Knowledge graph-driven manufacturing process optimization system

Abstract

The invention relates to the technical field of process optimization, in particular to a knowledge graph-driven manufacturing process optimization system, which comprises a graph node generation module, a dependency relationship construction module, a semantic entity matching module, a process path screening module and a trend-driven early warning module. According to the method, semantic implicit relation recognition is achieved through sequential alignment and numerical difference comparison, quantitative screening between the machining size and the tolerance grade is introduced in path construction, and it is ensured that a path connection structure is optimized under the condition that the size precision constraint is met; in combination with the time sequence trend splitting and continuous fluctuation consistency discrimination mechanism of the equipment operation state, the dynamic marking and early warning annotation set generation of the process deviation trend can be completed at the node path level, and the closed-loop processing from knowledge modeling, path screening to trend pre-judgment is realized. And the process path adaptability, the parameter configuration precision and the abnormality identification advancement are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of process optimization, and particularly to a knowledge graph-driven manufacturing process optimization system. Background Art

[0002] The technical field of process optimization mainly focuses on improving the efficiency, product quality, and resource utilization rate in the manufacturing process, specifically including multiple aspects such as manufacturing parameter configuration, process path adjustment, process coordination scheduling, and equipment operation optimization. The technologies in this field usually combine industrial engineering, control theory, and algorithm modeling to achieve dynamic optimization control of key indicators such as energy consumption, time cost, and quality fluctuations in the manufacturing process. In the context of informatization and intelligence, process optimization is widely introducing specific technical solutions such as multi-objective optimization algorithms, model predictive control, data-driven Bayesian optimization, and reinforcement learning to address process constraints, non-linear coupling, and uncertain disturbance problems in high-complexity manufacturing scenarios.

[0003] Among them, the knowledge graph-driven manufacturing process optimization system is a process optimization solution based on semantic association and structured knowledge representation, aiming to achieve explicit modeling and reasoning of manufacturing knowledge by constructing a knowledge graph containing multiple types of entities and relationships such as process parameters, equipment performance, product structure, and quality feedback. The system is used to support manufacturing enterprises in making rapid and intelligent decision recommendations in scenarios such as process planning, anomaly diagnosis, and parameter adjustment, and to improve the accuracy and adaptability of process plan selection by mining semantic paths and entity relationships in the knowledge graph.

[0004] Traditional optimization systems lack the ability to identify differences between operation entities with semantic heterogeneous expressions when constructing a manufacturing process knowledge graph. Path construction is only based on static structure reasoning, and no dynamic constraint mechanism for adapting machining dimensional tolerances is formed, resulting in problems of accumulated dimensional errors caused by inaccurate matching of control parameter ranges in path selection. In anomaly prediction, it mostly relies on single-point parameter threshold judgment and is difficult to accurately identify potential anomalies caused by continuous fluctuation trends during the operation stage. A typical situation is that although the spindle vibration value does not exceed the limit but is in a continuous rising state, it is still difficult to be captured, reducing the timeliness of process anomaly response and the foresight of decision-making, and limiting the ability to actively intervene in manufacturing process disturbances in complex scenarios. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies existing in the prior art and propose a knowledge graph-driven manufacturing process optimization system.

[0006] To achieve the above purpose, the present invention adopts the following technical solution: A knowledge graph-driven manufacturing process optimization system, the system includes: The graph node generation module obtains process documents, assembly flowcharts, and equipment control item descriptions in the numerical control processing line, constructs three types of knowledge graph entity nodes, establishes an entity basic network structure, and generates a process graph node set; The dependency relationship construction module maps and groups the cutting parameter ranges and equipment linkage marks based on the process graph node set, forms a dependency linkage item table, and embeds an association path in the graph structure in the form of edge connection to generate a graph dependency structure network; The semantic entity matching module calls the graph dependency structure network, obtains the equipment operation entities with semantic heterogeneous expressions in the graph, sequentially performs sequential alignment judgment and numerical offset calculation, and sets graph interaction connections according to the co-occurrence frequency in the structure network to generate an entity semantic fusion graph; The process path screening module calls the control parameter edges connected to each process node in the path based on the entity semantic fusion graph, statistically outputs the difference between the machining dimension value and the corresponding upstream and downstream nodes, establishes a path connectivity structure between nodes and marks the path preference level to generate a process path preference graph.

[0007] The improvements of the present invention are that the process graph node set includes an operation action node set, a component structure node set, a control parameter node set, a node pointing relationship set, and an initial graph connection structure; the graph dependency structure network includes a step number sequence, an equipment linkage mark set, a control parameter edge set, a parameter dependency connection path set, and a node-to-node linkage structure; the entity semantic fusion graph includes a semantic alignment node mapping, an offset difference annotation node, a parameter co-occurrence frequency record, a semantic heterogeneous node group, and a fused path topology structure; the process path preference graph includes a path node connectivity structure, an error elimination path list, a preference level mark group, a machining target matching index, and an optional path set.

[0008] The improvements of the present invention are that the graph node generation module includes: The document parsing sub-module obtains process documents, assembly flowcharts, and equipment control item descriptions in the numerical control processing line, extracts the operation step content recorded in the process document, the part numbers represented in the assembly flowchart, and each type of equipment control field in the equipment control item description, splits the three types of text items into semantic paragraphs respectively, counts the occurrence frequency of keywords, matches the operation word list, part dictionary, and control parameter standard items, establishes the word frequency distribution data of operation type word items, structure type word items, and control type word items, and generates a word frequency distribution ratio matrix; The structure recognition sub-module judges the content dependency relationship between operation type, structure type, and control type word items based on the word frequency distribution ratio matrix, sets the word item groups with a dependency frequency higher than the structure matching reference value as candidate node groups, performs entity label normalization processing on the word items in the candidate node groups and maps them to the standard item codes to obtain a word item set with a unique entity identifier, and generates a standard word item entity comparison set; The node construction sub-module calls the standard term entity control set, sets three types of graph nodes according to the standard item codes of the operation class, structure class, and control class according to the term category, configures a directed connection relationship between the three types of nodes according to the term dependency path, establishes a structured node network with complete pointing edges, and generates a process graph node set.

[0009] The improvement of the present invention is that the dependency relationship construction module includes: The reference extraction sub-module is based on the process graph node set, extracts the corresponding numbers in the operation action nodes and control parameter nodes, obtains the processing instruction serial number in the operation node and the device control code associated in the parameter node, and screens the node pairs with the reference frequency of the number greater than two as the direct reference relationship unit by comparing the frequency of simultaneous occurrence of the numbers in the task flow list, and establishes an operation parameter reference frequency value; The parameter combination sub-module calls each group of operation step numbers and control parameter numbers according to the operation parameter reference frequency value, extracts the cutting depth, feed speed, and spindle speed under the corresponding process link, classifies and marks them in combination with the device type label, takes the device number as the main key, reconstructs the sequence of the operation step numbers in chronological order, groups the parameter items into a fixed field set, and establishes a combined sequence mapping result; The path connection sub-module sets a directed connection path for the dependency relationship between nodes according to the combined sequence mapping result, calls each grouped field corresponding operation action node and control parameter node in the graph, writes the path structure into the node edge set, and marks the path source device number to generate a graph dependency structure network.

[0010] The improvement of the present invention is that the semantic entity matching module includes: The structure extraction sub-module calls the graph dependency structure network, extracts the corresponding application device number, operation sequence number, and parameter value field according to the parameter items corresponding to the control parameter group in each node, groups the device numbers in the parameter group, arranges the operation sequence fields in the group in ascending order of the number, maps the parameter value fields according to the node path number to generate a distribution set, and establishes a structure distribution relationship value; The difference judgment sub-module screens the parameter group set with the same sequence number under each device number according to the structure distribution relationship value, extracts the parameter value fields in the corresponding nodes, calculates the absolute value of the difference between the parameter values and the overall distribution standard deviation under the same number, calculates the relative difference offset degree, and extracts the structure common nodes and path numbers according to the parameter group with the relative difference offset degree value lower than the offset difference threshold to generate an offset alignment node group; The semantic fusion sub-module statistically calculates the frequency values of node paths in the graph structure based on the offset-aligned node groups, assigns interaction identifiers to path groups with frequency values greater than the co-occurrence threshold, updates the set of node-to-connection edges in the entity graph, and establishes an entity semantic fusion graph.

[0011] The improvements of the present invention include that the process path screening module comprises: The path extraction sub-module, based on the entity semantic fusion graph, extracts the path structures under each group of nodes according to the node groups pointing to the same processing target, records the process node numbers and the edge information of the connection control parameters in the paths, establishes the pairing relationship between the path numbers and the control parameter edges, and generates the process path structure quantity; The error judgment sub-module extracts the processed dimension values output by each process node and the processed dimension values of the corresponding upstream and downstream nodes according to the process path structure quantity, calculates the absolute difference between the nodes and compares it with the dimension error critical value in the IT tolerance standard level to obtain the processing deviation quantity through calculation, summarizes the error marks for the path nodes with a processing deviation quantity greater than zero, screens out the corresponding path structures, obtains the compliance path number sequence, and establishes the processing deviation elimination information; The grade marking sub-module, based on the processing deviation elimination information, re-statistically calculates the average difference of the node pairs in each path according to the compliance path number sequence, divides the paths into multiple grade range segments according to the proportional distribution between the statistical result and the minimum tolerance of the tolerance grade, adds a grade identifier field to each path, and establishes a process path optimization graph.

[0012] The improvements of the present invention include that the system further comprises: The trend-driven warning module calls the process path optimization graph, collects the real-time control parameter time series values during operation, compares the consistency of the fluctuation directions within each cycle segment and statistically calculates the number of continuous cycles. When the nodes have consistent continuous fluctuation directions and the number of cycles exceeds the standard value, the corresponding edges in the graph are marked as warning edges, and a graph anomaly trend annotation set is generated; The graph anomaly trend annotation set specifically includes a cycle consistency marked edge, a continuous offset path index, an equipment anomaly fluctuation label, a warning path connection identifier, and a dynamic trend annotation mapping table.

[0013] The improvements of the present invention include that the trend-driven warning module comprises: The time series acquisition sub-module calls the process path optimization graph, extracts the periodic change sequences of the spindle vibration value, the workpiece temperature rise value, and the spindle speed value in the processing stage according to the operation control parameter time series values of the equipment associated with the path nodes, segments the parameter time series, and establishes a processing stage time series set; The fluctuation recognition sub-module calls the time series set of the processing stage, identifies the fluctuation directions of three types of parameters, namely the spindle vibration value, the workpiece temperature rise value, and the spindle speed value, within each cycle segment, determines the positive and negative relationships of the numerical differences between adjacent data points in each segment, counts the number of paragraphs in the same direction, calculates the fluctuation trend intensity, marks the continuous trend change cycle segments according to the magnitude of the fluctuation trend intensity value, and establishes the trend continuous intensity value. Based on the trend continuous intensity value, the anomaly marking sub-module determines whether there is a situation where the continuous cycle trend intensity on the edge corresponding to the path node exceeds the control parameter fluctuation standard value, marks the edges that meet the conditions in the path graph as warning edges, and establishes the graph anomaly trend annotation set.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In the present invention, by extracting semantic entities from process documents, assembly processes, and equipment control items and performing structured construction, the multi-dimensional relationships among processing operations, component structures, and control parameters can be covered. A dependency linkage network is established based on the reference relationships and parameter mappings between nodes. The heterogeneous expressions of equipment operation entities are integrated during the semantic comparison process. The implicit semantic relationships are identified through sequential alignment and numerical difference comparison. A quantitative screening between processing dimensions and tolerance grades is introduced in the path construction to ensure the optimization of the path connection structure under the constraint of meeting dimensional accuracy. Combining the time series trend splitting and continuous fluctuation consistency discrimination mechanism of the equipment operation state, it is possible to dynamically mark the process deviation trend and generate the warning annotation set at the node path level, realizing the closed-loop processing from knowledge modeling, path screening to trend prediction, effectively improving the adaptability of the process path, the accuracy of parameter configuration, and the advance of anomaly recognition. Brief Description of the Drawings

[0015] Figure 1 is the system flow chart of the present invention; Figure 2 is the flow chart of the graph node generation module of the present invention; Figure 3 is the flow chart of the dependency relationship construction module of the present invention; Figure 4 is the flow chart of the semantic entity matching module of the present invention; Figure 5 is the flow chart of the process path screening module of the present invention; Figure 6 is the flow chart of the trend-driven warning module of the present invention. Detailed Embodiments

[0016] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0017] In the description of the present invention, it should be understood that the orientation or positional relationship indicated by the terms "length", "width", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, in the description of the present invention, the meaning of "a plurality of" is two or more unless otherwise specifically defined.

[0018] Please refer to Figure 1 , the present invention provides a technical solution: a knowledge graph-driven manufacturing process optimization system, the system includes: The graph node generation module obtains the process documents, assembly flowcharts and equipment control item descriptions in the numerical control processing line, extracts the operation items in the process documents, the component nodes in the assembly flowcharts and the control parameter fields in the equipment control item descriptions, respectively constructs three types of knowledge graph entity nodes, correspondingly sets the pointing relationships between the operation action nodes, component structure nodes and control parameter nodes, establishes the entity basic network structure, and generates the process graph node set; The dependency relationship construction module extracts the direct reference relationships between the operation action nodes and the control parameter nodes based on the process graph node set, classifies and combines the involved operation step numbers, cutting parameter ranges and equipment linkage marks, restores the sequence of the operation step numbers, maps and groups the cutting parameter ranges and equipment linkage marks, forms a dependency linkage item table, and embeds the association path in the graph structure in the form of edge connection to generate the graph dependency structure network; The cutting parameter range is widely used in the manufacturing industry to represent the operation boundary parameters such as spindle speed, feed speed, cutting depth, etc.; The semantic entity matching module calls the graph dependency structure network, identifies the distribution structure of the control parameter groups within the nodes according to each node and path, establishes a multi-dimensional comparison matrix according to the application equipment, adjustment sequence and numerical difference of the parameters, obtains the equipment operation entities with semantic heterogeneous expressions in the graph, sequentially performs sequence alignment judgment and numerical offset calculation, and filters the operation nodes with implicit association relationships according to the offset difference threshold, and sets the graph interaction connection according to the co-occurrence frequency in the structure network to generate the entity semantic fusion graph; Based on the entity semantic fusion graph, the process path screening module extracts all optional path structures under the same processing target, calls the control parameter edges connected to each process node in the path, calculates the differences between the processed dimension values and the corresponding upstream and downstream nodes, compares each difference with the critical value in the IT tolerance standard level one by one, excludes the paths with excessive errors, establishes the path connection structure between nodes, marks the path preference level, and generates the preferred process path graph; The trend-driven warning module calls the preferred process path graph, collects the time series values of real-time control parameters during operation according to the path node-related equipment, splits the time series of spindle vibration value, workpiece temperature rise value and spindle speed value into corresponding periodic segments according to the processing stage, compares the consistency of the fluctuation directions within each periodic segment and counts the number of continuous periods. When the nodes have consistent continuous fluctuation directions and the number of periods exceeds the standard value, mark the corresponding edge in the graph as a warning edge and generate the graph abnormal trend annotation set; The process graph node set includes the operation action node set, the component structure node set, the control parameter node set, the set of pointing relationships between nodes and the initial connection structure of the graph. The graph dependency structure network includes the step number sequence, the equipment linkage mark set, the control parameter edge set, the parameter dependency connection path set and the inter-node linkage structure. The entity semantic fusion graph includes semantic alignment node mapping, offset difference annotation nodes, parameter co-occurrence frequency records, semantic heterogeneous node groups and the fused path topology structure. The preferred process path graph includes the path node connection structure, the error elimination path list, the preferred level mark group, the processing target matching index and the set of optional paths. The graph abnormal trend annotation set is specifically the periodic consistency marked edge, the continuous offset path index, the equipment abnormal fluctuation label, the warning path connection identifier and the dynamic trend annotation mapping table.

[0019] Please refer to Figure 2 , the graph node generation module includes: The document parsing sub-module obtains the process documents, assembly flowcharts and equipment control item descriptions in the numerical control processing line, extracts the operation step content recorded in the process documents, the part numbers represented in the assembly flowcharts and each type of equipment control field in the equipment control item descriptions, splits the three types of text items into semantic paragraphs respectively, counts the frequency of keyword occurrences, matches the operation word list, part dictionary and control parameter standard items, and establishes the word frequency distribution data of operation-related word items, structure-related word items and control-related word items, and generates the word frequency distribution ratio matrix; Obtain the process documents, assembly flowcharts, and equipment control item descriptions in the numerical control processing line. Sequentially extract the operation step content, part numbers, and control field names. Taking five process documents recorded in the 2023 processing line as examples, each process document contains 15 to 30 operation instructions. Each instruction corresponds to the sequence code and action keyword of the processing link. Split the process document by paragraphs, and extract the three types of keywords with the highest occurrence frequencies in each paragraph to form an operation word frequency statistical set. For the numbered structure marked in the assembly flowchart, extract the numbered items, position identifiers, and their adjacency relationship information in the figure, and count the frequency of repeated numbers to form a part number word frequency statistical set. Further analyze the content of the equipment control item descriptions, extract the control fields starting with "speed", "feed", "temperature", etc., match the standard control parameter set, compare the spelling similarity and root word correlation degree of the field names, extract the corresponding control type keywords, count the occurrence frequency of each control word and record its unit and default value range. For example, the spindle speed is in the unit of "rpm", and the common range is 500 to 12000 rpm, to form a control parameter word frequency statistical set. After the above three types of texts are split by text paragraphs and keyword matching, three types of word frequency statistical data are formed, and the word frequency item groups of the operation class, structure class, and control class are constructed respectively. Use the word frequency normalization method to standardize the word frequency to the interval [0,1], and construct a word frequency distribution ratio matrix. Each unit in the matrix represents the relative frequency ratio of a certain type of word item in the current text source. See Table 1; Table 1 Word Frequency Ratio Matrix of Process Document Samples

[0020] As shown in Table 1, the control class word frequency is relatively low in some documents, corresponding to a narrow coverage of such word items in the numerical control document content. According to the actual ratio matrix data, the word frequency distribution ratio matrix is obtained.

[0021] Based on the word frequency distribution ratio matrix, the structure recognition sub-module judges the content dependency relationship between the operation class, structure class, and control class word items. Set the word item groups with a dependency frequency higher than the structure matching reference value as the candidate node groups. Normalize the entity labels of the word items in the candidate node groups and map them to the standard item codes to obtain a set of word items with unique entity identifiers, and generate a standard word item entity comparison set; Based on the word frequency distribution ratio matrix, the content dependence relationship between operation class, structure class, and control class terms is judged. For the judgment of the dependence relationship between terms, the frequency of co-occurrence of terms in the same text segment is cumulatively counted. For example, if the operation class term "clamping" and the structure class term "workpiece A5" co-occur 7 times in 10 documents, the co-occurrence frequency of the two is 70%. The structure matching reference value is set at 60%. According to this reference, the term pairs with co-occurrence frequencies exceeding this value form a candidate node group; then the candidate terms are mapped to the standard item codes. For example, "spindle speed" is mapped to the equipment control standard field "SPINDLE_SPEED". The equipment standard field set defined in GB / T 3811-2008 is used as the mapping dictionary to construct a standardized label list; for the term combinations with co-occurrence frequencies of 78% ["feeding" + "transfer device B7"] and 83% ["cooling" + "cooling unit X3"], it is confirmed that they meet the set reference value, and a unique entity code is constructed for them. Mapping and matching are carried out according to the standard structural part numbering system to complete the normalization process of entity labels and obtain a standard term entity comparison set.

[0022] The node construction sub-module calls the standard term entity comparison set, sets three types of graph nodes according to the operation class, structure class, and control class standard item codes according to the term category, configures a directed connection relationship between the three types of nodes according to the term dependence path, establishes a structured node network with complete directed edges, and generates a process graph node set; Call the operation class, structure class, and control class standard item codes in the constructed standard term entity comparison set. Construct the node "NODE_ACT_CLAMP" with the operation "clamping", construct the node "NODE_COMP_A5" with the structural part "workpiece A5", and construct the node "NODE_CTRL_SSPEED" with the control field "SPINDLE_SPEED". Determine the connection direction according to the co-occurrence relationship between terms. If the operation class appears before the structure class, set it as the starting point of the directed edge pointing to the structure class node; further set an edge between the operation class and the control class. The edge weight value depends on the co-occurrence frequency of the terms and the coding matching score of the standard terms. If the co-occurrence frequency of "clamping" and "SPINDLE_SPEED" is 65% and the matching score is 90, the edge weight calculation is To ensure the effectiveness of the edge, the weight threshold is set at 0.5, and the following formula is used to judge whether to retain the edge relationship: ; Among them, represents the normalized word frequency value of the operation class term in the paragraph, represents the normalized word frequency value of the control class term, represents the co-occurrence frequency of the two, represents the semantic similarity score of the two in the standard item coding. If the calculated result value exceeds 0.5, a connection edge relationship is established, and finally a process map node set is constructed. The benefit of the formula is that by comprehensively considering the word frequency difference and semantic similarity, the map connection relationship can more accurately reflect the actual dependency relationship in the control logic. Among them, represents the normalized word frequency value of the th operation word item in the paragraph, represents the normalized word frequency value of the th control word item, represents the co-occurrence frequency of the th word item pair, ; this result indicates that its weight is too low and no edge is set either. Only combinations with a weight greater than 0.5 are embedded in the process map node network.

[0023] Please refer to Figure 3 . The dependency relationship construction module includes: The reference extraction sub-module, based on the process map node set, extracts the corresponding numbers in the operation action nodes and control parameter nodes, obtains the processing instruction numbers in the operation nodes and the device control codes associated in the parameter nodes, and by comparing the frequencies of simultaneous occurrence of the numbers in the task flow list, filters out the node pairs with a reference frequency greater than two as direct reference relationship units to establish the operation parameter reference frequency value; When the reference extraction sub-module performs the number extraction of operation action nodes and control parameter nodes based on the process map node set, it first calls all the nodes marked as operation types from the node set, sequentially reads the processing instruction numbers stored in each node, sets the number format as a four-digit Arabic numeral code (such as "OPC1" to "OPC120"), and then performs the same number extraction on the control parameter nodes. The numbers in the control parameter nodes are the encoding of the device control fields, and the three-segment marking rule "EQ-XY-Z" is adopted, where "XY" is the device type code and "Z" is the control parameter serial number. For example, "EQ-MC-7" represents the 7th item of the control parameters of the CNC milling machine. Then, from the task flow list, a task number index table is constructed, and the number of times each group of operation instruction numbers and control parameter numbers appear together in the task flow sequence is counted. For example, "OPC12" and "EQ-MC-7" appear simultaneously in the task numbers "T-008", "T-017", and "T-021". Since their common occurrence times are 3, which is higher than the set threshold of 2, they are screened as reference relationship node pairs. The setting of this threshold is based on the co-line reference frequency distribution of standard production line tasks. By statistically analyzing 200 historical task samples, the average co-occurrence times of operation-parameter pairs are 1.68, and the integer value corresponding to the 70% quantile is taken as the screening threshold. Therefore, the threshold is set to 2 to ensure that only the core pairs of repeated calls are retained. The qualified node pairs are included in the reference relationship set, and their co-occurrence frequency is recorded as the reference frequency value. If a group of operation parameter node pairs such as "OPC12-EQ-MC-7" appears 3 times, its corresponding reference frequency value is recorded as 3, which constitutes the basis for subsequent parameter combination judgment. The operation parameter reference frequency value includes the operation instruction number, the control parameter number, and the reference frequency.

[0024] According to the operation parameter reference frequency value, the parameter combination sub-module calls each group of operation step numbers and control parameter numbers, extracts the cutting depth, feed rate, and spindle speed under the corresponding process link, classifies and marks them in combination with the device type label, reconstructs the sequence of operation step numbers in chronological order with the device number as the primary key, groups the parameter items into a fixed field set uniformly, and establishes a combined sequence mapping result; Based on the set of reference frequency values output by the previous sub-module, the parameter combination sub-module sequentially extracts data for each pair of operation step numbers and control parameter numbers. First, it calls the task execution log with the operation step number as the index to obtain the set processing parameter items set in the corresponding process link. Among them, the cutting depth is recorded in the unit of "mm", the feed rate is recorded in "mm / min", and the spindle speed is recorded in "rpm". Then, it reads the device information associated with the control parameter number and identifies its device type label. For example, "MC" represents a CNC milling machine, "DR" represents a drilling unit, etc. After packing and marking the three parameters and the device type, it establishes a multi-field mapping matrix. For the node pair "OPC12-EQ-MC-7" in task number "T-017", the query result shows that the cutting depth is 0.5 mm, the feed rate is 450 mm / min, the spindle speed is 1200 rpm, and the device type label is "MC". According to such data, the operation step numbers are sorted by task time, and its execution order is reconstructed as "OPC11→OPC12→OPC15" to form a mapping sequence. Then, the three parameter fields are grouped, and the fixed field set is set as [cutting depth, feed rate, spindle speed, device label]. With the device number "MC-001" as the primary key, the corresponding relationship between the sequence and the parameter set is constructed to generate the combined sequence mapping result. The following table lists the sample combinations: Table 2 Sample Table of Process Task Parameter Combinations

[0025] As shown in Table 2, in different step combinations, the parameter values are slightly adjusted according to the device type and task load, forming a grouped data structure with time sequence constraints. The combined sequence mapping result includes the device number, the operation step order, and the grouped parameter field values.

[0026] Based on the combined sequence mapping result, the path connection sub-module sets a directed connection path for the dependency relationship between the nodes by calling each grouped field corresponding operation action node and control parameter node in the graph. It writes the path structure into the node edge set and marks the path source device number to generate the graph dependency structure network; Based on the combined sequence mapping result, the path connection sub-module sequentially reads the operation action nodes and control parameter nodes corresponding to each device number, searches for the corresponding node instances in the graph structure, and retrieves whether there is parameter dependency or instruction call logic between them. By comparing the fields "call order" and "parameter dependency index" in the node attribute table, if the call order is less than the dependency index and there is a nested relationship in the parameter fields, a directed path relationship is established. Taking the device number "MC-001" as an example, its operation sequence is "OPC11→OPC12→OPC15", and the corresponding control parameter nodes are "EQ-MC-5", "EQ-MC-7", "EQ-MC-9". The path direction is recorded as "OPC11→EQ-MC-5→OPC12→EQ-MC-7→OPC15→EQ-MC-9" in the node edge connection, and the source device of the path field is marked as "MC-001". In the path connection logic, if a certain path direction violates the device dependency rule, such as setting the "feed rate" after calling the "spindle speed", the path setting is cancelled. Finally, the structured path is written into the node edge set. The judgment reference value for the path connection to be established is that the parameter dependency matching rate ≥ 85%. This ratio is calculated from the field structure similarity in the standard device control link. 12 groups of standard reference paths are set, and the average matching rate is 87.3%. Therefore, 85% is used as the reference value for constructing the effective path. In the example, the matching rate of the path "OPC11→EQ-MC-5→OPC12→EQ-MC-7" is 91%, which meets the set standard. The graph dependency structure network specifically includes the node path direction group, the device number mapping mark, and the control relationship edge weight.

[0027] Please refer to Figure 4 , the semantic entity matching module includes: The structure extraction sub-module calls the graph dependency structure network, extracts the corresponding application device number, operation sequence number, and parameter value field according to the parameter items corresponding to the control parameter group in each node, groups the device numbers in the parameter group, arranges the operation sequence fields in the group in ascending order of the number, maps the parameter value field according to the node path number to generate a distribution set, and establishes a structure distribution relationship value; When the structure extraction sub-module calls the graph dependence structure network, it first obtains the parameter item information included in the control parameter group in each node, and sequentially parses the corresponding device number field, operation sequence number field, and parameter value field for each control parameter group. The specific call form is to read the three columns of "Dev_ID", "Step_Seq", and "Param_Value" in the structure node data table. The device numbers such as "EQ-101" and "EQ-202" are represented in a unified format. The operation sequence numbers are an ascending sequence group such as "001", "002", "003", etc. After pre-normalization processing, the parameter value field is represented by a floating point number in the range of 0 to 1. Example data is shown in Table 4. Then, all control parameter groups are grouped according to the device number. For example, there are 5 parameter sets under the device "EQ-101", which appear in the operation step numbers 001 to 005 respectively. Then the grouping marks are "EQ-101-G1" to "EQ-101-G5". Inside each group, they are sorted in ascending order according to the operation sequence number to ensure the consistency of the operation sequence execution. Then, with the node path number field as the index, the parameter value field is sequentially mapped on the path sequence to form a distribution vector set. To clarify the distribution relationship, the aggregation trend of the parameter vector sequence on the path number is calculated in each device group to obtain the parameter structure distribution relationship value. This value is generated by calculating the sequence similarity of each group of parameters on the path and the numerical distribution consistency, and the numerical range is 0 to 1. The higher the value, the stronger the structure stability. For example, for the device "EQ-101-G2", its parameter value sequence [0.55, 0.63, 0.61] corresponds to the path sequence [ND-1, ND-3, ND-5], showing a consistent distribution structure in the path node graph, and the structure distribution relationship value is 0.89. The structure distribution relationship value specifically includes the path number sequence, the parameter normalization value sequence, and the sequence distribution trend index.

[0028] The difference judgment sub-module filters the set of parameter groups with the same sequence number under each device number according to the structure distribution relationship value, extracts the parameter value field in the corresponding node, calculates the absolute value of the difference between the parameter values with the same number and the overall distribution standard deviation, using the formula: ; Calculate to obtain the relative difference offset degree. According to the parameter groups with the relative difference offset degree value lower than the offset difference threshold, extract the structure common nodes and path numbers to generate an offset alignment node group; Among them, represents the normalized value of the th control parameter in the th device, represents the normalized value of the corresponding th control parameter in the reference device, represents the The average of the normalized values of all parameters in a device parameter group, represents the average of the normalized values of all parameters in the reference device parameter group, represents the total number of control parameters participating in the calculation for each group, represents the relative difference offset degree between two device operation parameter groups, which measures the overall offset degree of the value distribution structures of the two groups of parameters; The difference judgment sub-module extracts the parameter groups with the same operation sequence number set under different device numbers based on the aforementioned structure distribution relationship value, and obtains the group pairs with consistent sequence numbers under multiple devices through structure index matching. For example, both "EQ-101-G2" and "EQ-202-G2" contain operation steps 001 to 003. On the premise of ensuring the same number of parameters, the parameter value fields with the same number in the two groups are extracted respectively and corresponding matrices are constructed. Let the two sets of data be and , such as "[0.55, 0.63, 0.61]" for "EQ-101-G2" and "[0.52, 0.64, 0.59]" for "EQ-202-G2". The relative difference offset degree is calculated using the following formula: ; where, represents the normalized value of the th control parameter in the rd device, and this value is obtained from the original control parameters (such as spindle speed, feed speed, etc.) through maximum-minimum normalization transformation; represents the normalized value of the th corresponding control parameter in the reference device; represents the arithmetic mean of the normalized values of all control parameters of this device participating in the comparison, and the calculation method is ; represents the average of the normalized values of the control parameters at the same position in the reference device; represents the total number of control parameters participating in the comparison for each group; represents the relative difference offset degree between two device parameter groups.

[0029] The structural logic of the above formula is: The numerator part: , represents the absolute value difference between each corresponding parameter item, and accumulates and sums them to reflect the total offset amount between the overall parameters; The denominator part: , represents the square root of the sum of the sum of squares of deviations from the mean within each of the two groups of parameters, that is, the combined amount of standard deviations, which is used to measure the degree of fluctuation of the parameter values themselves; Overall logic: Normalize the ratio of the absolute difference to the combined standard deviation. If the means of the two groups of parameters are close, with small fluctuations but large differences, then the value is large; otherwise, it is small. This structure avoids the situation where the comparison fails due to excessive fluctuations in the distribution of the original parameters, making the comparison robust and widely applicable.

[0030] In the numerical example, set the parameter group "EQ-101-G2" as , with an average value of , and set the parameter group "EQ-202-G2" as , with an average value of . Substitute each value into the formula for calculation: ; ; ; Among them, the offset difference threshold is set to 0.6. The setting basis is as follows: The median of the S values calculated from a total of 50 groups of control parameters in the typical device comparison dataset is 0.59. Select it as the acceptable upper limit and round up to 0.6. Then compare the S value with the threshold. Because , it is judged that the offset difference between this parameter group is acceptable. Furthermore, the common structure nodes and path numbers can be extracted to establish an offset alignment node group. This result shows that the control parameters executed by this pair of devices under the same operation steps have a high degree of consistency and have a mapping-dependent feature in terms of structure, which can be used for subsequent structure fusion and semantic binding processing.

[0031] The semantic fusion sub-module counts the frequency values of the node paths in the graph structure according to the offset alignment node group, assigns an interaction identifier to the path group with a frequency value greater than the co-occurrence threshold, updates the set of node pointing connection edges in the entity graph, and establishes an entity semantic fusion graph; Based on the offset-aligned node group, the semantic fusion sub-module traverses all path instances in the graph structure, counts the occurrence frequencies of each path number in the above node group, records the reference times of each path in the graph structure using a path statistical table, and filters out path groups with a co-occurrence frequency greater than the co-occurrence threshold. The co-occurrence threshold is set to the path occurrence times ≥ 8. Based on the summary statistics in the graph task dataset, the average occurrence times of the same-node path group is 7.4, which is set as the upper limit integer value 8. For example, for the path group "P-01, P-02, P-03" that appears 12 times, 9 times, and 11 times respectively, all meeting the threshold conditions, it is assigned an interaction identifier "SIG-P". In the graph database, the corresponding path pointing edge set is updated, and a field "Semantic_Bind" with its value "SIG-P" is added to indicate that this path group forms a semantic fusion relationship, and finally an entity semantic fusion graph is formed. In this graph structure, each fusion path corresponds to a group of offset-aligned nodes, with the corresponding fusion label and co-occurrence device group number marked. The entity semantic fusion graph includes a fusion path list, a semantic interaction identifier set, and a fusion device number relationship table.

[0032] Please refer to Figure 5 , the process path screening module includes: Based on the entity semantic fusion graph, the path extraction sub-module extracts the path structure under each group of nodes according to the node group pointing to the same processing target, records the process node numbers and connection control parameter edge information in the path, establishes a pairing relationship between the path number and the control parameter edge, and generates a process path structure quantity; After obtaining the node group in the entity semantic fusion graph that points to the same processing target, first, according to the unique identifier of each node in the graph, extract the node set aggregated under the same target, and read the processing step numbers contained in the paths of each node therein to construct a path structure index. Further, based on the annotation information of the connection edges, obtain each control parameter edge item contained in each path segment one by one, extract the device action instruction number and parameter field corresponding to the control parameter edge, and establish a one-to-one pairing array between the path number and the control edge item. For example, for the path P-101 corresponding to the node group G1, its process node sequence is N1 → N2 → N3, and the control edges are C1, C2, and C3 in turn. The corresponding pairing relationship is established as {P-101: [(N1, C1), (N2, C2), (N3, C3)]}, and then the generation of the multi-element mapping structure set of the path structure and the control parameter edge item is completed, which serves as the composition basis of the process path structure quantity.

[0033] Based on the process path structure quantity, the error judgment sub-module extracts the processed dimension values output by each process node and the processed dimension values of the corresponding upstream and downstream nodes, calculates the absolute difference between the nodes and compares it with the dimension error critical value in the IT tolerance standard grade. The formula is used: ; Obtain the processing deviation amount through calculation, summarize the error marks for the path nodes with a processing deviation amount greater than zero, screen out the corresponding path structures, obtain the compliance path number sequence, and establish the processing deviation elimination information; Among them, represents the normalized value of the processing dimension output by the upstream node of the th path node, represents the normalized value of the processing dimension input by the downstream node of the th path node, represents the absolute value of the measured error of the th path node, represents the IT tolerance critical value corresponding to the th path node, represents the total number of nodes participating in the judgment in the path, represents the overall processing deviation amount of the path, which is used to measure the cumulative level of processing errors among all nodes on this path; According to the control parameter edges involved in each path in the process path structure, extract the processing dimension fields in each path node, construct a processing dimension mapping set, and call the upstream dimension fields output by each process node and the input dimension fields received by its direct downstream nodes , subtract the two and take the absolute value respectively to obtain the actual processing deviation value of each node, and at the same time extract the absolute value of the measured error of each node , as well as its corresponding IT tolerance critical value , square and sum the error ratio terms , and combine the number of nodes in each path to calculate the total deviation expression , and the specific calculation formula is as follows: ; Among them: is the normalized value of the processing dimension output by the upstream node of the th path node, which is expressed as the target dimension provided by the previous device or node after processing; is the normalized value of the processing dimension input by the downstream node of the th path node, that is, the requirement specification of the input dimension for the next process; is the actual processing deviation value of the th path node, which is collected by the on-line dimension measurement system and is the offset between the dimension after processing and the target dimension; is the The IT-level critical value corresponding to each node is taken from the error tolerance limit value corresponding to the standard size paragraph; represents the total number of nodes participating in error judgment in the path; By dividing by the total number of nodes to obtain the average deviation amount, the calculation results can be made comparable.

[0034] As shown in Table 3, four path numbers P-101 to P-104 are selected. Their upstream and downstream size values, measured errors, and IT tolerance values are obtained respectively, and then substituted into the above formula in sequence. After calculating the size difference and error term of each path node, the mean value is obtained , the sum of the squares of the error terms is 1.20, the square root is 1.0954, and the final deviation amount is .

[0035] Table 3 Process path processing error data table:

[0036] This result indicates that the final deviation amount is less than zero, indicating that the actual processing errors of all nodes on the current path are within the tolerance range. This path belongs to a compliant path and is retained.

[0037] The benefit of the formula is that through the composite operation of the processing size difference and the standard error tolerance limit value, it can comprehensively measure the overall error distribution of all nodes in the path structure, not only reflecting the average error level, but also effectively weakening the influence of local extreme errors, thereby improving the robustness and accuracy of path rejection judgment and avoiding the situation of misjudging the path quality due to single-point errors.

[0038] The grade marking sub-module, based on the processing deviation rejection information, re-statistics the average difference of node pairs in each path according to the compliant path number sequence. According to the proportional distribution between the statistical result and the minimum tolerance value of the tolerance grade, the path is divided into multiple grade range segments, and a grade identification field is added to each path to establish a process path optimization map; After extracting all the retained path number sequences from the obtained processing deviation rejection amounts, re-statistics are respectively performed on the size differences formed by adjacent node pairs in each path, and the average difference of each path is calculated. For example, the size differences between three nodes of path P-101 are 0.10mm, 0.15mm, and 0.05mm in sequence, and the average difference is mm, compare this value with the minimum tolerance value in the tolerance grade. For example, if the minimum tolerance value is 0.02 mm, the current path is in the proportional interval of 5 times the tolerance range. When dividing, the grading basis is set as follows: a tolerance multiple less than 2 is defined as a first-level path, 2 to 4 times as a second-level path, and more than 4 times as a third-level path. Therefore, P-101 is classified into the third-level path paragraph, and a grade field Grade = 3 is added to it. At the same time, a grade annotation label is added to the atlas structure for subsequent calls of the process path optimization atlas.

[0039] Please refer to Figure 6 , the trend-driven warning module includes: The time-series acquisition sub-module calls the process path optimization atlas, extracts the periodic change sequences of the spindle vibration value, workpiece temperature rise value, and spindle speed value in the processing stage according to the time-series values of the operation control parameters of the devices associated with the path nodes, segments the parameter time series, and establishes a time series set for the processing stage; According to the recorded path numbers and their node sets in the process path optimization atlas, the manufacturing equipment identifiers associated with the nodes are extracted one by one, and the control parameter data recorded in the operation logs of each device or the industrial control system is called to collect the spindle vibration value, workpiece temperature rise value, and spindle speed value in its processing execution stage, forming three independent time series record sets. To enhance data adaptability, the three time series are associated according to the node timestamps, segmented with the time window as the periodic boundary. If the timestamps of nodes N1~N3 with the path number PT-201 during processing are 13 s, 27 s, and 41 s respectively, the spindle vibration value within each period is extracted with a 14 s period window , workpiece temperature rise value and spindle speed value , each period segment generates an independent sequence paragraph, and its segment number f is recorded, forming a time series set after segmentation. If a node lacks data in a period segment, it is filled by taking the median of the adjacent time window data to avoid breakpoint affecting trend recognition. Finally, a processing stage time series set based on the one-to-one mapping of path nodes and time windows is formed in the processing stage, and the time window length is set to a fixed 14 s to match the periodic frequency of the spindle vibration signal in the processing stage.

[0040] The fluctuation recognition sub-module calls the processing stage time series set, identifies the fluctuation directions of the three parameters of the spindle vibration value, workpiece temperature rise value, and spindle speed value in each period segment, judges the positive and negative relationship of the numerical differences between adjacent data points in each segment, and counts the number of paragraphs in the same direction. The formula is used: ; Calculate to obtain the fluctuation trend intensity, mark the continuous trend change period segments according to the size of the fluctuation trend intensity value, and establish the trend continuous intensity value; Among them, represents the increment normalization value of the spindle vibration value in the th cycle segment, represents the measurement time normalization value in the th cycle segment, represents the acceleration amplitude normalization value in the th cycle segment, represents the increment normalization value of the workpiece temperature rise in the th cycle segment, represents the normalization reference of the initial workpiece temperature rise, represents the increment normalization value of the spindle speed value in the th cycle segment, represents the spindle reference speed normalization value, represents the number of cycle segments including the temperature rise and speed trends, represents the number of cycle segments including the vibration trend, represents the trend continuous intensity value; Call each parameter sequence in the machining stage time series set, and compare the positive and negative relationships of the numerical differences between adjacent sampling points in the same cycle segment in turn. If the values of two consecutive sampling points are increasing, mark the fluctuation direction as positive, otherwise as negative. Statistically, the length of the consistent interval of the fluctuation directions of the three types of parameters in each cycle segment is calculated, and the increment normalization value of the spindle vibration value in each cycle segment is extracted. The normalization method is to divide the instantaneous change value by the maximum vibration range; the measurement time is extracted as the normalization value of the total sampling time of the cycle segment, and is standardized by dividing the total segment duration by the maximum segment duration; the acceleration amplitude is extracted as the ratio of the maximum vibration amplitude of the current cycle segment to the maximum value of the overall data segment. Substitute the above three parameters into the vibration trend term for cycle summation, and then take the temperature rise trend term and the speed trend term , and use their corresponding initial values , for normalization respectively. The final trend value is calculated as: ; In this formula: represents the spindle vibration change amplitude in the th cycle segment. After normalization, it represents the vibration fluctuation trend; represents the sampling duration of this cycle segment, reflecting the vibration duration interval; represents the acceleration peak amplitude in this segment, used to measure the vibration intensity; Indicates the temperature rise change of this section of the workpiece; Is the normalized value of the initial temperature of the workpiece during processing, used as the normalization reference for temperature rise; Is the spindle speed change; Is the normalized value of the reference speed at the start of spindle processing; Represents the total number of cycle segments containing the spindle vibration trend; Indicates the total number of cycle segments containing the temperature rise and speed trend terms; The operation logic is: by summing the vibration trend intensity terms in all cycle segments and subtracting the mean of the temperature rise and speed trend terms, a measure of the fluctuation trend intensity is formed , used to mark the strength of trend coherence.

[0041] Taking the vibration trend sampling data shown in Table 4 as an example for calculation: Table 4 Vibration Trend Cycle Segment Sample Data Table

[0042] Using the data of the first three cycle segments in the above table for calculation: The cumulative part of the vibration term is: ; The average part of the temperature rise and speed terms is: ; The final trend intensity value , this result indicates that the trend fluctuation amplitude is low, belonging to the stable section of the fluctuation trend.

[0043] The benefit of the formula is that through the combined evaluation of the three-dimensional parameters of vibration, temperature rise, and speed, the global nature of trend recognition is enhanced. Especially the expression form of the vibration part after the compound adjustment of acceleration and sampling time effectively alleviates the risk of misjudgment of abnormal fluctuations caused by measurement cycle differences, ensuring the relative consistency and periodic stability of trend intensity.

[0044] The abnormal annotation sub-module, based on the trend continuous intensity value, determines whether there is a situation where the continuous cycle trend intensity on the edge corresponding to the path node exceeds the control parameter fluctuation standard value, marks the edges that meet the conditions in the path graph as warning edges, and establishes an abnormal trend annotation set for the graph; According to the calculation result of the trend continuous intensity value , judge each path node one by one to determine whether the value in the cycle segment associated with its adjacent edge is continuously outside the standard fluctuation control parameter range. Set the control threshold to ±0.025. If it is satisfied within 3 consecutive cycle segments, or Then, a warning label is attached to this edge in the path structure, and this structure is marked as an abnormal trend area in the processing technology graph. For example, for the node pair (N2→N3) in the path numbered PT-203, there are three consecutive vibration trend values of 0.031, 0.037, and 0.029 respectively, all of which are higher than the standard value. Then its connecting edge is marked as a warning edge and enters the abnormal trend annotation set of the graph. This set is retained as the basis for subsequent path optimization, rejection, or regulation. The trend intensity of all edges is marked and displayed in an interval range, in the format of "0.029±0.004", indicating the concentration of fluctuations in the continuous trend performance of this edge under a specific path.

[0045] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A knowledge graph-driven manufacturing process optimization system, characterized in that The system includes: The graph node generation module obtains process documents, assembly flowcharts, and equipment control item descriptions in the numerical control processing line, constructs three types of knowledge graph entity nodes, establishes an entity basic network structure, and generates a process graph node set. The dependency relationship construction module, based on the process graph node set, maps and groups the cutting parameter ranges and equipment linkage marks to form a dependency linkage item table, and embeds the association path in the graph structure in the form of edge connection to generate a graph dependency structure network. The semantic entity matching module calls the graph dependency structure network, obtains the equipment operation entities with semantic heterogeneous expressions in the graph, sequentially performs sequential alignment judgment and numerical offset calculation, and sets the graph interaction connection according to the co-occurrence frequency in the structure network to generate an entity semantic fusion graph. The process path screening module, based on the entity semantic fusion graph, calls the control parameter edges connected to each process node in the path, statistically outputs the difference between the machining dimension value and the corresponding upstream and downstream nodes, establishes a path connectivity structure between nodes and marks the path preference level to generate a process path preference graph. The process path screening module includes: The path extraction sub-module, based on the entity semantic fusion graph, extracts the path structure under each group of nodes according to the node group pointing to the same machining target, records the process node numbers and the information of the connected control parameter edges in the path, establishes the pairing relationship between the path number and the control parameter edge, and generates the process path structure quantity.

2. The knowledge graph-driven manufacturing process optimization system according to claim 1, wherein The process graph node set includes an operation action node set, a component structure node set, a control parameter node set, a node pointing relationship set, and an initial graph connection structure. The graph dependency structure network includes a step number sequence, an equipment linkage mark set, a control parameter edge set, a parameter dependency connection path set, and a node linkage structure. The entity semantic fusion graph includes a semantic alignment node mapping, an offset difference annotation node, a parameter co-occurrence frequency record, a semantic heterogeneous node group, and a fused path topology structure. The process path preference graph includes a path node connectivity structure, an error elimination path list, a preference level mark group, a machining target matching index, and an optional path set.

3. The knowledge graph-driven manufacturing process optimization system according to claim 2, wherein The graph node generation module includes: The document parsing sub-module obtains process documents, assembly flowcharts, and equipment control item descriptions in the numerical control processing line, extracts the operation step content recorded in the process document, the part numbers represented in the assembly flowchart, and each type of equipment control field in the equipment control item description, splits the three types of text items into semantic paragraphs respectively, statistically counts the occurrence frequency of keywords, matches the operation word list, part dictionary, and control parameter standard items, and establishes the word frequency distribution data of operation type word items, structure type word items, and control type word items to generate a word frequency distribution ratio matrix. The structure recognition sub-module, based on the word frequency distribution ratio matrix, judges the content dependency relationship between operation type, structure type, and control type word items, sets the word item groups with a dependency frequency higher than the structure matching reference value as candidate node groups, performs entity label normalization processing on the word items in the candidate node groups and maps them to the standard item codes to obtain a word item set with unique entity identifiers, and generates a standard word item entity comparison set. The node construction sub-module calls the standard term entity control set, sets three types of graph nodes according to the term category based on the standard item codes of the operation class, structure class, and control class, configures a directed connection relationship between the three types of nodes according to the term dependency path, establishes a structured node network with complete pointing edges, and generates a process graph node set.

4. The knowledge graph-driven manufacturing process optimization system according to claim 3, characterized in that The dependency relationship construction module includes: The reference extraction sub-module, based on the process graph node set, extracts the corresponding numbers in the operation action nodes and control parameter nodes, obtains the processing instruction sequence number in the operation nodes and the device control code associated in the parameter nodes, and by comparing the frequencies of simultaneous occurrence of the numbers in the task flow list, filters out the node pairs with a reference frequency greater than two as the direct reference relationship units, and establishes the operation parameter reference frequency value; The parameter combination sub-module, according to the operation parameter reference frequency value, calls each group of operation step numbers and control parameter numbers, extracts the cutting depth, feed rate, and spindle speed under the corresponding process link, classifies and marks them in combination with the device type label, and takes the device number as the primary key, reconstructs the sequence of operation step numbers in chronological order, groups the parameter items into a fixed field set, and establishes the combined sequence mapping result; The path connection sub-module, according to the combined sequence mapping result, sets a directed connection path for the dependency relationship between the nodes according to the operation action nodes and control parameter nodes that call each grouped field in the graph, writes the path structure into the node edge set, and marks the path source device number to generate the graph dependency structure network.

5. The knowledge graph-driven manufacturing process optimization system according to claim 4, wherein The semantic entity matching module includes: The structure extraction sub-module calls the graph dependency structure network, extracts the corresponding application device number, operation sequence number, and parameter value field according to the parameter items corresponding to the control parameter group in each node, groups the device numbers in the parameter group, arranges the operation sequence fields in the group in ascending order of the number, maps the parameter value fields according to the node path number to generate a distribution set, and establishes the structure distribution relationship value; The difference judgment sub-module, according to the structure distribution relationship value, filters out the parameter group sets with the same sequence number under each device number, extracts the parameter value fields in the corresponding nodes, calculates the absolute value of the difference between the parameter values and the overall distribution standard deviation under the same number, operates to obtain the relative difference offset degree, and according to the parameter groups with a relative difference offset degree value lower than the offset difference threshold, extracts the structure common nodes and path numbers to generate the offset alignment node group; The semantic fusion sub-module, according to the offset alignment node group, counts the frequency value of the node path in the graph structure, assigns an interaction identifier to the path group with a frequency value greater than the co-occurrence threshold, updates the node pointing connection edge set in the entity graph, and establishes the entity semantic fusion graph.

6. The knowledge graph-driven manufacturing process optimization system according to claim 5, wherein The process path screening module further includes: The error judgment sub-module extracts the processed dimension values output by each process node and the processed dimension values of the corresponding upstream and downstream nodes according to the process path structure quantity, calculates the absolute difference between the nodes, and compares the difference with the dimension error critical value in the IT tolerance standard grade to obtain the processing deviation quantity through operation. It summarizes the error marks for the path nodes with a processing deviation greater than zero, filters out the corresponding path structures, obtains the compliance path number sequence, and establishes the processing deviation elimination information. The grade marking sub-module, based on the processing deviation elimination information and according to the compliance path number sequence, re-statistics the average difference of node pairs in each path. According to the proportional distribution between the statistical result and the minimum tolerance value of the tolerance grade, the path is divided into multiple grade range segments, and a grade identification field is added to each path to establish a process path optimization map.

7. The knowledge graph-driven manufacturing process optimization system according to claim 6, characterized in that The system further includes: The trend-driven warning module calls the process path optimization map, collects the real-time control parameter time series values during operation, compares the consistency of the fluctuation directions in each cycle segment and counts the number of continuous cycles. When there is a continuous consistent fluctuation direction at the node and the number of cycles exceeds the standard value, the corresponding edge in the map is marked as a warning edge to generate a map abnormal trend annotation set. The map abnormal trend annotation set specifically includes a cycle consistency marked edge, a continuous offset path index, an equipment abnormal fluctuation label, a warning path connection identifier, and a dynamic trend annotation mapping table.

8. The knowledge graph-driven manufacturing process optimization system according to claim 7, characterized in that The trend-driven warning module includes: The time series acquisition sub-module calls the process path optimization map, extracts the periodic change sequences of the spindle vibration value, the workpiece temperature rise value, and the spindle speed value in the processing stage according to the operation control parameter time series values of the equipment associated with the path nodes, segments the parameter time series, and establishes a processing stage time series set. The fluctuation identification sub-module calls the processing stage time series set, identifies the fluctuation directions of the three parameters of the spindle vibration value, the workpiece temperature rise value, and the spindle speed value in each cycle segment, judges the positive and negative relationship of the numerical differences between adjacent data points in each segment and counts the number of paragraphs with the same direction, calculates the fluctuation trend intensity, marks the continuous trend change cycle segments according to the size of the fluctuation trend intensity value, and establishes a trend continuous intensity value. The abnormal annotation sub-module, based on the trend continuous intensity value, judges whether there is a situation where the continuous cycle trend intensity of the edge corresponding to the path node exceeds the control parameter fluctuation standard value, marks the edges that meet the conditions in the path map as warning edges, and establishes a map abnormal trend annotation set.

Images