Automatic template management system for operation and maintenance tasks

Abstract

The invention relates to the technical field of operation and maintenance management, and discloses an operation and maintenance task-oriented automatic template management system, which comprises the following steps of: uniformly expressing an equipment tree in a fixed sequence of projects, branch companies, machine room sub-parts / sub-items and first-level / second-level / third-level equipment through node indexing-hierarchical coding-alignment to generate a stable index vector; cross-project naming differences and hierarchy missing or redundancy are converted into recoverable index offset, the same object is positioned consistently in different tree versions, candidate screening is not affected by hierarchy jitter, a type extraction module executes alias merging on a type table to obtain type mapping, then type vectors are extracted from a task list according to the type mapping, and the type vectors are extracted from the task list. According to the method, alias crossing and granularity inconsistency are converged to a unified key space, ambiguity is eliminated in an extraction link, and template matching is accurately restrained, so that near-synonym error locking is reduced, the accuracy and stability of template automatic locking are improved, and manual proofreading and backspacing are reduced.

Description

Technical Field

[0001] This invention relates to the field of operation and maintenance management technology, and more specifically, to an automated template management system for operation and maintenance tasks. Background Technology

[0002] In typical operation and maintenance management systems, the management terminal is used to dispatch tasks and configure the approval process, while the mobile terminal is used to accept orders and fill in electronic reports. The system needs to select a report template and generate an archived version based on the task type and equipment information. The application objects cover a variety of equipment such as high-voltage cabinets, low-voltage cabinets, transformers, generators, ATS and UPS. The equipment information is represented by a tree-like archive of first-level, second-level, and third-level equipment in the project, branch, computer room, sub-project, and item. The system operation is constrained by the parallel operation of multiple projects and the growth of the template library, as well as by mobile network fluctuations and offline scenarios, and by terminal interaction latency and processing capabilities. Existing technologies usually adopt the method of category merging and keyword filtering, combined with hierarchical mapping constraints and historical behavior statistics, and reduce fluctuations through time smoothing and filtering.

[0003] However, in existing technologies, there are objectively existing unstable factors such as naming differences and mapping changes of device tree nodes in different projects and stages, and alias overlap and granularity inconsistency of task types in multiple departments and scenarios. These factors together weaken the consistency of category merging and hierarchical mapping and cause fluctuations in template set inference, thus affecting the stability of automatic selection results. Therefore, the technical problem that needs to be solved is how to improve the stability of device tree node recognition and reduce the ambiguity rate of task types while achieving the accuracy of automatic template locking.

[0004] In view of this, the present invention proposes an automated template management system for operation and maintenance tasks to solve the above problems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, this invention provides an automated template management system for operation and maintenance tasks.

[0006] To achieve the above objectives, the present invention provides the following technical solution: Firstly, an automated template management system for operation and maintenance tasks is provided, including: The device tree is located based on the task list to obtain the node index. The task list is a sequence of task items dispatched by the management terminal, and the device tree is a hierarchical data structure with the project as the root. Hierarchical encoding is performed based on node indexes to obtain hierarchical vectors. Alignment processing is then performed on the device tree based on the hierarchical vectors to obtain stable index vectors. The hierarchical vectors represent hierarchical features in the order of project, branch, data center division, data center item, first-level equipment, second-level equipment, and third-level equipment. Obtain the type table, perform alias merging on the type table to obtain the type mapping, and perform type extraction on the task list based on the type mapping to obtain the type vector. Here, the type table refers to the key-value mapping of the task type directory. The template library is filtered based on the stable index vector to obtain a candidate set. The template library is a collection of report templates, and a report template refers to the structured descriptive data used to generate electronic reports. The candidate set is matched based on the type vector to obtain the locked set, which is the set of indices of the target template.

[0007] In some embodiments, the method for performing location processing on the device tree based on the task list to obtain the node index includes: Based on the task list, perform field extraction processing to obtain the project number column; based on the task list, perform field extraction processing to obtain the device number column; based on the task list, perform sequential encoding processing to obtain the task sequence number column. Based on the project number column and the device tree, perform path acquisition processing to obtain the path index table. Based on the path index table and the device number column, perform key construction processing to obtain the hierarchical key sequence. Based on the hierarchical key sequence and the task sequence number column, perform key normalization processing to obtain the normalized key sequence. Based on the normalized key sequence and the device tree, perform node matching processing to obtain a candidate node set. Based on the candidate node set and the path index table, perform conflict annotation processing to obtain an ambiguity record table. Based on the candidate node set and the ambiguous record table, a disambiguation and merging process is performed to obtain a stable node set. Based on the stable node set, an index summarization process is performed to obtain the node index.

[0008] In some embodiments, the method for performing hierarchical encoding based on node indices to obtain hierarchical vectors includes: Based on the node index and device tree, perform path parsing to obtain the parsed path table; based on the parent-child relationship of the device tree, perform parent-child extraction to obtain the parent-child mapping table. Based on the parsing path table and the parent-child mapping table, perform layer encoding processing to obtain the layer vector. Based on the parsing path table, perform layer depth calculation processing in the order of project, branch company, computer room branch, computer room sub-item, first-level equipment, second-level equipment, and third-level equipment to obtain the layer depth vector. Based on the layer position vector and layer depth vector, perform gap detection processing to obtain the gap matrix. Based on the gap matrix and the parent-child mapping table, perform cross-level verification processing to obtain the cross-level marker set. Based on the over-level tag set and the parsing path table, perform order adjustment processing to obtain the adjustment vector. Based on the adjustment vector and the node index, perform alignment calculation processing to obtain the initial alignment vector. Based on the initial alignment vector and the layer vector, an encoding fusion process is performed to obtain the alignment encoding vector. Based on the alignment encoding vector and the adjustment vector, a dimension compression process is performed to obtain the layer vector.

[0009] In some embodiments, the method for obtaining a layer vector by performing layer-level encoding processing based on the parsed path table and parent-child mapping table includes: Based on the parsing path table and the parent-child mapping table, perform layer order extraction processing to obtain a layer name list. Based on the layer name list and the parsing path table, perform sequence mapping processing to obtain a layer sequence list. Based on the layer sequence table and the parent-child mapping table, perform cross-level detection processing to obtain the layer marker set. Based on the layer marker set and the layer sequence table, perform gap filling processing to obtain the filling list. Based on the layer sequence list and the padding list, perform bit weight generation processing to obtain the bit weight vector. Based on the bit weight vector and the layer sequence list, perform encoding calculation processing to obtain the layer bit encoding vector. Based on the layer encoding vector and the layer name list, perform dimension reshaping to obtain the layer vector.

[0010] In some embodiments, the method for obtaining a hierarchical vector by performing dimensionality compression based on the aligned encoding vector and the adjusted vector includes: Based on the alignment encoding vector and the adjustment vector, range statistics processing is performed to obtain the component range table. Based on the alignment encoding vector and the component range table, threshold generation processing is performed to obtain the threshold list. Based on the adjustment vector and the threshold list, gating calculation is performed to obtain the gating weights. Based on the aligned encoding vector and the gating weights, weight superposition is performed to obtain the superimposed vector. Based on the superposition vector and component range table, perform principal axis selection processing to obtain the principal axis index. Based on the principal axis index and the superposition vector, perform component projection processing to obtain the projection matrix. Here, the principal axis is the linear projection direction that maximizes the variance of the superposition vector. Based on the projection matrix and the gate weights, perform dimension aggregation to obtain an aggregated vector. Based on the aggregated vector and the principal axis index, perform scale compression to obtain a hierarchy vector.

[0011] In some embodiments, the method for performing alias merging on the type table to obtain the type mapping includes: Based on the type table, perform term extraction processing to obtain a term list; based on the task list, perform type counting processing to obtain a frequency vector. Based on the term list and type table, perform alias indexing to obtain the alias index. Based on the alias index and frequency vector, perform conflict detection to obtain the conflict matrix. Based on the conflict matrix and frequency vector, conflict weight processing is performed to obtain the conflict weights. Based on the alias index and conflict weights, cluster merging processing is performed to obtain the merged index. Based on the merged index and the term list, a mapping generation process is performed to obtain a mapping list. Based on the mapping list and the type table, an encoding normalization process is performed to obtain the type mapping.

[0012] In some embodiments, the method for obtaining conflict weights by performing conflict weight processing based on the conflict matrix and frequency vector includes: Based on the frequency vector, normalization is performed to obtain the normalized frequency; based on the collision matrix, effective masking is performed to obtain the matrix mask. Based on the collision matrix and matrix mask, intensity calculation is performed to obtain the collision metric. Based on the collision metric and normalized frequency, frequency modulation is performed to obtain the modulation vector. Based on the modulation vector, a sliding window smoothing process is performed to obtain a smoothed vector. Based on the smoothed vector and the matrix mask, an anomaly suppression process is performed to obtain a suppressed vector. Based on the suppression vector and the normalized frequency, coefficient mapping is performed to obtain the weight coefficients. Based on the weight coefficients, threshold shaping is performed to obtain the conflict weights.

[0013] In some embodiments, the method for obtaining type mappings by performing encoding normalization processing based on the mapping list and the type table includes: Based on the mapping list and type table, perform encoding extraction processing to obtain the encoding dictionary table; based on the type table, perform priority generation processing to obtain the priority sequence. Based on the mapping list and the encoding dictionary table, perform index building processing to obtain the encoding index table. Based on the encoding index table and the priority sequence, perform conflict identification processing to obtain the conflict list table. Based on the conflict list and priority sequence, priority adjudication is performed to obtain the target type set. Based on the target type set and the coding index table, consistency verification is performed to obtain the consistency vector. Based on the consistency vector and the target type set, a mapping generation process is performed to obtain a mapping index table. Based on the mapping index table and the encoding dictionary table, an encoding normalization process is performed to obtain the type mapping.

[0014] In some embodiments, the method for performing filtering on the template library based on the stable index vector to obtain a candidate set includes: Based on the stable index vector and template library, perform attribute extraction processing to obtain template metadata, and perform field mapping processing based on the stable index vector to obtain the matching condition set; Based on the template metadata and the matching condition set, perform device filtering to obtain a device filter set; based on the template metadata and the matching condition set, perform type filtering to obtain a type filter set. Based on the equipment filter set and the type filter set, perform set intersection and union processing to obtain a preliminary selection list. Based on the preliminary selection list and the template library, perform validity verification processing to obtain a valid list. Based on the valid list and the matching condition set, perform index aggregation processing to obtain the candidate set.

[0015] In some embodiments, the method for performing matching processing on the candidate set based on the type vector to obtain the locked set includes: Based on the type vector and the candidate set, feature expansion processing is performed to obtain a similarity matrix. Based on the type vector, weight generation processing is performed to obtain a matching priority sequence. Based on the similarity matrix and the matching priority sequence, a threshold filtering process is performed to obtain a weight threshold table. Based on the weight threshold table and the candidate set, an index filtering process is performed to obtain a matching index table. Based on the matching index table and the candidate set, perform index extraction processing to obtain the initial template set. Based on the initial template set and the type vector, perform conflict detection processing to obtain the matching conflict table. Based on the matching conflict table and the matching priority sequence, perform priority adjudication processing to obtain the adjudication index table. Based on the adjudication index table and the initial template set, perform index summarization processing to obtain the lock set.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: The hierarchical alignment module of this invention forms a stable index vector through "node index - hierarchical encoding - alignment", which enables the device tree to be uniformly expressed in a fixed order of project, branch, computer room division / sub-item, first-level equipment, second-level equipment, and third-level equipment. Naming differences, hierarchical missing and redundancy caused across projects or stages are converted into recoverable index offsets. The same physical object obtains consistent positioning on different tree versions. Candidate set screening is no longer affected by hierarchical jitter. The node positioning hit rate and traceability are improved, the number of false screenings and backtracking rematches is reduced, and the overall output is more stable. The type extraction module of this invention first performs alias merging on the type table to obtain a type mapping, and then extracts type vectors from the task list using the type mapping. Alias ​​intersections and granularity inconsistencies of task types are converged to a unified key space. The source of ambiguity is resolved in the extraction stage, the template matching constraint on the candidate set is more precise, the mis-locking of near-synonyms is significantly reduced, and the type interpretation remains consistent across multiple departments and scenarios. Ultimately, this is reflected in the improved accuracy and stability of automatic template locking in the locking set stage, and the reduced need for manual proofreading and rollback reruns. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the structure of an automated template management system for operation and maintenance tasks according to the present invention; Figure 2 This is a flowchart illustrating an automated template management method for operation and maintenance tasks according to the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains. The terms “first,” “second,” and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as “comprising” or “including” mean that the element or object preceding the term covers the element or object listed after the term and its equivalents, without excluding other elements or objects. Terms such as “connection” or “linked” are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect.

[0020] Example 1 Please see Figure 1 As shown, this embodiment discloses an automated template management system for operation and maintenance tasks, including: Node positioning module: used to perform positioning processing on the device tree according to the task list to obtain the node index. The task list is a sequence of task items dispatched by the management terminal, and the device tree refers to a hierarchical data structure with the project as the root. Methods for performing location processing on the device tree based on the task list to obtain node indexes include: Based on the task list, perform field extraction processing to obtain the project number column; based on the task list, perform field extraction processing to obtain the device number column; based on the task list, perform sequential encoding processing to obtain the task sequence number column. Based on the project number column and the device tree, perform path acquisition processing to obtain the path index table. Based on the path index table and the device number column, perform key construction processing to obtain the hierarchical key sequence. Based on the hierarchical key sequence and the task sequence number column, perform key normalization processing to obtain the normalized key sequence. Based on the normalized key sequence and the device tree, perform node matching processing to obtain a candidate node set. Based on the candidate node set and the path index table, perform conflict annotation processing to obtain an ambiguity record table. Based on the candidate node set and the ambiguous record table, a disambiguation and merging process is performed to obtain a stable node set. Based on the stable node set, an index summarization process is performed to obtain the node index.

[0021] In this embodiment, the task list is a sequence of task items dispatched by the management terminal, the project number column is a sequence of project numbers extracted from the task list by field, the device number column is a sequence of device numbers extracted from the task list by field, the task sequence number column is a sequence number obtained by performing sequential encoding processing on the task list according to time or import order, and the device tree is a hierarchical data structure with the project as the root and containing hierarchical relationships such as project, branch, computer room, division, computer room, item, first-level device, second-level device, third-level device, etc.

[0022] It should be noted that the path index table is a set of hierarchical path sequences parsed from the device tree based on the project number column; the hierarchical key sequence is a combination key sequence of hierarchical position and device number constructed based on the path index table and the device number column; the normalized key sequence is a key sequence obtained by normalizing the hierarchical key sequence and the task sequence number column, the purpose of which is to form comparable and transferable stable key values ​​under the same project or the same level for subsequent matching and disambiguation processing; the candidate node set is a set of node numbers obtained by performing node matching processing in the device tree based on the normalized key sequence; the ambiguity record table is a set of one-to-many or many-to-one conflict entries recorded in the candidate node set and the path index table; the stable node set is a set of nodes obtained by performing disambiguation and merging processing on the candidate node set according to the ambiguity record table; and the node index is an index result that can directly locate device tree nodes obtained by performing index summarization processing on the stable node set.

[0023] Understandably, the field extraction process converts the project and device identifiers in the task list into computable project number and device number columns. The sequential encoding process provides a stable comparison benchmark in the task sequence dimension. The path acquisition process generates a path index table based on the parent-child relationship of the device tree to lock the topological position of the project to the target level. The key construction process binds the path position with the device number to form a hierarchical key sequence. The key normalization process unifies the key format under the same level and project to eliminate time drift and writing differences to facilitate consistent matching in the future.

[0024] In this embodiment, the node matching process retrieves the corresponding node in the device tree based on the normalized key sequence to form a candidate node set. The conflict marking process registers the case where one normalized key hits multiple nodes or multiple normalized keys hit the same node as an ambiguity record table. The disambiguation and merging process eliminates conflicts to obtain a stable node set while maintaining path consistency and sequence number consistency. The index summarization process sorts the stable node set by task sequence number column and generates a node index for direct use in subsequent hierarchical coding and template filtering stages.

[0025] For example, the task list contains three tasks with project number columns P01, P01, and P02, equipment number columns D11, D12, and D21, and task sequence number columns 1, 2, and 3. In the equipment tree, the path corresponding to project P01 is P01, Branch Company A, Data Center Branch A1, and Data Center Sub-item A1a, and the path corresponding to project P02 is P02, Branch Company B, Data Center Branch B1, and Data Center Sub-item B1a. The above path index table represents the hierarchical position of each task and serves as the path input for constructing the subsequent hierarchical key sequence.

[0026] Furthermore, based on the path index table and the device number column, hierarchical key sequences P01A1aD11, P01A1aD12, and P02B1aD21 are generated. Based on the hierarchical key sequences and the task sequence number column, normalized key sequences K01, K02, and K03 are generated. Based on the normalized key sequences, candidate node sets {n110, n111, n210} are obtained by searching the device tree. If there are duplicate devices in the device tree, causing K01 to hit n110 and n112, it is registered in the ambiguity record table as {n110, n112} corresponding to K01 to provide conflict details.

[0027] For example, the disambiguation and merging process consistently retains n110 and n111 after comparing the hierarchical positions of the path index table and the candidate node set, and removes n112 which is inconsistent with the path, to obtain a stable node set {n110, n111, n210}. The index summarization process generates a node index set {(1, n110), (2, n111), (3, n210)} according to the task sequence. This node index is used to directly locate device tree nodes in subsequent hierarchical encoding processes and maintain a one-to-one correspondence between tasks and nodes to reduce the impact of path drift in the calculation.

[0028] The hierarchical alignment module is used to perform hierarchical encoding processing based on node indexes to obtain hierarchical vectors. Based on the hierarchical vectors, the device tree is aligned to obtain stable index vectors. The hierarchical vectors represent hierarchical features in the order of project, branch, data center division, data center item, first-level equipment, second-level equipment, and third-level equipment. Methods for performing hierarchical encoding based on node indices to obtain hierarchical vectors include: Based on the node index and device tree, perform path parsing to obtain the parsed path table; based on the parent-child relationship of the device tree, perform parent-child extraction to obtain the parent-child mapping table. Based on the parsing path table and the parent-child mapping table, perform layer encoding processing to obtain the layer vector. Based on the parsing path table, perform layer depth calculation processing in the order of project, branch company, computer room branch, computer room sub-item, first-level equipment, second-level equipment, and third-level equipment to obtain the layer depth vector. Based on the layer position vector and layer depth vector, perform gap detection processing to obtain the gap matrix. Based on the gap matrix and the parent-child mapping table, perform cross-level verification processing to obtain the cross-level marker set. Based on the over-level tag set and the parsing path table, perform order adjustment processing to obtain the adjustment vector. Based on the adjustment vector and the node index, perform alignment calculation processing to obtain the initial alignment vector. Based on the initial alignment vector and the layer vector, an encoding fusion process is performed to obtain the alignment encoding vector. Based on the alignment encoding vector and the adjustment vector, a dimension compression process is performed to obtain the layer vector.

[0029] In this embodiment, the parsed path table is a set of path sequences obtained by tracing back level by level in the device tree according to the node index; the parent-child mapping table is a set of pairwise relationships between parent nodes and child nodes in the device tree; the layer vector is a fixed-length vector obtained by sequentially encoding the project, branch, computer room, sub-item, first-level equipment, second-level equipment, and third-level equipment in the path; the layer depth vector is the layer depth sequence of the corresponding path; the gap matrix is ​​a Boolean matrix indicating whether each task is missing at each layer; the skip-level marker set is a set of task markers that have skipped levels or crossed layers; the adjustment vector is a set of displacement amounts after order correction for skipping levels or gaps; the initial alignment vector is a set of layer index sequences aligned in the path space according to the adjustment vector; the alignment encoding vector is a numerical expression obtained by fusing the initial alignment vector and the layer vector; and the layer vector is the final fixed-length vector obtained by dimensional compression of the alignment encoding vector.

[0030] It should be noted that the path resolution process uses node indexes to read the hierarchical chain from the root to the target node in the device tree to form a resolution path table. Its purpose is to provide a consistent hierarchical context for subsequent layer encoding and layer depth calculation. The parent-child extraction process collects the parent-child mapping table from the topology of the device tree as a comparison benchmark for cross-level verification. Both ensure that all subsequent consistency judgments rely on the same topology source to avoid deviations caused by multi-source paths.

[0031] For example, using the node index set {(1, n110), (2, n111), (3, n210)} from Example 1, the device tree parsing yields the parsing path table, as shown in Table 1: Table 1 Furthermore, the layer encoding process maps each layer in the parsed path table to numerical bits in a fixed order and concatenates them into a layer vector. The layer depth calculation process counts the effective layers of each path to form a layer depth vector. The purpose is to provide a dual-channel expression that includes both layer order and layer depth so that gaps and skips can be objectively determined and the numerical execution of alignment calculation can be supported.

[0032] For example, the following layer position vector and layer depth vector are generated for the three paths in the table above. The layer position vector uses a seven-bit code to correspond to the layer position from the project to the third-level equipment, and the layer depth vector records the number of valid layers, as shown in Table 2: Table 2 In this embodiment, the gap detection process generates a gap matrix based on the layer position vector and layer depth vector to identify any missing layer. The skip-level verification process uses a parent-child mapping table to check whether there is a skip-level phenomenon that crosses the parent-child relationship in the path and produces a skip-level marker set. The purpose is to make the path anomalies explicit in a structured way to facilitate subsequent order adjustment and alignment calculations and avoid ambiguity caused by direct fuzzy correction on the original path.

[0033] For example, if Task 2 is missing at the secondary device layer, resulting in the sixth position of the layer vector being empty, then the gap matrix is ​​recorded as 1; otherwise, it is 0. Simultaneously, if Task 3 is directly connected to the secondary device E2a under project P02 without passing through the primary device E2, then the skip-level tag set contains the tag entries for Task 3, as shown in Table 3 below: Table 3 It should be noted that the sequence adjustment process corrects the displacement of abnormal layers based on the cross-level marker set to obtain the adjustment vector. The alignment calculation process then shifts or fills in the layer index to generate the initial alignment vector. The purpose is to obtain a set of cross-task comparable sequence coordinates without changing the actual parent-child structure of the device tree, so as to provide a stable reference for subsequent encoding fusion.

[0034] For example, if Task 2 has a gap in the sixth position, the vector is adjusted to [0, 0, 0, 0, 0, +1, 0] to fill the gap during encoding. After alignment calculation, the initial alignment vector is consistent with the layer position. However, Task 3 needs to insert a placeholder in the fifth position due to skipping levels, so its initial alignment vector is [2, 1, 1, 1, 1, 1, 1]. On this basis, the encoding fusion process is performed to superimpose the initial alignment vector and the layer position vector with position weights to obtain the alignment encoding vector, so as to retain the dual information of layer position and alignment.

[0035] Understandably, dimensionality compression performs gated aggregation and scale compression on the aligned encoding vector to obtain a hierarchical vector. The hierarchical vector, as a fixed-length structure, serves as a numerical representation of the stable index vector in the subsequent template selection and matching process. Its technical effect is to enhance the stability of device tree node recognition and reduce encoding ambiguity caused by hierarchical irregularities through chain constraints of path parsing and cross-level verification, thereby providing a consistent upstream input for automatic template locking.

[0036] The methods for obtaining the layer vector by performing layer-level encoding processing based on the parsed path table and parent-child mapping table include: Based on the parsing path table and the parent-child mapping table, perform layer order extraction processing to obtain a layer name list. Based on the layer name list and the parsing path table, perform sequence mapping processing to obtain a layer sequence list. Based on the layer sequence table and the parent-child mapping table, perform cross-level detection processing to obtain the layer marker set. Based on the layer marker set and the layer sequence table, perform gap filling processing to obtain the filling list. Based on the layer sequence list and the padding list, perform bit weight generation processing to obtain the bit weight vector. Based on the bit weight vector and the layer sequence list, perform encoding calculation processing to obtain the layer bit encoding vector. Based on the layer encoding vector and the layer name list, perform dimension reshaping to obtain the layer vector.

[0037] Methods for obtaining hierarchical vectors by performing dimensionality compression based on the aligned encoding vector and the adjusted vector include: Based on the alignment encoding vector and the adjustment vector, range statistics processing is performed to obtain the component range table. Based on the alignment encoding vector and the component range table, threshold generation processing is performed to obtain the threshold list. Based on the adjustment vector and the threshold list, gating calculation is performed to obtain the gating weights. Based on the aligned encoding vector and the gating weights, weight superposition is performed to obtain the superimposed vector. Based on the superposition vector and component range table, perform principal axis selection processing to obtain the principal axis index. Based on the principal axis index and the superposition vector, perform component projection processing to obtain the projection matrix. Here, the principal axis is the linear projection direction that maximizes the variance of the superposition vector. Based on the projection matrix and the gate weights, perform dimension aggregation to obtain an aggregated vector. Based on the aggregated vector and the principal axis index, perform scale compression to obtain a hierarchy vector.

[0038] In this embodiment, the layer name list is a sequence of layer names extracted from the parsing path table, such as project branch, company, computer room, division, computer room, item, first-level equipment, second-level equipment, and third-level equipment. The layer sequence list is a sequence of sequential indexes formed by mapping the layer name list to specific positions in the parsing path table. The layer position tag set is a set of tags obtained by detecting skipping levels and gaps based on the parent-child mapping table. The filling list is a sequence of placeholder or filler symbols generated for the gap position. The position weight vector is a weight sequence allocated according to the importance of the layer position. The layer position encoding vector is a numerical expression obtained by combining the layer sequence list and the position weight vector bit by bit. The layer position vector is a fixed-length vector obtained by performing dimension reshaping on the layer position encoding vector.

[0039] It should be noted that the purpose of the layer order extraction process is to uniformly project the parsing path table onto the coordinate system of the same layer name list. The order mapping process ensures the comparability of layer order lists of different projects. The skip detection process identifies cross-layer jumps based on the parent-child mapping table and generates a layer position label set. The gap filling process makes the missing layers explicit and forms a computable filling list. The position weight generation process assigns greater weights to important layers to improve the sensitivity to key layers. The encoding calculation process combines the layer order and weights to form a layer position encoding vector. The dimension shaping process regularizes the encoding results into fixed dimensions to facilitate subsequent alignment and aggregation with other tasks.

[0040] For example, using the node index set {(1, n110), (2, n111), (3, n210)} from Example 1, the parsed path table corresponds to three paths, and the layer name list is fixed as {Project, Branch, Data Center Division, Data Center Item, Level 1 Equipment, Level 2 Equipment, Level 3 Equipment}. Based on this, the layer sequence is shown in Table 4: Table 4 Furthermore, based on the parent-child mapping table, skip-level detection and gap detection are performed. If Task 2 has a gap in the secondary equipment, the layer marker set records it as a gap. If Task 3 involves a skip-level from the project to the secondary equipment, the layer marker set records it as a skip-level. A filling list is generated for the gap locations for subsequent encoding calculations. An example is shown in Table 5. Table 5 Understandably, the bit weight generation process sets bit weights for each layer, such as [64, 32, 16, 8, 4, 2, 1]. The encoding calculation process multiplies the layer sequence list with the bit weight vector bit by bit or maps them bit by bit to obtain the layer bit encoding vector. The dimension shaping process normalizes the encoding vectors of different tasks to the same length for comparison and downstream operations. Examples are shown in Table 6. Table 6 In this embodiment, the component range table is a statistical range record table for each component of the aligned encoding vector, the threshold list is a segmented threshold sequence generated based on the component range table, the gating weight is a weight vector calculated based on the adjustment vector and the threshold list, the superposition vector is a vector after the aligned encoding vector is weighted by the gating weight, the principal axis index is the projection direction index selected based on the maximum variance criterion of the superposition vector, the projection matrix is ​​the linear projection result of the superposition vector on the principal axis index, the aggregation vector is the vector obtained by aggregating the projection matrix according to the gating weight, and the hierarchical vector is the fixed-length result after scaling the aggregated vector.

[0041] It should be noted that range statistical processing makes the components of different tasks have comparable intervals, threshold generation processing discretizes the continuous range into a computable list of thresholds, gating calculation processing assigns higher gating weights to sensitive components based on the adjustment vector to enhance the response to the adjusted layer, weight superposition processing obtains a superimposed vector to integrate encoding and gating information, principal axis selection processing determines the most discriminative direction based on the maximum variance criterion, component projection processing maps the superimposed vector to this direction to form a projection matrix, and dimension aggregation and scale compression obtain a uniformly sized hierarchical vector while maintaining discriminability for subsequent template selection and matching.

[0042] For example, the alignment encoding vectors are: Task 1 [64, 32, 16, 8, 4, 2, 1], Task 2 [64, 32, 16, 8, 0, 2, 1], and Task 3 [128, 32, 16, 8, 4, 2, 1]. The component range table records the minimum and maximum values ​​of each component as follows. The threshold list generates corresponding threshold intervals by dividing the data into three segments, as shown in Table 7. Table 7 Furthermore, assuming the adjustment vector has a positive displacement in the 6th component of Task 2 to enhance the response of that layer, the gating calculation process yields gating weights such as [1, 1, 1, 1, 1, 1, 1, 1] for Task 1, [1, 1, 1, 1, 1, 1.5, 1] ​​for Task 2, and [1, 1, 1, 1, 1, 1, 1] for Task 3. The weight superposition process forms the superimposed vector as shown in Table 8. Table 8 Understandably, the principal axis selection process obtains the principal axis index as 1 when the variance of component 1 is the largest. The component projection process projects the three superimposed vectors onto the principal axis index to obtain the projection matrix column vector as [64, 64, 128]. The dimension aggregation process combines the gating weights to normalize and aggregate the projections of each task to obtain the aggregate vector as [64, 66.0, 128]. The scale compression process linearly scales the aggregate vector to a fixed interval such as [0, 1] to form a hierarchical vector while maintaining the difference between tasks.

[0043] Compared to existing technologies that directly involve single-level encoding in subsequent matching, the aforementioned level encoding and dimension compression link transforms hierarchical semantics into stable and comparable hierarchical vectors through cross-level explicitness, gap filling, position weight control, and gated compression. It exhibits better anti-ambiguity and alignment stability in scenarios with irregular levels, naming differences, and path changes, thereby providing consistent and robust input for subsequent template selection and matching.

[0044] Type extraction module: used to obtain the type table, perform alias merging on the type table to obtain the type mapping, and perform type extraction on the task list according to the type mapping to obtain the type vector. Here, the type table refers to the key-value mapping of the task type name directory. Methods for performing alias merging on type tables to obtain type mappings include: Based on the type table, perform term extraction processing to obtain a term list; based on the task list, perform type counting processing to obtain a frequency vector. Based on the term list and type table, perform alias indexing to obtain the alias index. Based on the alias index and frequency vector, perform conflict detection to obtain the conflict matrix. Based on the conflict matrix and frequency vector, conflict weight processing is performed to obtain the conflict weights. Based on the alias index and conflict weights, cluster merging processing is performed to obtain the merged index. Based on the merged index and the term list, a mapping generation process is performed to obtain a mapping list. Based on the mapping list and the type table, an encoding normalization process is performed to obtain the type mapping.

[0045] Based on the conflict matrix and frequency vector, conflict weight processing is performed to obtain the conflict weights. Methods for obtaining conflict weights include: Based on the frequency vector, normalization is performed to obtain the normalized frequency; based on the collision matrix, effective masking is performed to obtain the matrix mask. Based on the collision matrix and matrix mask, intensity calculation is performed to obtain the collision metric. Based on the collision metric and normalized frequency, frequency modulation is performed to obtain the modulation vector. Based on the modulation vector, a sliding window smoothing process is performed to obtain a smoothed vector. Based on the smoothed vector and the matrix mask, an anomaly suppression process is performed to obtain a suppressed vector. Based on the suppression vector and the normalized frequency, coefficient mapping is performed to obtain the weight coefficients. Based on the weight coefficients, threshold shaping is performed to obtain the conflict weights.

[0046] In this embodiment, the term list is a sequence of type names extracted from the type table, the frequency vector is a vector that counts the number of times each type or its aliases appear on the same day or in the same batch according to the task list, the alias index is an index structure that establishes a one-to-one or many-to-one correspondence between the term list and the standard names in the type table, the conflict matrix is ​​a Boolean matrix showing the mutual exclusion or overlap between aliases, the conflict weight is a weight coefficient calculated by combining the conflict matrix and the frequency vector, the merge index is the aggregation result of aliases to standard names obtained under the constraint of conflict weight, the mapping list is a key-value table expanded according to the merge index, and the type mapping is the final mapping result after uniformly encoding the mapping list.

[0047] Understandably, the term extraction process transforms the natural language type table into a computable list of terms and an index space; the type counting process provides a frequency vector to reflect recent usage intensity; the alias indexing process categorizes terms like "inspection" and "daily inspection" into the candidate set of "inspection report"; the conflict detection process generates a conflict matrix based on literal similarity or historical misuse relationships; the conflict weighting process weights the conflict matrix according to the frequency vector to highlight high-frequency and easily confused alias pairs; the clustering and merging process outputs a merged index under the guidance of the conflict weights; and the mapping generation and encoding normalization process ensures that the final type mapping is unique across the entire database and can be directly used for subsequent matching.

[0048] For example, the example type table and task list statistics are as follows. The term list and frequency vector are given in the table, which serves as the input for conflict detection and weight calculation. An example is shown in Table 9: Table 9 Furthermore, the alias index establishes a candidate set along the standard name dimension, and the conflict matrix records word pairs that are easily confused in the same task scenario, where 1 indicates conflict or high confusion, and 0 indicates no conflict, as shown in Tables 10 and 11: Table 10 Table 11 It should be noted that the conflict weighting process first linearly normalizes the frequency vector to obtain the normalized frequency to eliminate batch differences. Then, a matrix mask is generated based on the conflict matrix to focus on the real conflict location. Subsequently, the conflict metric is calculated and frequency modulation is performed using the normalized frequency to obtain the modulation vector. The suppression vector is obtained through sliding window smoothing and anomaly suppression. Finally, the suppression vector is mapped to the weight coefficient and thresholded to obtain the conflict weight for subsequent clustering and merging.

[0049] For example, after the frequency vector is normalized, the normalized frequencies are {inspection 0.90, daily inspection 0.35, maintenance 0.45, upkeep 0.25, fault 0.55, repair 0.30}. After applying the matrix mask to the conflict matrix, only the diagonal elements of the table above are retained. For each pair of aliases, the conflict metric is calculated and multiplied by the normalized frequency to obtain the modulation vector. Then, a sliding window of length three is applied to the modulation vector to obtain the smoothing vector. The single-point anomaly handling bit is set to zero to form a suppression vector. This suppression vector is used to generate weight coefficients to give conflict weights higher than the threshold in the three groups of "inspection-daily inspection", "maintenance-upkeep", and "fault-repair".

[0050] In this embodiment, the clustering merging process merges synonyms in the candidate set of each standard name according to the conflict weight from high to low to obtain a merge index. The mapping generation process expands the merge index into a mapping list and maintains a one-to-one mapping relationship. The encoding normalization process assigns a unique type code to each standard name and outputs a type mapping. This type mapping is directly called by the subsequent matching process to generate a matching priority sequence and a weight threshold table.

[0051] For example, based on the above weighting results, the merged index is {Inspection Report ← {Inspection, Daily Inspection}, Maintenance Report ← {Maintenance, Maintenance}, Handover Start-up Report ← {Handover Start-up}, Fault Report ← {Fault, Repair}, Other Report ← {Other}}. The mapping list expands accordingly to {Inspection → Inspection Report, Daily Inspection → Inspection Report, Maintenance → Maintenance Report, Maintenance → Maintenance Report, Handover Start-up → Handover Start-up Report, Fault → Fault Report, Repair → Fault Report, Other → Other Report}. The encoding is normalized and assigned type codes such as {Inspection Report T01, Maintenance Report T02, Handover Start-up Report T03, Fault Report T04, Other Report T05}, forming a type mapping to ensure consistency and computability across the entire database.

[0052] Compared to the static merging method based solely on word similarity in existing technologies, the above-mentioned link links the conflict matrix and frequency vector to form conflict weights and explicitly uses these weights in cluster merging. Therefore, it can effectively reduce type ambiguity and output stable type mappings in high-frequency mixed-use scenarios, while retaining structured intermediate quantities throughout the process for review and engineering implementation.

[0053] The methods for obtaining type mappings by performing encoding normalization processing based on the mapping list and type table include: Based on the mapping list and type table, perform encoding extraction processing to obtain the encoding dictionary table; based on the type table, perform priority generation processing to obtain the priority sequence. Based on the mapping list and the encoding dictionary table, perform index building processing to obtain the encoding index table. Based on the encoding index table and the priority sequence, perform conflict identification processing to obtain the conflict list table. Based on the conflict list and priority sequence, priority adjudication is performed to obtain the target type set. Based on the target type set and the coding index table, consistency verification is performed to obtain the consistency vector. Based on the consistency vector and the target type set, a mapping generation process is performed to obtain a mapping index table. Based on the mapping index table and the encoding dictionary table, an encoding normalization process is performed to obtain the type mapping.

[0054] In this embodiment, the encoding dictionary table is a lookup table from the standard types extracted from the type table to the unique codes; the priority sequence is a type priority sequence generated according to business rules and the effective time sequence; the encoding index table is an index structure that maps the terms in the mapping list to the target codes according to the encoding dictionary table; the conflict list table is a set of conflict entries generated when the same term corresponds to multiple target codes; the target type set is a set of standard types retained after priority adjudication; the consistency vector is a Boolean vector that verifies the consistency between the target type set and the encoding index table; the mapping index table is a set of paired indexes from the target type to the code; and the type mapping is the final key-value mapping result obtained after normalizing the mapping index table.

[0055] It should be noted that the encoding extraction process extracts the established standard names and historical codes from the type table to form an encoding dictionary table. The priority generation process provides a priority sequence based on the effective time and the system's built-in order without introducing subjective evaluation. The index construction process aligns the mapping list to the encoding space so that conflicts can be uniformly identified according to the encoding dimension. The conflict identification process locates one-to-many or many-to-one entries at the encoding level to generate a conflict list table. The priority adjudication process performs deterministic selection of conflict entries according to the priority sequence to obtain the target type set. The consistency verification process uses a consistency vector to explicitly mark whether it is consistent with the encoding index table. The mapping generation and encoding normalization processes ensure that the output type mapping is unique and comparable in the entire database and can be directly called for subsequent matching.

[0056] For example, the example type table and historical codes are as follows. The code dictionary table records the relationship between standard names and codes. The priority sequence is generated according to "fault reports take precedence over inspection reports, which take precedence over maintenance reports, which take precedence over handover and startup reports, which take precedence over other reports," and is used for subsequent adjudication and normalization. An example is shown in Table 12. Table 12 The priority sequence is T04, T01, T02, T03, T05. This means that the index building process aligns the terms in the mapping list to the encoding space to obtain the encoding index table, and the conflict identification process identifies cases where the same term points to multiple encodings, forming a conflict list table. This reduces semantic-level alias conflicts to encoding-level conflicts for easier deterministic adjudication. An example is shown in Table 13. Table 13 Understandably, after aligning the coding index table with the priority sequence, the conflict identification process did not find any cases where the same term corresponds to multiple target codes. Therefore, the conflict list is recorded as a non-conflicting entry, and the coding field when involved is an empty placeholder. This result indicates that the mapping list is consistent with the historical codes in this batch of data, and there is no need to enter the abnormal decision-making process.

[0057] Furthermore, if the same term is incorrectly pointed to different codes in the historical database, it is written into the conflict list table and priority adjudication is performed according to the priority sequence to obtain the target type set. At the same time, the consistency verification process compares the target type set with the code index table one by one and outputs a consistency vector so as to find residual inconsistencies before normalization.

[0058] Furthermore, the priority adjudication process directly confirms the target type set as {T01, T02, T03, T04, T05} under the premise of no conflict. The consistency verification process compares the target type set with the encoding index table item by item to obtain the consistency vector [1, 1, 1, 1, 1], where the value 1 indicates complete consistency. This consistency vector serves as a pre-verification signal for the encoding normalization process, ensuring that the type mapping of the final output is one-to-one within this batch and can be directly called by the matching process.

[0059] For example, the mapping generation process expands the target type set and mapping list to obtain a mapping index table, and the encoding normalization process unifies the encoding format and case sensitivity while ensuring a one-to-one relationship, thereby outputting type mappings for downstream matching to generate matching priority sequences and weight threshold tables, as shown in Table 14: Table 14 Compared to the existing technology that directly groups by word similarity and then manually assigns codes, the above-mentioned coding normalization link uses the coding dictionary as an anchor and locates and adjudicates conflicts at the coding layer. This makes the generation of type mapping independent of human experience and has verifiable intermediate results. The technical effect is to ensure the uniqueness and stability of type mapping in scenarios where the task type has multiple source inputs and is inconsistent with historical codes, reduce type ambiguity in subsequent template matching, and improve the repeatability of automatic locking.

[0060] Template filtering module: This module is used to perform filtering on the template library based on the stable index vector to obtain a candidate set. The template library is a collection of report templates, and a report template refers to the structured descriptive data used to generate electronic reports. Methods for filtering the template library based on stable index vectors to obtain a candidate set include: Based on the stable index vector and template library, perform attribute extraction processing to obtain template metadata, and perform field mapping processing based on the stable index vector to obtain the matching condition set; Based on the template metadata and the matching condition set, perform device filtering to obtain a device filter set; based on the template metadata and the matching condition set, perform type filtering to obtain a type filter set. Based on the equipment filter set and the type filter set, perform set intersection and union processing to obtain a preliminary selection list. Based on the preliminary selection list and the template library, perform validity verification processing to obtain a valid list. Based on the valid list and the matching condition set, perform index aggregation processing to obtain the candidate set.

[0061] In this embodiment, the stable index vector is a set of device tree location indexes obtained by aligning the hierarchical vectors; the template library is a set of report templates covering device families such as high-voltage switchgear, low-voltage switchgear, transformers, generators, ATS, and UPS; the template metadata consists of attributes such as device family, report type, applicable voltage level, and required fields extracted from the template library; and the matching condition set is a set of device families and hierarchical fields obtained by reverse lookup in the device tree based on the stable index vector, which is used to drive subsequent filtering.

[0062] The attribute extraction process extracts template metadata from the template library to establish a comparable space. The field mapping process maps the hierarchical fields in the stable index vector to a set of matching conditions to form conditions such as device family, report type and voltage level. The purpose of both is to describe "task-side conditions" and "template-side attributes" with the same attribute domain, avoiding fuzzy matching caused by cross-domain comparison.

[0063] Furthermore, the device filtering process obtains the device filter set in the device family dimension based on the template metadata and the matching condition set, while the type filtering process obtains the type filter set in the report type dimension. The purpose of separating the two is to decouple "device matching" and "type matching" to reduce the sensitivity of misjudgment in a single filtering and to provide independent input for subsequent set operations.

[0064] The set intersection and union processing performs combination calculations of intersection, union, or difference on the equipment filter set and the type filter set to generate a preliminary selection list. The combination method is determined by the explicit constraints in the matching condition set. For example, if the equipment family is "high voltage cabinet" and the type is "inspection report", the intersection is used. If the type is "fault report" and template sharing is allowed across equipment families, the union is used, thereby minimizing template omissions and redundancy.

[0065] For example, the effective verification process uses the initial selection list as an index to back-check the template library to verify the compatibility of required fields and voltage levels, etc., and generates an effective list. Subsequently, the index summary process extracts template numbers based on the effective list and the matching condition set, and removes duplicates by device family and type dimension, finally obtaining a candidate set. The candidate set, as the input for subsequent matching steps, can stably converge to the template range consistent with the stable index vector without relying on manual interpretation.

[0066] Template matching module: Used to perform matching processing on the candidate set according to the type vector to obtain the lock set, where the lock set is the index set of the target template.

[0067] Methods for obtaining the locked set by performing matching processing on the candidate set based on the type vector include: Based on the type vector and the candidate set, feature expansion processing is performed to obtain a similarity matrix. Based on the type vector, weight generation processing is performed to obtain a matching priority sequence. Based on the similarity matrix and the matching priority sequence, a threshold filtering process is performed to obtain a weight threshold table. Based on the weight threshold table and the candidate set, an index filtering process is performed to obtain a matching index table. Based on the matching index table and the candidate set, perform index extraction processing to obtain the initial template set. Based on the initial template set and the type vector, perform conflict detection processing to obtain the matching conflict table. Based on the matching conflict table and the matching priority sequence, perform priority adjudication processing to obtain the adjudication index table. Based on the adjudication index table and the initial template set, perform index summarization processing to obtain the lock set.

[0068] In this embodiment, the type vector is a fixed-length encoded sequence obtained by type mapping, the candidate set is a set of template numbers screened by the stable index vector, the similarity matrix is ​​a matrix measuring the item-by-item similarity between the type vector and the template attributes of the candidate set, the matching priority sequence is a weighted order given according to the importance of the type, the weight threshold table is a list of segmented thresholds generated on the similarity matrix according to the priority weight, the matching index table is a list of template indexes obtained by threshold screening, the initial template set is a subset of templates extracted from the candidate set according to the matching index table, the matching conflict table is a conflict record of the initial template set coexisting on the same device family or the same type, the adjudication index table is a set of template indexes adjudicated according to the matching priority sequence, and the locked set is a set of target template numbers obtained after index aggregation.

[0069] Feature expansion processing expands the type encoding and matching fields of the candidate set template into a feature domain consistent with the type vector to form a comparable space. The similarity matrix is ​​calculated using known metrics such as one-dimensional cosine or Hamming consistency. The weight generation processing outputs a matching priority sequence based on the type code order in the type vector. Its purpose is to explicitly distinguish the contribution of types such as inspection reports and maintenance reports within the same metric domain, thereby providing a stable weighted reference for subsequent threshold screening.

[0070] Furthermore, the threshold screening process sets segmented thresholds on the similarity matrix according to the matching priority sequence and generates a weighted threshold table. The index screening process then extracts template numbers that meet the thresholds from the candidate set to form a matching index table. These two steps transform continuous similarity values ​​into operable index conditions to reduce the uncertainty of boundary samples, while ensuring that the dual consistency of device family and type is simultaneously constrained.

[0071] The index extraction process extracts templates from the candidate set according to the matching index table to obtain a preliminary template set. The conflict detection process checks whether there are multiple templates of similar type under the same device family or whether there are multiple voltage level templates under the same type in the preliminary template set and generates a matching conflict table. This design explicitly exposes conflicts with a structured list rather than implicitly covering them, which makes it easier for subsequent decisions to have a deterministic basis.

[0072] For example, if the type vector corresponds to a fault report that takes precedence over an inspection report, and the candidate set contains both a high-voltage cabinet fault template and a high-voltage cabinet inspection template, then the matching conflict table records the two template numbers under the high-voltage cabinet item. The priority decision processing first retains the fault template according to the matching priority sequence, and when there are multiple voltage levels of the same type, it selects a voltage level according to the known level order to generate a decision index table. This decision ensures that the template selection satisfies both the business priority order and the voltage level constraint.

[0073] The index aggregation process checks the consistency between the adjudication index table and the initial template set, and outputs the locked set after deduplication in the device family and type dimensions. The locked set is delivered to the report generation stage as the final template number set. The entire process reduces the fluctuation of similarity measurement through feature expansion and threshold filtering, and avoids ambiguity of type and voltage level through explicit conflict detection and priority adjudication, thereby stably obtaining template locking results consistent with the type vector without relying on manual judgment.

[0074] Example 2 Please see Figure 2 As shown, based on the same inventive concept, this embodiment discloses an automated template management method for operation and maintenance tasks. For details not covered in this embodiment, please refer to the relevant sections of Embodiment 1. The method includes: S10: Perform location processing on the device tree according to the task list to obtain the node index. The task list is a sequence of task items dispatched by the management terminal, and the device tree refers to a hierarchical data structure with the project as the root. S20: Perform hierarchical encoding processing based on node index to obtain hierarchical vector. Perform alignment processing on device tree according to hierarchical vector to obtain stable index vector. The hierarchical vector represents hierarchical features in the order of project, branch, computer room division, computer room item, first-level equipment, second-level equipment, and third-level equipment. S30: Obtain the type table, perform alias merging on the type table to obtain the type mapping, and perform type extraction on the task list according to the type mapping to obtain the type vector. Here, the type table refers to the key-value mapping of the task type directory. S40: Perform a filtering process on the template library based on the stable index vector to obtain a candidate set. The template library is a set of report templates, and a report template refers to the structured descriptive data used to generate electronic reports. S50: Perform matching processing on the candidate set according to the type vector to obtain the locked set, where the locked set is the index set of the target template.

[0075] The accompanying drawings of the embodiments of this invention only involve the structures involved in the embodiments of this invention. Other structures can refer to the general design. In the absence of conflict, the features of the same embodiment and different embodiments of this invention can be combined with each other. The above are only specific implementations of this invention, but the protection scope of this invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this invention should be included within the protection scope of this invention. Therefore, the protection scope of this invention should be determined by the protection scope of the claims.

Claims

1. An automated template management system for operation and maintenance tasks, characterized in that, include: Node positioning module: used to perform positioning processing on the device tree according to the task list to obtain the node index. The task list is a sequence of task items dispatched by the management terminal, and the device tree refers to a hierarchical data structure with the project as the root. The hierarchical alignment module is used to perform hierarchical encoding processing based on node indexes to obtain hierarchical vectors. Based on the hierarchical vectors, the device tree is aligned to obtain stable index vectors. The hierarchical vectors represent hierarchical features in the order of project, branch, data center division, data center item, first-level equipment, second-level equipment, and third-level equipment. Type extraction module: used to obtain the type table, perform alias merging on the type table to obtain the type mapping, and perform type extraction on the task list according to the type mapping to obtain the type vector. Here, the type table refers to the key-value mapping of the task type name directory. Template filtering module: This module is used to perform filtering on the template library based on the stable index vector to obtain a candidate set. The template library is a collection of report templates, and a report template refers to the structured descriptive data used to generate electronic reports. Template matching module: Used to perform matching processing on the candidate set according to the type vector to obtain the lock set, where the lock set is the index set of the target template.

2. The automated template management system for operation and maintenance tasks according to claim 1, characterized in that, Methods for performing location processing on the device tree based on the task list to obtain node indexes include: Based on the task list, perform field extraction processing to obtain the project number column; based on the task list, perform field extraction processing to obtain the device number column; based on the task list, perform sequential encoding processing to obtain the task sequence number column. Based on the project number column and the device tree, perform path acquisition processing to obtain the path index table. Based on the path index table and the device number column, perform key construction processing to obtain the hierarchical key sequence. Based on the hierarchical key sequence and the task sequence number column, perform key normalization processing to obtain the normalized key sequence. Based on the normalized key sequence and the device tree, perform node matching processing to obtain a candidate node set. Based on the candidate node set and the path index table, perform conflict annotation processing to obtain an ambiguity record table. Based on the candidate node set and the ambiguous record table, a disambiguation and merging process is performed to obtain a stable node set. Based on the stable node set, an index summarization process is performed to obtain the node index.

3. The automated template management system for operation and maintenance tasks according to claim 2, characterized in that, Methods for performing hierarchical encoding based on node indices to obtain hierarchical vectors include: Based on the node index and device tree, perform path parsing to obtain the parsed path table; based on the parent-child relationship of the device tree, perform parent-child extraction to obtain the parent-child mapping table. Based on the parsing path table and the parent-child mapping table, perform layer encoding processing to obtain the layer vector. Based on the parsing path table, perform layer depth calculation processing in the order of project, branch company, computer room branch, computer room sub-item, first-level equipment, second-level equipment, and third-level equipment to obtain the layer depth vector. Based on the layer position vector and layer depth vector, perform gap detection processing to obtain the gap matrix. Based on the gap matrix and the parent-child mapping table, perform cross-level verification processing to obtain the cross-level marker set. Based on the over-level tag set and the parsing path table, perform order adjustment processing to obtain the adjustment vector. Based on the adjustment vector and the node index, perform alignment calculation processing to obtain the initial alignment vector. Based on the initial alignment vector and the layer vector, an encoding fusion process is performed to obtain the alignment encoding vector. Based on the alignment encoding vector and the adjustment vector, a dimension compression process is performed to obtain the layer vector.

4. The automated template management system for operation and maintenance tasks according to claim 3, characterized in that, The methods for obtaining the layer vector by performing layer-level encoding processing based on the parsed path table and parent-child mapping table include: Based on the parsing path table and the parent-child mapping table, perform layer order extraction processing to obtain a layer name list. Based on the layer name list and the parsing path table, perform sequence mapping processing to obtain a layer sequence list. Based on the layer sequence table and the parent-child mapping table, perform cross-level detection processing to obtain the layer marker set. Based on the layer marker set and the layer sequence table, perform gap filling processing to obtain the filling list. Based on the layer sequence list and the padding list, perform bit weight generation processing to obtain the bit weight vector. Based on the bit weight vector and the layer sequence list, perform encoding calculation processing to obtain the layer bit encoding vector. Based on the layer encoding vector and the layer name list, perform dimension reshaping to obtain the layer vector.

5. An automated template management system for operation and maintenance tasks according to claim 3, characterized in that, Methods for obtaining hierarchical vectors by performing dimensionality compression based on the aligned encoding vector and the adjusted vector include: Based on the alignment encoding vector and the adjustment vector, range statistics processing is performed to obtain the component range table. Based on the alignment encoding vector and the component range table, threshold generation processing is performed to obtain the threshold list. Based on the adjustment vector and the threshold list, gating calculation is performed to obtain the gating weights. Based on the aligned encoding vector and the gating weights, weight superposition is performed to obtain the superimposed vector. Based on the superposition vector and component range table, perform principal axis selection processing to obtain the principal axis index. Based on the principal axis index and the superposition vector, perform component projection processing to obtain the projection matrix. Here, the principal axis is the linear projection direction that maximizes the variance of the superposition vector. Based on the projection matrix and the gate weights, perform dimension aggregation to obtain an aggregated vector. Based on the aggregated vector and the principal axis index, perform scale compression to obtain a hierarchy vector.

6. An automated template management system for operation and maintenance tasks according to claim 1, characterized in that, Methods for performing alias merging on type tables to obtain type mappings include: Based on the type table, perform term extraction processing to obtain a term list; based on the task list, perform type counting processing to obtain a frequency vector. Based on the term list and type table, perform alias indexing to obtain the alias index. Based on the alias index and frequency vector, perform conflict detection to obtain the conflict matrix. Based on the conflict matrix and frequency vector, conflict weight processing is performed to obtain the conflict weights. Based on the alias index and conflict weights, cluster merging processing is performed to obtain the merged index. Based on the merged index and the term list, a mapping generation process is performed to obtain a mapping list. Based on the mapping list and the type table, an encoding normalization process is performed to obtain the type mapping.

7. An automated template management system for operation and maintenance tasks according to claim 6, characterized in that, Based on the conflict matrix and frequency vector, conflict weight processing is performed to obtain the conflict weights. Methods for obtaining conflict weights include: Based on the frequency vector, normalization is performed to obtain the normalized frequency; based on the collision matrix, effective masking is performed to obtain the matrix mask. Based on the collision matrix and matrix mask, intensity calculation is performed to obtain the collision metric. Based on the collision metric and normalized frequency, frequency modulation is performed to obtain the modulation vector. Based on the modulation vector, a sliding window smoothing process is performed to obtain a smoothed vector. Based on the smoothed vector and the matrix mask, an anomaly suppression process is performed to obtain a suppressed vector. Based on the suppression vector and the normalized frequency, coefficient mapping is performed to obtain the weight coefficients. Based on the weight coefficients, threshold shaping is performed to obtain the conflict weights.

8. An automated template management system for operation and maintenance tasks according to claim 7, characterized in that, The methods for obtaining type mappings by performing encoding normalization processing based on the mapping list and type table include: Based on the mapping list and type table, perform encoding extraction processing to obtain the encoding dictionary table; based on the type table, perform priority generation processing to obtain the priority sequence. Based on the mapping list and the encoding dictionary table, perform index building processing to obtain the encoding index table. Based on the encoding index table and the priority sequence, perform conflict identification processing to obtain the conflict list table. Based on the conflict list and priority sequence, priority adjudication is performed to obtain the target type set. Based on the target type set and the coding index table, consistency verification is performed to obtain the consistency vector. Based on the consistency vector and the target type set, a mapping generation process is performed to obtain a mapping index table. Based on the mapping index table and the encoding dictionary table, an encoding normalization process is performed to obtain the type mapping.

9. An automated template management system for operation and maintenance tasks according to claim 1, characterized in that, Methods for filtering the template library based on stable index vectors to obtain a candidate set include: Based on the stable index vector and template library, perform attribute extraction processing to obtain template metadata, and perform field mapping processing based on the stable index vector to obtain the matching condition set; Based on the template metadata and the matching condition set, perform device filtering to obtain a device filter set; based on the template metadata and the matching condition set, perform type filtering to obtain a type filter set. Based on the equipment filter set and the type filter set, perform set intersection and union processing to obtain a preliminary selection list. Based on the preliminary selection list and the template library, perform validity verification processing to obtain a valid list. Based on the valid list and the matching condition set, perform index aggregation processing to obtain the candidate set.

10. An automated template management system for operation and maintenance tasks according to claim 9, characterized in that, Methods for obtaining the locked set by performing matching processing on the candidate set based on the type vector include: Based on the type vector and the candidate set, feature expansion processing is performed to obtain a similarity matrix. Based on the type vector, weight generation processing is performed to obtain a matching priority sequence. Based on the similarity matrix and the matching priority sequence, a threshold filtering process is performed to obtain a weight threshold table. Based on the weight threshold table and the candidate set, an index filtering process is performed to obtain a matching index table. Based on the matching index table and the candidate set, perform index extraction processing to obtain the initial template set. Based on the initial template set and the type vector, perform conflict detection processing to obtain the matching conflict table. Based on the matching conflict table and the matching priority sequence, perform priority adjudication processing to obtain the adjudication index table. Based on the adjudication index table and the initial template set, perform index summarization processing to obtain the lock set.

Images