on all parallel task nodes, a complete node device adaptation result is formed. The structure data provides device-level constraint support for subsequent scheduling path construction.

[0093] The task path construction submodule extracts the response time period field corresponding to the device number in each group of binding results based on the node device adaptation results, sets the executable time interval for each task node, sorts the paths according to the sequential number between task nodes, combines the device number with the task node and its executable window information, and establishes the task flow processing result.

[0094] By traversing each task node record, the system extracts the set of device IDs bound to it, and then reads the response time period field corresponding to each device from the device status operation analysis table. For example, if the time period bound to EQ131 is 08:00-10:30 and that of EQ132 is 09:00-11:00, the system will take the intersection of these time periods as the executable time window of the task node. The system performs a segment intersection operation on the available time periods of multiple device IDs, that is, the maximum start time to the minimum end time of multiple response time periods is used as the actual schedulable window of the task. If task T022 is bound to EQ131 and EQ132, then its execution window is 09:00-10:30, and the system assigns this interval to the task. The executable time field of T022 is then sorted by the sequential number of the task nodes, either from smallest to largest or globally based on scheduling priority. This is combined with the device number of each task and the calculated execution time interval to form a scheduling path segment information of "task index → ​​device number → execution window". Finally, the system combines all path segments to form a full-process task execution structure. Each record in the structure represents a specific task path node, which includes three pieces of information: task number, bound device, and executable time period. The path structure is used to fully construct the multi-task scheduling network and feasibility execution graph, supporting subsequent task graph visualization and path allocation analysis.

[0095] Please see Figure 2 and Figure 7 It also includes a scheduling window control module. Based on the task flow processing results, the scheduling window control module calculates the ratio of task delay time to equipment recovery cycle to generate a delay coefficient, adjusts the task activation time field, writes it into the task control table, and obtains the optimized task flow.

[0096] The optimized task flow includes activation time adjustment, delay coefficient mapping relationship, and task control field update records;

[0097] The scheduling window control module includes:

[0098] The delay coefficient calculation submodule obtains the task node sequence number in the task flow processing result, extracts the previous task delay duration field corresponding to each task node and the recovery cycle field corresponding to the current device number, pairs the fields according to the task number order and calculates the corresponding numerical ratio to generate the task delay coefficient.

[0099] First, all task records are extracted according to the sequential numbering of task nodes. The preceding task delay duration field for each task node is read one by one. This field records the cumulative waiting time in minutes caused by abnormalities or delays in the preceding task. For example, the preceding delay for T006 is 18 minutes. Simultaneously, the system extracts the device number bound to the current task node and obtains the device's recovery cycle field from the device's operating information. The recovery cycle field is defined as the period after the device completes fault repair or maintenance and becomes available again, in hours. For example, the recovery cycle for EQ205 is 2 hours. The system then converts the delay duration and recovery cycle to a unified time unit, using minutes as the base for conversion (2 hours equals 120 minutes). Field pairing is performed for each task, comparing the preceding delay time with the device recovery cycle to calculate the composite structure of the delay coefficient as follows:

[0100] Delay coefficient λ i It consists of two independent dimensions: the percentage of delays in preceding tasks. Equipment recovery pending item Its expression is:

[0101]

[0102] Where, λ i This represents the delay coefficient for task i. Indicates the delay time (in minutes) of the preceding task of task i. This represents the average delay time for all tasks. Recovery cycle (in minutes) for the device bound to the current task. This represents the average equipment recovery cycle. The formula independently standardizes and sums the delay effects from both the task chain and equipment status, forming a dimensionless composite coefficient to avoid unit conflicts. For example: Task T008 has a delay of 24 minutes, equipment EQ302 has a recovery cycle of 3 hours (180 minutes), the system's overall average delay is 20 minutes, and the average equipment recovery cycle is 150 minutes. Then the delay coefficient is:

[0103]

[0104] The above coefficients represent the time load of a task affected by both preceding and equipment factors. The system traverses all tasks, extracts fields, converts units, and substitutes formulas in numerical order, and finally generates a delay coefficient table for each task. The structure includes four items: "task number → delay duration → recovery cycle → delay coefficient", which are used as weight input factors in the time field adjustment logic calculation.

[0105] The delay coefficient measures the time load intensity of a single task in the scheduling process caused by the delay of preceding tasks and the impact of equipment resource recovery. The larger the value, the more significant the time delay impact on the execution path of the task. As a key calculation factor in the scheduling adjustment process, this coefficient is mainly used to determine whether the task activation time needs to be delayed and the specific duration of the delay. In the actual calculation process, firstly, the delay time of the preceding tasks for each task is extracted. This data reflects the waiting time caused by the failure of the preceding process to be completed in time. For example, task A is delayed by 15 minutes, task B is delayed by 25 minutes, and the average delay time of all tasks is assumed to be 20 minutes. Secondly, the recovery cycle field of the equipment bound to the task is extracted. This value refers to the waiting time required for the equipment to be restored to availability after maintenance or failure. For example, the recovery cycle of equipment X is 150 minutes, equipment Y is 180 minutes, and the average is 165 minutes. The system normalizes these two values ​​separately, using the method of dividing the actual value of the task by the global average to eliminate the influence of dimensionality. Then, the two normalized results are summed to form the final delay coefficient.

[0106] The time field adjustment submodule calls the activation time field of each task node in the task flow processing result according to the task delay coefficient, calculates the time offset between the delay coefficient and the original activation time field, updates the field content according to the adjusted time of each task node, and obtains the activation time adjustment result.

[0107] First, the system retrieves the original activation time field of all task nodes from the task flow processing results, in the format "04-14, 08:00:00". Then, it reads the corresponding delay coefficient value for each task node, denoted as λ. i Then, the required adjustment time for each task is calculated using a preset time offset base value. This base value depends on the system's standard activation offset baseline value, with a default value set to 10 minutes. The offset is calculated as: time offset value Δt i =λ i ×t base , where Δt i λ represents the time (in minutes) by which the activation time of task i should be shifted backward. i Let t be the time delay factor (dimensionless) for this task. base The standard unit time base value for activating the offset (unit: minutes, default 10 minutes). For example, if the delay factor for task T008 is 2.4 and the standard offset base is 10 minutes, then the task needs to be delayed by: Δt i=2.4 × 10 = 24 minutes; The system adds 24 minutes to the original activation time "08:00:00" to adjust it to "08:24:00". All task nodes execute the same offset adjustment logic and update their activation time field. The system records five data items as an adjustment result list: "task number → original activation time → delay coefficient → offset duration → adjusted activation time", which are then called by the subsequent scheduling path structure reconstruction module.

[0108] Based on the activation time adjustment results, the scheduling structure update submodule performs field value replacement operations on the activation field in the control table corresponding to each task node, writes the adjusted time into the corresponding activation field position, and reconstructs the scheduling path and resource binding information according to the task number order to establish the optimized task flow.

[0109] By reading the control table data structure of each task node, locating the activation field in the control table, and performing a field value replacement operation, the adjusted time overwrites the original activation time field, and recording the modification version number for log traceability, the system then reconstructs the scheduling path between tasks according to the task number order. It performs sorting logic checks on the dependencies between tasks to ensure that the time relationship between preceding and subsequent tasks meets the logical chain constraints. If there is a case where the activation time is less than the completion time of the preceding task, the system shifts the start time of the current task backward, making its start time greater than or equal to the completion time of all preceding tasks plus a buffer period. The buffer period can be set to the minimum system execution cycle, such as 5 minutes. In addition, the system simultaneously reconstructs resource binding information, writes and verifies the device status table based on the device number bound to the task node and its new activation time range. If a device has a task conflict within the new activation time period, the system issues a conflict warning flag, indicating that there may be resource overlap in the current task path. Finally, the system generates an optimized task flow table, which records the task number, rewritten activation time, task sequence number, and resource binding field for each task, and outputs it for the scheduling optimization engine to call, supporting subsequent overall process rebalancing.

[0110] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications 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 protection scope of the present invention.

Claims

1. An AGI-based process processing system, characterized in that: The system includes: The semantic density parsing module obtains the task execution instruction log issued by the enterprise scheduler, extracts the process step name and resource station number fields from the task instructions, counts the number of keywords and number items, calculates the semantic density deviation, and merges task groups based on the semantic density deviation to generate a task fusion candidate list. The fusion node construction module extracts the fields of conflicting task resource call time and device usage from the task fusion candidate list, performs time segment intersection comparison and conflict field investigation, and generates a set of parallel task nodes. The process rhythm analysis module obtains the material tag field corresponding to the task in the parallel task node set, clusters all time data of process number to form standard process time, compares the current arrival time offset with the task, and generates a process rhythm offset list. The equipment task matching module extracts the equipment resource pool field of the parallel task nodes in the process rhythm offset list, combines the equipment operating status and response time period, establishes a task-equipment rhythm matching relationship, binds the equipment and sets an executable window, and generates the task flow processing result.

2. The AGI-based process processing system according to claim 1, characterized in that: The task fusion candidate list includes a semantic density deviation threshold, semantic correlation between adjacent tasks, and fusion task group index. The parallel task node set specifically includes resource call time coverage relationship, equipment conflict exclusion mark, and task synchronization start mark. The process rhythm offset list includes process rhythm offset, standard process time difference, and task combination rhythm comparison relationship. The task flow processing result includes task node binding structure, equipment rhythm matching relationship, and task executable time window.

3. The AGI-based process processing system according to claim 1, characterized in that: The semantic density parsing module includes: The field extraction submodule obtains the task execution instruction log issued by the enterprise scheduler, extracts the process step name and resource station number fields corresponding to each task instruction, identifies the number of keyword items corresponding to each content of the field, identifies the number of equipment number, station number and tool number in the resource station number field, and generates the field structure density based on the extracted field content and quantity distribution. The semantic computation submodule constructs the density vector of each task based on the field structure density using the AGI natural language processing model, calculates the difference between the density vectors of two adjacent tasks in the task sequence, obtains the semantic deviation between each group of tasks, and compares the difference with the set semantic fusion judgment threshold to obtain the semantic deviation degree between adjacent tasks. The task fusion identification submodule filters task pairs with semantic deviation less than the semantic fusion judgment threshold based on the semantic deviation of adjacent tasks, combines and classifies the task pairs according to their order in the original task sequence, identifies the semantic similarity tags between tasks in each combination, establishes a fusion task group index, and obtains a task fusion candidate list.

4. The AGI-based process processing system according to claim 3, characterized in that: To identify semantic similarity labels between tasks in each combination, the formula is: Sim calculates the keyword similarity between task A and task B. k (A,B), identify semantic similarity tags based on keyword similarity; Where A and B represent the sets of process step keywords in task A and task B, respectively, and x is each keyword in set A∩B, that is, the common keywords that appear in task A and task B. ∑ x∈A∩B min(f A (x),f B (x)) represents the sum of the smaller values ​​of the occurrence counts of all common keywords in the two tasks, ∑ x∈A∪B max(f A (x),f B (x) represents the sum of the largest number of occurrences of all keywords that appear in any task.

5. The AGI-based process processing system according to claim 1, characterized in that: The fusion node construction module includes: The time intersection comparison submodule obtains the resource call time field of each group of tasks in the task fusion candidate list, identifies the start and end time segments of each group of tasks, performs an intersection comparison on the time segments between all tasks, filters the intersection segments with continuous coverage relationship, calculates the coverage integrity ratio of each group of intersection segments, and obtains the resource call coverage set. The conflict field screening submodule obtains the device usage conflict field of each group of tasks based on the resource call coverage set, determines whether there are overlapping and duplicate items in the device number field between tasks, marks the task combination with the intersection mark in the conflict field, compares the coverage integrity ratio with the ratio threshold, obtains the task combination that meets the coverage requirements and has no intersection mark, and establishes a conflict-free fusion pair list. The task node generation submodule recombines the time period content and device number field of each group of tasks according to the list of conflict-free fusion pairs, constructs the node structure in the task scheduling relationship graph, configures the parallel identifier and task path index number for each combination item, and generates a set of parallel task nodes.

6. The AGI-based process processing system according to claim 1, characterized in that: The process rhythm analysis module includes: The material identification submodule obtains the material tag field corresponding to each task in the parallel task node set, identifies the association information between the material number and the process number in the material tag field, calls the task index number in the parallel task node, constructs the binding structure between the process number and the task node, establishes a material correspondence table for each task node, and generates task process mapping results. The time extraction submodule extracts the process record time field corresponding to the process number in the bound task according to the task process mapping result, gathers the same type of data according to the process number, performs time period classification and clustering on the process number time field, counts the frequency center position in the time period of each type of process, and obtains the standard process time interval. The rhythm comparison submodule compares the current arrival time field of each task in the parallel task node set with the standard process time interval, calculates the offset time between the current arrival time of each task and the center of the standard process time interval, and combines them according to the order of task nodes to establish a process rhythm offset list.

7. The AGI-based process processing system according to claim 1, characterized in that: The device task matching module includes: The resource status identification submodule extracts the equipment resource pool field corresponding to each parallel task node in the process rhythm offset list, identifies the current running status and response time field corresponding to the equipment number field in the resource pool, establishes a mapping set between equipment number and corresponding status information according to the equipment number order, calls the task node index information for matching and sorting, and generates equipment status operation analysis results. The rhythm adaptation mapping submodule compares the rhythm offset of each task with the response time interval of the corresponding device number based on the device status operation analysis results, filters the device numbers that meet the response time correspondence, and binds the filtered device numbers to the task nodes to generate node device adaptation results. Based on the node device adaptation results, the task path construction submodule extracts the response time period field corresponding to the device number in each group of binding results, sets the executable time interval for each task node, sorts the paths according to the sequential numbering of the task nodes, combines the device number with the task node and its executable window information, and establishes the task flow processing result.

8. The AGI-based process processing system according to claim 1, characterized in that: It also includes a scheduling window control module. Based on the task flow processing results, the scheduling window control module calculates the ratio of task delay time to equipment recovery cycle to generate a delay coefficient, adjusts the task activation time field, writes it into the task control table, and obtains the optimized task flow. The optimized task flow includes activation time adjustment, delay coefficient mapping relationship, and task control field update record.

9. The AGI-based process processing system according to claim 1, characterized in that: The scheduling window control module includes: The delay coefficient calculation submodule obtains the task node sequence number in the task flow processing result, extracts the previous task delay duration field corresponding to each task node and the recovery cycle field corresponding to the current device number, pairs the fields according to the task number order and calculates the corresponding numerical ratio to generate the task delay coefficient. The time field adjustment submodule calls the activation time field of each task node in the task flow processing result according to the task delay coefficient, calculates the time offset between the delay coefficient and the original activation time field, updates the field content according to the adjusted time of each task node, and obtains the activation time adjustment result. Based on the activation time adjustment result, the scheduling structure update submodule performs field value replacement operation on the activation field in the control table corresponding to each task node, writes the adjusted time to the corresponding activation field position, and reconstructs the scheduling path and resource binding information according to the task number order to establish the optimized task flow.

10. The AGI-based process processing system according to claim 9, characterized in that: The formula for calculating the delay factor is as follows: Where, λ i This represents the delay coefficient for task i. This indicates the delay time of the preceding task of task i. This represents the average delay time for all tasks. The recovery cycle for the device bound to the current task. This represents the average equipment recovery cycle.

Images