Low-code development interface AI intelligent generation method and system

Abstract

The application discloses a low-code development interface AI intelligent generation method and system, relates to the technical field of low-code development technology, and comprises the following steps: extracting a first production line identifier and a first process route identifier, generating a first propagation path through reverse positioning in a workshop plan coordinate system, and converting a spoken address into a physical topology diffusion source; searching for an existing form item record to generate a first fence area, and converting a discrete field attribution state into a geometric space barrier; superimposing processing on the two to extract a first conflict area, distributing a disambiguation strategy according to the position of the conflict area, and accurately identifying same-name different-meaning business objects. The application breaks the flat search mode of only judging the same field name, completes workshop topology level pre-rehearsal and quality control before the field falls into the canvas, avoids wrong association of the same name spoken address in the multi-production line interweaving scene, and guarantees the accuracy of the look board summary and material tracing.

Description

Technical Field

[0001] This application relates to the field of low-code development technology, and in particular to a method and system for generating AI-powered low-code development interfaces. Background Technology

[0002] Existing low-code interface generation methods typically retrieve fields with the same name from a pre-made form item library based on the colloquial terms entered by business personnel and place them onto the corresponding canvas.

[0003] Furthermore, at a machine processing plant where stamping, injection molding, and final assembly lines were operating simultaneously, the stamping line foreman verbally stated "add one column for material loading," the injection molding line foreman verbally stated "add one column for material loading," and the final assembly line foreman verbally stated "add one column for material loading."

[0004] The existing low-code interface generation method, upon hearing the colloquial expression "material loading," selects a form item named "material loading" from a pre-made form item library and places it onto the work order canvases of the three production lines. All three work orders have the same field name: "material loading." However, "material loading" in the stamping production line refers to steel coil loading, in the injection molding production line it refers to plastic granule loading, and in the final assembly production line it refers to component loading. The physical objects, units of measurement, and upstream and downstream processes of these three production lines are completely different.

[0005] When the workshop's large screen summarizes the daily material loading volume, the existing low-code development interface generation method treats the three fields with the same name "Material Loading" as the same type of object and merges them, resulting in the number of steel coils, the number of bags of plastic granules, and the number of parts being superimposed in the same column.

[0006] Furthermore, when performing material traceability, the existing low-code development interface generation method uses the "material loading" field name to perform a reverse lookup, simultaneously returning the steel coil batch, granular material batch, and component batch, causing confusion in the traceability chain.

[0007] Existing low-code interface generation methods assume that the same colloquial term corresponds to the same business object, only determining whether the field names are the same, without distinguishing the production line location, workstation location, and material context of the colloquial term. In workshops with multiple production lines, colloquial expressions with the same name but different meanings are mistakenly associated with the same business object, causing both Kanban summary and material traceability to be distorted. Summary of the Invention

[0008] In view of the aforementioned problems, this application is hereby filed.

[0009] Therefore, this application provides a low-code development interface AI intelligent generation method and system, which can solve the problem of misassociating the same colloquial terms with the same business object in workshops with multiple production lines, resulting in distortion of Kanban summary and material traceability.

[0010] To solve the above-mentioned technical problems, this application provides the following technical solution: In the first aspect, this application provides a low-code development interface AI intelligent generation method, including: obtaining a spoken voice package input by a workshop team leader through an industrial tablet, performing natural language semantic analysis on the spoken voice package, extracting spoken title text, and extracting a first production line identifier, a first access point identifier, and a first process route identifier from the spoken conversation context based on semantic understanding. The first production line identifier is used to describe the production line that the narrator has logged into and bound to; The first access point identifier is used to describe the workshop wireless access point connected to the industrial tablet held by the speaker. The first process route identifier is used to describe the process route to which the work order opened by the speaker in this session belongs; Based on the first production line identifier and the first access point identifier, the first verbal anchor point is obtained by reverse lookup in the workshop floor plan coordinate system; based on the first process route identifier, the production line process layout description is retrieved, and the process node responsible for receiving materials is determined from the production line process layout description as the first intention landing point; A first propagation path is generated based on the first spoken anchor point and the first intention landing point; Retrieve existing form item records with the same name as the spoken title text from the pre-made form item library, locate the field ownership center point and associated endpoint on the workshop floor plan according to the production line identification of each existing form item record, and generate a first fenced area based on the field ownership center point and the associated endpoint. The first propagation path and the first fenced area are graphically overlaid, and the overlapping line segments are extracted as the first conflict zone. Based on the segment position of the first conflict zone on the first propagation path, a field disambiguation handling strategy is distributed.

[0011] Preferably, the step of performing natural language semantic analysis on the spoken voice package to extract the spoken address text includes: The spoken voice packet is converted into a text sequence, entity recognition is performed on the text sequence, and business object descriptive words are extracted as the spoken title text; The text sequence is classified and identified to determine that the business action corresponding to the colloquial address text is a new field. Based on semantic dependency analysis, production line environment descriptors, network environment descriptors, and process environment descriptors are extracted from the context of the text sequence and mapped to the first production line identifier, the first access point identifier, and the first process route identifier, respectively.

[0012] Preferably, generating the first propagation path based on the first spoken anchor point and the first intention landing point includes: Retrieve the corresponding production line process layout description based on the first process route identifier; The production line process layout description includes embedded relative coordinate references of the workshop plane and material connection relationships for each process node from the raw material warehouse interface point to the finished product off-line station; Based on semantic similarity comparison, the first verbal anchor point is mapped to the coordinates of the process node in the production line process layout description and defined as a verbal process point. The coordinates of the process node closest to the raw material warehouse interface point and responsible for receiving materials in the production line process layout description are defined as the first intention landing point. In the coordinate system of the workshop floor plan, a vector connection is made with the verbal process point as the starting point and the first intention landing point as the ending point to generate the first propagation path with directionality.

[0013] Preferably, generating the directional first propagation path includes: The material connection lines on the workshop floor plan are traversed along the direction of the first propagation path. When the traversal reaches the graphic boundary of the adjacent production line section, the real-time material occupancy status of the adjacent production line section is read through the real-time process status database. Based on semantic matching, the real-time material occupancy status of the adjacent production line section is matched with the material injection direction corresponding to the spoken term text to establish a connection relationship. When a succession relationship is determined, the identifier and graphic outline of the adjacent production line section are connected to the first propagation path, and the next propagation path is regenerated from the process boundary on the other side of the adjacent production line section away from the first verbal anchor point. When it is determined that there is no connection, the extension of the first propagation path in the corresponding direction is terminated.

[0014] Preferably, terminating the extension of the first propagation path in the corresponding direction when it is determined that no connection exists includes: When traversing to an adjacent production line section, the production line identifier and material category identifier of the adjacent production line section are retrieved simultaneously. When the material category received by the adjacent production line section is not the same as the material category of the production line corresponding to the first production line identifier, the material category termination condition is triggered, and the propagation path extension in the corresponding direction is stopped. When the production line identifier of the adjacent production line section is inconsistent with the first production line identifier and there is no process route association with shared raw material supply, the production line isolation termination condition is triggered, and the propagation path extension in the corresponding direction is stopped. When the real-time status of the adjacent production line section is the material end consumption point, the material end termination condition is triggered, and the propagation path extension in the corresponding direction is stopped. The production line section that triggers the termination condition is defined as the intended cut-off section, and the boundary position of the intended cut-off section in the workshop plan is written into the end coordinate slot of the first propagation path.

[0015] Preferably, generating the first fenced area based on the field's home center point and the associated endpoint includes: Based on knowledge graph association query, determine the field belonging center point and the associated endpoint of the existing form item record in the workshop floor plan coordinate system corresponding to the production line to which the field belongs; In the coordinate system of the workshop floor plan, the field ownership center point is connected to each of the associated endpoints by a straight line segment to generate the field ownership connection line; When the field belonging connection line satisfies the closed polygon enclosure condition and the associated endpoints are distributed in different quadrants of the field belonging center point, the adjacent associated endpoints are connected end to end according to the outer geometric arrangement order of the associated endpoints on the workshop floor plan to generate a closed first enclosure area as the first fence area. When the field belonging connection line does not meet the closed polygon enclosure condition, the field belonging connection line is used as the geometric center line, and the field is geometrically expanded outward at equal distances to both sides of the center line according to the preset belonging width parameter to generate a first strip area as the first fence area.

[0016] Preferably, the step of graphically overlaying the first propagation path and the first fenced area to extract overlapping line segments as the first conflict zone includes: The first fenced area is overlaid on the workshop topology layer where the first propagation path is located. The overlapping line segments in the first propagation path that fall completely inside the first fenced area are cut off and extracted as candidate conflict segments through line segment clipping. Parse the field creation time in the existing form item record, read the preset ownership retention time of the corresponding cross-production line same-name field from the system configuration table, and accumulate the field creation time and the preset ownership retention time on the time axis to generate the field ownership valid time period; Extract the current oral trigger time, determine whether the current oral trigger time is within the valid time period to which the field belongs, and read the effective activation identifier from the existing form item record; When the current verbal trigger time is within the valid time period of the field and the activation flag is still enabled, the candidate conflict segment is activated and written into the first conflict area; when the current verbal trigger time exceeds the valid time period of the field or the activation flag is disabled, the candidate conflict segment is erased.

[0017] Preferably, the step of distributing the field disambiguation handling strategy based on the segment position of the first conflict zone on the first propagation path includes: The first dividing point is the geometric boundary of the production line section symbol where the first verbal anchor point is located on the workshop floor plan, and the second dividing point is the geometric boundary of the section symbol intended to be cut off on the workshop floor plan. Based on the first segmentation point and the second segmentation point, the first propagation path is divided into a first source segment, a first relay segment, and a first boundary segment; The first source segment encompasses the physical location of the current production line section of the narrator and the graphic area adjacent to the process pin; The first relay segment is sandwiched between the first source segment and the first boundary segment; The first boundary segment encloses the graphic area of ​​the section intended to be cut off. The first conflict zone is spatially matched with the first source segment, the first relay segment, and the first boundary segment, and the corresponding field disambiguation handling strategy is distributed according to the matching results.

[0018] Preferably, the step of distributing the corresponding field disambiguation processing strategy based on the matching result includes: When the first conflict zone overlaps with the first source segment, the writing channel from the form item corresponding to the spoken title text to the work order canvas is cut off. The conflict cause description text is generated based on natural language understanding processing. The conflict cause description text, along with the production line identifier, field creation time, and field material category of the existing form item record, is written into the conflict source data slot of the field interception area. The position to be placed on the work order canvas is flashed and highlighted on the industrial tablet screen, and a warning window pops up, displaying the conflict cause description text. In response to the team leader's confirmation operation, the production line prefix is ​​added to the form item corresponding to the colloquial title text based on natural language generation processing and placed into the work order canvas. When the first conflict zone coincides with the first relay segment, upstream and downstream process nodes belonging to different production lines are searched along the upstream and downstream of the first conflict zone. A relay verification line is pulled across the first fence area in the workshop floor plan memory. A breakpoint crossing verification result is generated based on topological connectivity verification. When the breakpoint crossing verification result is topological convergence, in response to the triggering of multiple operators entering their work numbers and dynamic passwords to complete the dual electronic signature verification, the material category suffix of the spoken title text is marked on the form item based on natural language generation processing and placed into the work order canvas.

[0019] Secondly, this application also provides a low-code development interface AI intelligent generation system, including: a spoken semantic parsing module, used to obtain spoken voice packets input by the workshop team leader through an industrial tablet, perform natural language semantic parsing on the spoken voice packets, extract spoken title text, and extract the first production line identifier, the first access point identifier, and the first process route identifier from the spoken conversation context based on semantic understanding. The propagation path generation module is used to obtain the first verbal anchor point by reverse lookup and positioning in the workshop floor plan coordinate system based on the first production line identifier and the first access point identifier, to determine the first intention landing point by retrieving the production line process layout description based on the first process route identifier, and to generate the first propagation path based on the first verbal anchor point and the first intention landing point. The fence area generation module is used to retrieve existing form item records with the same name as the spoken title text from the pre-made form item library, locate the field ownership center point and associated endpoint on the workshop floor plan according to the production line identification of each existing form item record, and generate the first fence area based on the field ownership center point and the associated endpoint. The conflict zone purification module is used to perform graphic overlay processing on the first propagation path and the first fenced area, and extract the overlapping line segments as the first conflict zone. The disambiguation strategy distribution module is used to distribute field disambiguation handling strategies according to the segment position of the first conflict zone on the first propagation path.

[0020] Implementing this application will have the following beneficial effects: This application provides a low-code development interface AI intelligent generation method and system. This invention extracts the first production line identifier and the first process route identifier, and uses reverse lookup to locate the first verbal anchor point and the first intention landing point in the workshop floor plan coordinate system to generate the first propagation path, transforming abstract verbal terms into concrete physical topological diffusion sources. It generates a first fenced area by retrieving existing form item records, transforming discrete field ownership status into geometric spatial barriers. The first propagation path and the first fenced area are graphically overlaid to extract the first conflict zone, and field disambiguation handling strategies are distributed based on the segment position of the first conflict zone on the first propagation path, accurately identifying business objects with the same name but different meanings from the source segment, relay segment, and boundary segment. This invention breaks away from the flat retrieval method that only judges the same field name, completing a workshop topology-level pre-visualization and check before the field falls into the canvas, avoiding the misassociation of verbal terms with the same name as the same business object in multi-production line intertwined scenarios, and ensuring the accuracy of Kanban summary and material traceability. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a low-code development interface AI intelligent generation method involved in this application; Figure 2 This is a flowchart illustrating the propagation path extension and termination judgment of a low-code development interface AI intelligent generation method involved in this application; Figure 3 This is a schematic diagram of the conflict segment activation and erasure judgment process of a low-code development interface AI intelligent generation method involved in this application; Figure 4 This application relates to a data flow and timing interaction topology among the modules of a low-code development interface AI intelligent generation system. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0024] Example 1, referring to Figures 1-3 This is the first embodiment of the present application, which provides a low-code development interface AI intelligent generation method, including: This application provides a method that can effectively solve the problems mentioned above. The following will describe in detail how to implement the low-code development interface AI intelligent generation method with multiple embodiments. Figure 1 A flowchart illustrating a low-code development interface AI intelligent generation method is shown, including: Step S1: Obtain the spoken voice packet input by the workshop team leader via an industrial tablet. Perform natural language semantic analysis on the spoken voice packet to extract the spoken address text. Based on semantic understanding, extract the first production line identifier, the first access point identifier, and the first process route identifier from the context of the spoken conversation. The first production line identifier describes the production line that the speaker has logged into and bound to. The first access point identifier describes the workshop wireless access point that the speaker's handheld industrial tablet is connected to. The first process route identifier describes the process route to which the work order opened by the speaker in this conversation belongs. It should be noted that in complex workshop environments with multiple production lines intertwined, business personnel often use colloquial language to express their requirements for the user interface.

[0025] However, due to differences in the physical referents of the same colloquial term across different production lines, directly performing flat searches based on field names can lead to severe misalignment. To complete a workshop-level topology pre-visualization and control before fields are included in the canvas, it is necessary to convert spoken speech packets containing simulated voice waveforms into computable digital feature sequences, and extract boundary parameters of physical location and process flow from the multidimensional environmental context. These parameters serve as the source input benchmark for subsequent spatial topology calculations.

[0026] In some embodiments, the step of performing natural language semantic parsing on the spoken voice packet to extract the spoken address text includes: Step S1.1: Convert the spoken voice packet into a text sequence, perform entity recognition on the text sequence, and extract the business object descriptive words as the spoken title text; Specifically, spoken speech packets, as a continuous acoustic analog signal, must undergo signal discretization and acoustic feature mapping processing before they can be converted into computable data.

[0027] Furthermore, upon receiving the spoken voice packet, the simulated sound wave is sliced ​​at a fixed time step, and the Mel-frequency cepstral coefficients within each time slice are extracted. Let the index of the time slice be denoted as . Then the first Acoustic feature vectors of each slice It can be obtained by taking the logarithm after performing a discrete Fourier transform and mapping to the Mel scale.

[0028] Furthermore, the acoustic feature vector sequence The input is fed into a Hidden Markov Statistical Model for state transition probability calculation by traversing a preset state transition matrix. With the observation probability matrix The phoneme state sequence corresponding to the maximum posterior probability is calculated using the Viterbi optimal path, and then the phoneme state sequence is concatenated into a text sequence based on the pronunciation dictionary. ,in For the first in the text Each word.

[0029] Furthermore, after obtaining the text sequence, entity boundaries are labeled using a conditional random field mathematical model. By constructing a joint probability distribution function of the observed sequence and the hidden state sequence, the state path that maximizes the probability value is sought. The specific calculation formula is as follows: ; In formula (1), Representing the The entity status label corresponding to each word segmentation and These represent the characteristic functions of nodes and edges, respectively. and This represents the corresponding allocation ratio. This represents the normalization factor.

[0030] Furthermore, after calculating the optimal state sequence through Viterbi optimization, the word segments labeled as business object descriptors are extracted as spoken title text. For example, in the voice package of the injection molding workshop foreman verbally describing material loading, after the above transformation and labeling process, the word "material loading" is labeled as a business object descriptor and extracted as spoken title text.

[0031] In some embodiments, the feature functions of nodes and edges can be quantitatively solved by constructing a binary indicator mapping relationship. Node feature functions Read the current word segmentation Part-of-speech tagging results and entity state labels The function outputs an integer 1 if the combination matches a pre-summarized part-of-speech tag pair, and an integer 0 otherwise. (Edge feature function) Read adjacent entity status labels and The function outputs an integer 1 if the state transition pair belongs to a pre-set valid business tag flow path, and an integer 0 otherwise.

[0032] It should be noted that by comparing the output value of the above function with the corresponding allocation ratio... and By performing a product operation and an accumulation operation, the confidence level of the mapping from the observed sequence to the hidden state sequence can be numerically quantified.

[0033] Step S1.2: Perform intent classification and recognition on the text sequence to determine that the business action corresponding to the spoken address text is a new field; Furthermore, after obtaining the text sequence, it is necessary to determine the business intent represented by the text sequence. Intent classification and recognition are performed by constructing a multi-class support vector machine model.

[0034] Furthermore, the text sequence Mapped to word vector matrix And calculate the geometric center point of the word vector matrix in multidimensional space. The support vector machine model constructs the separating hyperplane equation in multidimensional space. ,in It is the normal vector. This is a bias term.

[0035] Furthermore, the geometric center point Substitute the equation of the separating hyperplane into the equation and calculate its distance to each type of hyperplane. The calculation formula is: .

[0036] When the interval distance of a certain category reaches a maximum value and exceeds a preset distance threshold, the text sequence is determined to belong to that category. For example, when a text sequence is mapped with a "Add Material" column, if its geometric center point is closest to the hyperplane interval distance of the newly added field category, then the corresponding business action is determined to be a newly added field. The preset distance threshold mentioned here is preferably set in the range of 0.3 to 0.6, and is specifically obtained by collecting historical spoken text datasets, calculating the minimum distance from the boundary samples of each category to the reference hyperplane, and performing statistical fitting.

[0037] Step S1.3: Based on semantic dependency analysis, extract production line environment descriptors, network environment descriptors, and process environment descriptors from the context of the text sequence, and map them to the first production line identifier, the first access point identifier, and the first process route identifier, respectively.

[0038] In this embodiment, semantic dependency analysis is used to mine the syntactic dominance relationships between various word segments in a text sequence, thereby accurately extracting implicit contextual constraints. A dependency syntax tree is constructed using a shift-reduce parser based on shift and reduce operations. The parser maintains a word segment stack and a sequence to be processed. In each step, based on the top element of the stack and the first element of the sequence to be processed, it calculates local feature scores to determine whether to perform a shift, left reduce, or right reduce operation. This approach uses spoken terminology text as the core node and extends outwards to find modifying dependency nodes.

[0039] Furthermore, text fragments corresponding to dependent nodes with dependency arc tags of production line modification, network modification, and process modification are extracted. For the extracted environmental descriptors, an inverted index table is constructed and precisely matched in the workshop metadata dictionary. For example, if the extracted production line environmental descriptor is "stamping line," the corresponding string code L01 is found in the metadata dictionary and mapped to the first production line identifier.

[0040] If the extracted network environment descriptor is AP2, it is mapped to the first access point identifier.

[0041] If the process environment description term is mold assembly, it is mapped to the first process route identifier. Furthermore, if a complete environment description term cannot be directly extracted from the dependency parsing tree, the completion logic based on the sliding window mentioned above is invoked to backtrack several interaction rounds of text records and use part-of-speech tagging to extract noun entities for fuzzy matching completion.

[0042] For example, if the process environment description is missing, and the phrase "mold assembly work order" is found in the previous records, then the mold assembly is extracted and mapped. The preferred setting range for the size of the preset sliding window mentioned here is 2 to 5 rounds, and the specific method of obtaining it is based on the median value of the average continuous interaction conversation length of the workshop team leader.

[0043] Step S2: Based on the first production line identifier and the first access point identifier, locate the first verbal anchor point in the workshop floor plan coordinate system; retrieve the production line process layout description based on the first process route identifier, and determine the process node responsible for receiving materials from the production line process layout description as the first intention landing point; generate a first propagation path based on the first verbal anchor point and the first intention landing point. It should be pointed out that in order to transform abstract colloquial terms into concrete physical topological diffusion sources, it is necessary to establish a mapping channel from logical identifiers to spatial geometric coordinates. By extracting multidimensional environmental identifiers, cross-comparison and retrieval are performed in the coordinate system of a two-dimensional workshop floor plan to locate the initiation point of the verbal behavior in physical space, and the intended destination point is determined by combining the direction of process flow, thereby constructing a spatial connection curve with directional orientation.

[0044] In some embodiments, generating the first propagation path based on the first spoken anchor point and the first intent landing point includes: Step S2.1: Retrieve the corresponding production line process layout description according to the first process route identifier; the production line process layout description contains the workshop plane relative coordinate reference and material front and back connection relationship of each process node from the raw material warehouse interface point to the finished product off-line station; Furthermore, the production line process layout description is a set of spatial attributes stored in a relational database in the form of a structured topology diagram.

[0045] Furthermore, upon receiving the first process route identifier, a structured query statement is constructed using it as the primary key, and a query request is sent to the database. The returned production line process layout description contains a set of process nodes. Each process node All include two-dimensional planar relative coordinates and the adjacency matrix of material succession relationships .

[0046] Furthermore, in the adjacency matrix In, elements The rules for determining the value are as follows: ; In formula (2), the variable and This represents the index number of the node. This is achieved by traversing the adjacency matrix. This allows us to know the complete material flow relationship from the raw material warehouse interface point to the finished product delivery station. For example, in the production line process layout description of the final assembly line, there are coordinate data of several process nodes such as component loading, chassis assembly, and interior installation, as well as the adjacency matrix between them, thereby constructing a complete relative coordinate benchmark of the workshop plane and the material flow relationship.

[0047] Step S2.2: Based on semantic similarity comparison, the first verbal anchor point is mapped to the coordinates of the process node in the production line process layout description, and defined as a verbal process point; Meanwhile, the first verbal anchor point is a spatial coordinate point obtained by reverse lookup in the workshop floor plan coordinate system based on the first production line identifier and the first access point identifier. To associate this coordinate point with the discrete nodes in the production line process layout description, the coordinate point needs to be calculated. With each process node Euclidean distance between coordinates The calculation formula is as follows: ; In formula (3), and The x and y coordinates represent the first oral anchor point. and Representing the The horizontal and vertical coordinates of each process node are calculated. After calculating all distance values, the process node that achieves the minimum value is selected, and its coordinates are defined as the descriptive process point. For example, in the actual application in the stamping workshop, the coordinates obtained by reverse lookup of the first verbal anchor point are located in the X3Y4 ​​area of ​​the workshop plan. By calculating the distance between this coordinate and the three process nodes of steel coil feeding, punching, and scrap collection, it is found that the distance to the steel coil feeding process node reaches a minimum value. Therefore, the coordinates of the steel coil feeding process node are defined as the verbal process point.

[0048] Step S2.3: Define the coordinates of the process node closest to the raw material warehouse interface point and responsible for receiving materials in the production line process layout description as the first intention landing point; For example, the raw material warehouse interface point is the physical starting boundary for all purchased materials to enter the production line.

[0049] Furthermore, in the process node set In the process, a subset of candidate nodes whose attribute labels contain the responsibility of receiving materials are selected. Calculate the subset of candidate nodes. Coordinates of each node in the middle to the coordinates of the raw material warehouse interface point Manhattan distance The calculation formula is as follows: ; In formula (4), and The horizontal and vertical coordinates represent the candidate nodes.

[0050] Furthermore, among all the calculated Manhattan distances, the candidate node that achieves the minimum value is defined as the first intention landing point. For example, in an injection molding production line, the raw material silo interface point is the granule buffer silo. Calculations show that the Manhattan distance from the granule loading node to the buffer silo reaches a minimum value. Therefore, the coordinates of the granule loading node are defined as the first intended landing point. In this embodiment, the tolerance range for the Manhattan distance calculation is preferably set to 0.2 meters to 0.8 meters. The specific method of obtaining this range is based on the physical alignment error accuracy of the workshop transport trolley.

[0051] Step S2.4: Generate a first propagation path based on the first spoken anchor point and the first intention landing point; Understandably, after determining the starting and ending points in physical space, a topological connection curve reflecting the direction of material and instruction flow must be established as a reference line for subsequent judgment of name conflicts.

[0052] In some embodiments, generating the directional first propagation path includes: Step S2.41: Traverse the material connection lines on the workshop floor plan along the direction of the first propagation path. When traversing to the graphic boundary of the adjacent production line section, read the real-time material occupancy status of the adjacent production line section through the real-time process status database. For example, material connection lines are a set of vector line segments on a workshop floor plan that represent physical conveyor belts or navigation trajectories.

[0053] Furthermore, after generating the initial connection line of the first propagation path, from the verbal process point... Depart, following the path leading to the initial intended landing point. In the direction of scanning, the vector line segments of the material connection line that are crossed are scanned segment by segment.

[0054] When the scanning pointer touches the polygonal boundary of an adjacent production line section, a data retrieval request is sent to the real-time process status database to read the current real-time material occupancy status vector of that adjacent production line section. .

[0055] It should be noted that vectors It includes fields such as material identifier, material quantity, and material dwell time stamp. For example, when the propagation path crosses the boundary between the stamping production line and the final assembly production line, the real-time material occupancy status of a certain section of the final assembly production line is read as part code A001, quantity 20 pieces, and dwell time stamp is 5 minutes before the current time.

[0056] Step S2.42: Based on semantic matching, match the real-time material occupancy status of the adjacent production line section with the material injection direction corresponding to the spoken term text. On the other hand, semantic matching is performed by calculating the cosine similarity between the material attribute vector corresponding to the spoken title text and the real-time material occupancy status vector.

[0057] Furthermore, the spoken address text is mapped to a material attribute vector through word embedding. Map the material identifier in the real-time material occupancy status to a state vector. Calculate the cosine similarity between the two. The formula is as follows: ; In formula (5), Index representing the dimension of the vector. This represents the total dimension of the vector. and Representing the material attribute vector and state vector respectively in the th... The components of a dimension.

[0058] When cosine similarity If the similarity exceeds a preset similarity threshold, a connection is considered to exist. For example, in the context of stamping, the material attribute vector mapped from the colloquial term "material loading" is a steel coil feature vector, while the state vector read from an adjacent section is a plastic granule feature vector. The cosine similarity calculated between the two is 0.12, which is lower than the preset threshold of 0.75, therefore no connection is considered. The preferred setting range for the preset similarity threshold mentioned here is 0.7 to 0.85. Specifically, it is obtained by cross-validating a historical material confusion sample set and selecting a segmentation point with the highest accuracy.

[0059] Step S2.43: When it is determined that there is a succession relationship, the identifier and graphic outline of the adjacent production line section are connected to the first propagation path, and the next propagation path is regenerated from the process boundary on the other side of the adjacent production line section away from the first verbal anchor point. Specifically, when the cosine similarity meets the acceptance condition, the unique identifier of the adjacent production line section is written into the node queue of the first propagation path, and the sequence of the polygonal vertex of the graphic outline of the section on the workshop floor plan is added to the geometric buffer of the propagation path.

[0060] Furthermore, after the addition is completed, the addressing module jumps to the process boundary on the other side of the adjacent production line section, away from the first described anchor point. It uses the midpoint of this boundary as the new starting point and restarts the spatial vector connection calculation logic to generate the next propagation path. For example, after determining that there is a component connection between the stamping line and the final assembly line, the identifier and graphic outline of the final assembly line are added to the path, and the path is extended downstream from the western boundary of the final assembly line.

[0061] Step S2.44: When it is determined that there is no connection, the extension of the first propagation path in the corresponding direction is terminated.

[0062] Preferably, when the semantic matching determination does not meet the acceptance conditions, it means that the material flow chain has physically broken at this point, and the path extension logic must be cut off immediately to prevent the generation of a topological connection illusion that crosses unrelated production lines.

[0063] Figure 2 The topology of the data body for determining the extension and termination of the first propagation path when crossing the boundary of a work section is shown in this embodiment of the application. When traversing along the direction of the first propagation path to the graphical boundary of an adjacent production line section, the real-time material occupancy status of that section is read through the real-time process status database, and a semantic matching is used to determine whether there is a continuity relationship between the status and the material injection direction corresponding to the spoken terminology text. When a continuity relationship is determined to exist, the system connects the identifier and graphical outline of the adjacent production line section to the current path, and regenerates the next propagation path from the process boundary on the other side to achieve topology extension; when a continuity relationship is determined not to exist, the system immediately triggers termination conditions such as material category or production line isolation, stops the path extension in the corresponding direction, and forcibly writes the boundary position of the intended truncated section into the endpoint coordinate slot of the first propagation path, thereby completing the precise trimming and closed-loop construction of the physical flow chain.

[0064] In some embodiments, terminating the extension of the first propagation path in the corresponding direction when it is determined that no succession relationship exists includes: Steps S2.441 to S2.443 are explained in conjunction with the following: when traversing to an adjacent production line section, the production line identifier and material category identifier of the adjacent production line section are retrieved simultaneously. When the material category received by the adjacent production line section is not the same as the material category of the production line corresponding to the first production line identifier, the material category termination condition is triggered, and the propagation path extension in the corresponding direction is stopped. When the production line identifier of the adjacent production line section is inconsistent with the first production line identifier and there is no process route association with shared raw material supply, the production line isolation termination condition is triggered, and the propagation path extension in the corresponding direction is stopped. In particular, when handling path termination conditions, multiple physical attribute comparison logic must be introduced to prevent misjudgments.

[0065] When the traversal pointer touches the graphic boundary of an adjacent production line section, the production line identifier to which that section belongs is simultaneously retrieved from the workshop basic information ledger. Material category identifier .

[0066] Furthermore, Material category identifier corresponding to the first production line identifier A comparison is performed. Since material categories are typically stored in a multi-level classification tree, the length of the longest common subsequence is used to measure the similarity between the two categories. Let the string sequences corresponding to the material category identifiers be as follows: and , length is and Construct a two-dimensional state matrix The state transition equation is as follows: ; In formula (6), and The character index representing the string.

[0067] When the calculated longest common subsequence length When the material category depth is less than a preset category depth threshold, a material category termination condition is triggered, immediately stopping the extension of the propagation path in the corresponding direction and generating a material isolation log record. For example, the first production line is identified as the stamping line, and its material category is identified as the metal raw material category, while the material category of the adjacent section is identified as the plastic raw material category. The length of their longest common subsequence is only 1, which is much less than the preset category depth threshold of 3, thus triggering the material category termination condition. This operation ensures that even if the sections are physically adjacent, as long as the physical properties of the materials they process are fundamentally different, the tentative extension of the propagation path will be immediately terminated, cutting off the possibility of cross-business object association from the material flow dimension.

[0068] Furthermore, at the production line identification comparison level, when determining... and If there is a discrepancy, it is necessary to further consult the cross-production line shared raw material supply table.

[0069] Furthermore, and Use it as a join primary key query to access the shared raw material supply adjacency table in a relational database.

[0070] If the query returns an empty set, meaning there is no shared raw material supply in the process route, the production line isolation termination condition is triggered. At this point, the path extension thread in the current direction is cut off, and the temporary node cache in memory is released. For example, at the boundary between the textile workshop and the machining workshop, since there is neither a consistent production line identifier nor a shared raw material supply relationship, the production line isolation termination condition is directly triggered.

[0071] In this embodiment, the preset category depth threshold is preferably set within the range of 2 to 4, and is specifically obtained by setting the average hierarchical depth of the material classification dictionary in the enterprise manufacturing execution system. Through the above-mentioned rigorous verification in two dimensions, the complex physical connections on the workshop floor plan can be precisely trimmed at the logical level, retaining only the topological links with real business significance, thus providing a noise-free and clean input path for subsequent conflict determination.

[0072] Step S2.444: When the real-time status of the adjacent production line section is the material end consumption point, the material end termination condition is triggered, and the propagation path extension in the corresponding direction is stopped. In this application, the material end-of-life consumption point refers to the final process in the production process where no physical entity is output downstream.

[0073] Furthermore, during the traversal, the real-time status label field of adjacent production line sections is read. If the field contains a termination flag, the material termination condition is triggered, stopping the propagation path extension in the corresponding direction. For example, in the off-line inspection section of the automobile assembly line, the parts have been assembled into finished vehicles and are no longer output to other sections. Its real-time status is marked as the material termination consumption point, and the propagation path stops extending in this section.

[0074] Step S2.445: Define the production line section that triggers the termination condition as the intended cut-off section, and write the boundary position of the intended cut-off section in the workshop plan into the end coordinate slot of the first propagation path.

[0075] In this embodiment, once any termination condition is triggered, the corresponding production line section is defined as the intended cut-off section. The polygonal boundary of the intended cut-off section is extracted from the workshop floor plan, and the horizontal and vertical coordinates of the geometric centroid of the polygon are calculated. .

[0076] Furthermore, this coordinate value is forcibly written into the endpoint coordinate slot of the first propagation path, thereby completing the closed-loop construction of the entire propagation path. Endpoint coordinates The extraction formula is as follows: ; In formula (7), This represents the total number of vertices of the polygon intended to be truncated in the work section. Represents the vertex index. and Representing the first The horizontal and vertical coordinates of each vertex. For example, after triggering the material category termination condition, the plastic injection molding section is defined as the intended cut-off section, its polygonal geometric centroid coordinates are calculated and written into the endpoint coordinate slot, completing the complete path closed loop from the stamping anchor point to the plastic injection molding boundary.

[0077] It should be noted that if the graphic boundary of the section to be cut off has an extremely irregular concave shape, the calculation of the geometric centroid may result in a positional deviation. In this case, it can be replaced by calculating the center point of the circumscribed rectangle of the polygon as the coordinate of the endpoint to ensure that the physical location of the endpoint of the path is accurate.

[0078] Step S3: Retrieve existing form item records with the same name as the spoken title text from the pre-made form item library; locate the field ownership center point and associated endpoint on the workshop floor plan according to the production line identification of each existing form item record; and generate a first fenced area based on the field ownership center point and the associated endpoint. To transform discrete field ownership status into geometric spatial barriers, it is necessary to extract records with the same name from historical data and delineate isolation areas on a two-dimensional plane based on their physical binding relationships. A pre-built form item library serves as a persistent storage medium, preserving the business fields created each time and their production environment context parameters.

[0079] In some embodiments, generating the first fenced area based on the field's attribution center point and the associated endpoint includes: Step S3.1: Based on knowledge graph association query, determine the field belonging center point and the associated endpoint of the existing form item record in the workshop floor plan coordinate system corresponding to the production line to which the field belongs; Specifically, the knowledge graph stores the entity relationships between form items, production line identifiers, and physical locations in the form of triples.

[0080] Furthermore, after retrieving existing form item records with the same name as the spoken title text, the unique primary key of the record is used as the starting entity node, and a breadth-first traversal is performed along the belonging product line relationship edge in the knowledge graph.

[0081] Furthermore, during the traversal, all adjacent nodes are visited starting from the starting node, and then the same visit operation is performed on these adjacent nodes in turn until the target production line identifier node is found.

[0082] When traversing to the production line identifier node, continue expanding outward along the spatial relationship edge until the coordinates of the corresponding field's center point are obtained. and the set of coordinates of the externally associated endpoints of the production line. For example, when searching for an existing record named "material loading", the knowledge graph is used to find that the production line it belongs to is the injection molding line, and then the coordinates of the center point of the injection molding line on the workshop floor plan and the coordinates of the three related endpoints for external material handover are located.

[0083] Step S3.2: In the coordinate system of the workshop floor plan, connect the center point of the field to each of the associated endpoints with a straight line segment to generate the field belonging connection line; In this application, after obtaining the coordinate data, a connection matrix is ​​constructed in the memory of the workshop floor plan coordinate system using the line interpolation drawing function in computer graphics.

[0084] Furthermore, the field will be assigned to a central point. Starting from this point, respectively with the set Each associated endpoint in Perform the line connection operation. Let the connecting line vector be... Its mathematical expression is: ; In formula (8), and Representing the The x and y coordinates of each associated endpoint are used. By performing the above subtraction operation on all endpoints, a set of divergent field-attached connection lines is generated. For example, starting from the center point of the injection molding line, three straight line segments are drawn to the three material junction endpoints, forming a divergent connection structure.

[0085] Step S3.3: When the field belonging connection line satisfies the closed polygon enclosure condition and the associated endpoints are distributed in different quadrants of the field belonging center point, the adjacent associated endpoints are connected end to end according to the outer geometric arrangement order of the associated endpoints on the workshop floor plan to generate a closed first enclosure area as the first fence area. In practical applications, the enclosure condition of a closed polygon refers to the sum of the azimuth spans of the associated endpoints distributed around the center point reaching 360 degrees. First, calculate the enclosure condition for each connecting line. Azimuth angle relative to the positive direction of the horizontal axis The calculation formula is as follows: ; In formula (9), This represents the arctangent function.

[0086] Furthermore, the calculated azimuth angles are arranged in ascending order. If the difference between the maximum and minimum azimuth angles satisfies the closure condition, and the endpoints are distributed in four different quadrants, then adjacent associated endpoints are extracted sequentially according to the azimuth angle arrangement. These endpoints are then connected end-to-end using a linear interpolation function to form a closed polygonal region. This polygonal region is defined as the first enclosed area and written into memory as the first fence region. For example, if the three associated endpoints of an injection molding line are located in the first, third, and fourth quadrants respectively, and the sum of their azimuth angles satisfies the closure condition, then connecting them sequentially generates a triangular enclosed area as the first fence region.

[0087] Step S3.4: When the field belonging connection line does not meet the closed polygon enclosure condition, take the field belonging connection line as the geometric center line, and expand geometrically outward at equal distances to both sides of the center line according to the preset belonging width parameter to generate a first strip area as the first fence area.

[0088] In some embodiments, buffer generation logic is used when endpoints are concentrated in a single quadrant, making closure impossible. For each field-assigned connection line... Calculate its normal vector .

[0089] The vector is Then normalize it and multiply it by the preset home width parameter. Half of it, to obtain the offset vector The calculation formula is as follows: ; In formula (10), This represents the preset home width parameter.

[0090] Furthermore, the offset vector is added to and subtracted from the start and end points of the connecting line, respectively. This yields four outward-expanding vertices, which are then connected sequentially to generate a rectangular strip region.

[0091] Furthermore, all the strip-shaped areas generated by the connecting lines are spatially merged to form the final strip-shaped area as the first fenced area. For example, if all the associated endpoints of an existing record are located in the first quadrant, multiple rectangular strips are generated outward according to a preset attribution width parameter of 2 meters and then merged to form a strip-shaped fence covering the quadrant area. The preset attribution width parameter mentioned here is preferably set within the range of 1 meter to 4 meters, and is specifically obtained based on the average width of the physical isolation zone of the production line in the workshop.

[0092] Step S4: Perform graphic overlay processing on the first propagation path and the first fenced area, and extract the overlapping line segments as the first conflict zone. It should be noted that graphic overlay processing is the core step of performing Boolean operations on spatial features of different dimensions within the same coordinate system. By determining the intersection of the direction of spoken language propagation and the historical field's attribution range in geometric space, potential semantic collision areas can be accurately located.

[0093] In some embodiments, the step of graphically overlaying the first propagation path and the first fenced area to extract overlapping line segments as the first conflict zone includes: Step S4.1: Overlay the first fenced area onto the workshop topology layer where the first propagation path is located, and cut off and extract the overlapping line segments in the first propagation path that completely fall inside the first fenced area through line segment clipping processing, as candidate conflict segments. To accurately extract the path segments that fall inside the fence, a region coding and clipping mathematical mapping relationship is used.

[0094] Furthermore, the boundary of the first fenced area is considered as a clipping window, and each discrete vector line segment of the first propagation path is considered as the object to be clipped. A 4-bit region code is assigned to the endpoints of the line segments, with the most significant bit representing the endpoints being outside the upper, lower, right, and left boundaries of the window, respectively.

[0095] If the ordinate of an endpoint is greater than the upper boundary of the window, the highest value is set to 1, and so on. By performing a bitwise logical AND operation on the region codes of the two endpoints of the line segment, if the result is non-zero, the line segment is completely outside the window and is discarded; if the result is zero and both endpoint codes are zero, the line segment is completely inside the window.

[0096] Furthermore, vector line segments that are completely within the window are cut off and extracted, and stored in a candidate conflict segment set. For example, the first propagation path passes through the first triangular enclosure area generated by the injection molding line. After trimming, the three vector line segments located inside the triangle are extracted as candidate conflict segments.

[0097] Step S4.2: parse the field creation time in the existing form item record, read the preset ownership retention time of the corresponding cross-production line same-name field from the system configuration table, and add the field creation time and the preset ownership retention time together on the time axis to generate the field ownership valid time period; Furthermore, validating the time dimension is crucial to preventing misjudgments caused by historically invalid fields. This involves parsing the creation time field from existing form field records. And read the corresponding preset ownership retention time from the system configuration table. .

[0098] Furthermore, by summing the two values ​​on the timeline, the critical moment of attribution failure can be obtained. The calculation formula is as follows: ; In formula (11), Represented in timestamp format, In seconds. The time is created from the field. To the critical moment of ownership failure The formed left-closed and right-open interval This refers to the effective time period for which a field belongs. For example, if a field in an existing material loading record was created at 8:00:00 on the same day, and the preset retention period is 8 hours, the two are added together to generate the critical moment of 16:00:00 on the same day, which constitutes the 8-hour effective time period.

[0099] Step S4.3: Extract the current oral trigger time, determine whether the current oral trigger time is within the valid time period to which the field belongs, and read the effective activation identifier in the existing form item record; Furthermore, extract the current verbal trigger moment. This is then used as an input variable in the segmented function for logical judgment. The activation / enabling flag is read from existing form field records. Time validity determination function The mathematical expression is as follows: ; In formula (12), the variable This represents the moment when the narration was triggered.

[0100] when When the return value is 1, the determination time is within the valid time period. Simultaneously, the activation flag is read. The boolean value. For example, if the current verbal trigger time is 14, substituting it into the function yields a return value of 1, and the activation flag is read as true, indicating that both the time dimension and the state dimension meet the activation conditions.

[0101] Step S4.4: When the current verbal trigger time is within the valid time period of the field and the activation flag is still enabled, the candidate conflict segment is activated and written into the first conflict area; when the current verbal trigger time exceeds the valid time period of the field or the activation flag is disabled, the candidate conflict segment is deleted.

[0102] Optionally, under extreme boundary conditions, anomalies such as network time synchronization node jumps or out-of-order data packet arrivals may occur, causing non-linear rollback of the timestamp field in existing records of the same batch.

[0103] If a single logical decision is used directly, it is very easy to cause serious distortions such as false activation or missed activation in the conflict zone. Therefore, it is necessary to establish a closed-loop logic for full disaster recovery and defense.

[0104] The function is determined at that moment. Return 1 and When true, the set of candidate conflict segments is... All line segments in the region are activated, and their geometric coordinate sequences are written to the first collision zone memory pool. The activation operation is performed using the following formula: ; In formula (13), This represents the set of line segments that are ultimately written into the first conflict zone. Representative of candidates in the conflict segment. Represents the logical AND operator.

[0105] Conversely, when Return 0 or If the result is false, the corresponding candidate conflict segment is removed from the memory pool.

[0106] It is particularly important to note that when a clock jump causes... much smaller If the time difference exceeds the preset out-of-order tolerance threshold, an abnormal reset command will be triggered, suspending the existing time determination thread and sending a forced time synchronization signal to the clock synchronization service.

[0107] For example, in a situation where severe latency in the workshop network causes data packets to arrive 2 minutes late, the preset out-of-order tolerance threshold is 5 minutes. In this case, a reset is not triggered; instead, the current trigger time is recorded. After adding a delay compensation correction parameter of 120 seconds, the result is recalculated using the function. The preset out-of-order tolerance threshold mentioned here is preferably set within the range of 1 to 10 minutes, and is specifically obtained by calculating based on three times the standard deviation of the average jitter delay of the workshop wireless network.

[0108] Figure 3 The data topology of the judgment process for conflict segment activation and erasure in this embodiment is shown. First, the field creation time and preset attribution retention duration of the existing form item records are parsed to generate a valid field attribution time period; then, a spatiotemporal validity check is performed to determine whether the current verbal trigger time falls within the valid time period and whether the activation flag is still enabled. When the judgment condition is met, the system executes an activation command to write the candidate conflict segment into the first conflict zone memory pool to retain spatial overlap characteristics; when the judgment condition is not met (i.e., the time exceeds or the flag is disabled), the system executes an erasure command to remove the candidate conflict segment from memory, thereby preventing historically failed fields from causing misjudgment interference to the current workshop topology-level pre-rehearsal and ensuring the timing accuracy of conflict zone purification.

[0109] Step S5: Distribute field disambiguation handling strategies based on the segment position of the first conflict zone on the first propagation path.

[0110] It is not difficult to understand that the position of the conflict zone in the propagation path directly reflects the root nature of the semantic collision. By dividing the path into different segments and comparing the position of the conflict zone, it is possible to accurately identify business objects with the same name but different meanings from three dimensions: source, relay, and boundary, thereby triggering differentiated disambiguation handling logic.

[0111] In some embodiments, the step of distributing the field disambiguation handling strategy based on the segment position of the first conflict zone on the first propagation path includes: Step S5.1: The geometric boundary of the production line section symbol where the first oral anchor point is located on the workshop floor plan is taken as the first dividing point, and the geometric boundary of the section symbol intended to be cut off on the workshop floor plan is taken as the second dividing point. It should be noted that the determination of the path split point depends on the extraction of the extreme values ​​of the outer contour of the section symbol. All vertices of the polygon of the section symbol where the first anchor point is located are extracted, the minimum and maximum values ​​of the horizontal and vertical coordinates are calculated, and the circumscribed rectangle is constructed.

[0112] Furthermore, the four sides of the circumscribed rectangle are used as the boundaries of the first dividing point. Similarly, the same extreme value extraction operation is performed on the symbols of the intended truncation sections to construct circumscribed rectangles as the boundaries of the second dividing points. For example, the left side of the circumscribed rectangle of the stamping section is determined as the first dividing point, and the right side of the circumscribed rectangle of the injection molding truncation section is determined as the second dividing point.

[0113] Step S5.2: Based on the first segmentation point and the second segmentation point, the first propagation path is divided into a first source segment, a first relay segment, and a first boundary segment; the first source segment encompasses the physical location of the current production line section of the speaker and the graphic area adjacent to the process pin; the first relay segment is sandwiched between the first source segment and the first boundary segment; the first boundary segment encompasses the graphic area of ​​the section intended to be cut off. Furthermore, the parametric equations of the propagation path are divided into intervals by introducing the coordinates of the dividing points. Let the parametric equation of the first propagation path be... ,in The parameters corresponding to the first dividing point are: The parameters corresponding to the second dividing point are The formula for interval partitioning is as follows: ; In formula (14), The relative position parameters represent the path. Using the piecewise function described above, the first propagation path is strictly divided into three segments. The first source segment covers the work area where the team leader holds the industrial tablet, the first boundary segment covers the material cutting section, and the portion in between is the first relay segment.

[0114] Step S5.3: Perform spatial overlap matching between the first conflict zone and the first source segment, the first relay segment, and the first boundary segment, respectively, and distribute the corresponding field disambiguation handling strategy according to the matching results.

[0115] Furthermore, spatial overlap matching is performed by calculating the intersection-union ratio (IUR) of the conflict zone segment set and the segment sets at each segment level. Let the conflict zone segment set be... The set of segment lines is The formula for calculating the intersection-union ratio is: .

[0116] When the intersection-union ratio (IUR) of a certain interval reaches a maximum value, the field disambiguation processing strategy bound to the segment with the maximum IUR is automatically distributed.

[0117] In some embodiments, the step of distributing the corresponding field disambiguation processing strategy based on the matching result includes: Steps S5.311 to S5.313 are described in conjunction as follows: When the first conflict area overlaps with the first source segment, the writing channel from the form item corresponding to the spoken title text to the work order canvas is cut off; a conflict cause description text is generated based on natural language understanding processing, and the conflict cause description text, together with the production line identifier, field creation time, and field material category recorded in the existing form item, is written into the conflict source data slot of the field interception area; the position to be placed on the work order canvas is flashed and highlighted on the industrial tablet screen, and a warning window pops up, in which the conflict cause description text is displayed; Specifically, when the first conflict zone overlaps with the first source segment, it means that the currently described material object has a direct spatial and semantic collision with an existing historical field at the same physical starting point. At this point, the write channel to the work order canvas must be cut off. The write-off logic is implemented by modifying the write permission flag in memory. Let the write permission flag be... The overlap determination function is: The state transition equation is as follows: ; In formula (15), A value of 0 indicates that the channel is closed, while a value of 1 indicates that the channel is open.

[0118] When the channel is cut off, conflict elements are extracted based on natural language understanding and a preset syntactic template is applied to generate a text describing the cause of the conflict. The template filling process involves extracting the production line identifier, field creation time, and field material category from existing records and injecting them as parameters into a pre-built string template slot to generate a complete natural language description text.

[0119] Furthermore, the generated text, along with the existing record's production line identifier, field creation time, and field's material category, is packaged into a structured data frame and written to the conflict source data slot in the field interception area via memory addressing.

[0120] Furthermore, after the data writing is complete, the graphics rendering process of the industrial tablet is driven. The pixel coordinates of the area to be placed on the work order canvas are read, and the flashing highlight effect is generated by periodically alternating the red and green color channel values ​​of that area.

[0121] At the same time, the operating system's underlying pop-up drawing interface is called to render a warning window.

[0122] Furthermore, the conflict cause description text in the conflict source data slot is extracted and rendered into the text control of the warning window. For example, when the stamping line foreman verbally reports material feeding again and determines that the source overlaps with an existing material feeding record on the stamping line, the channel is cut off, the screen flashes red, and a warning window pops up displaying a conflict prompt: "The current stamping line already has a field named 'Material Feeding,' belonging to the material category 'Steel Coil,' created at 8:00 AM. Please confirm whether to force an addition."

[0123] It should be noted that in certain extreme cases, if the graphics rendering process is unable to draw the warning window in time due to exhaustion of video memory resources, it will automatically degrade to emitting three short alarm sounds via the buzzer built into the industrial tablet, and displaying a persistent notification icon in the terminal's status bar to ensure the delivery of conflict alarm information. The preferred setting range for the blinking high-brightness frequency mentioned here is 1Hz to 4Hz, specifically obtained by setting the optimal stimulation frequency based on the persistence of vision effect of the human eye.

[0124] Furthermore, after completing the entire closed loop of the above-mentioned cut-off, data slot writing, and interface alarm, the field placement action is physically frozen. Operators must make the next decision based on the information provided by the window, effectively blocking the direct overwrite operation of fields with the same name but different meanings at the source position.

[0125] Step S5.314: In response to the team leader's confirmation operation, the production line prefix is ​​added to the form item corresponding to the spoken title text and placed into the work order canvas based on natural language generation processing. Furthermore, if a physical trigger signal is received from the team leader clicking the confirmation button after the warning window pops up, an authorization flag will be written to the flag. Reset to 1.

[0126] Furthermore, before executing the placement and writing, the colloquial address text is reconstructed using natural language processing. The abbreviation of the first production line identifier is extracted as a prefix, and this prefix is ​​combined with the colloquial address text using a string concatenation function. The concatenation formula is as follows: ; In formula (16), This represents the text of the new field after reconstruction. Represents the production line prefix text. Represents the original spoken language terminology text.

[0127] Furthermore, the generated new field text is serialized and written into the component tree data structure of the work order canvas to complete the field placement. For example, in response to the team leader's confirmation trigger, the material loading text is appended with the stamping prefix, generating stamping_loading and falling into the canvas.

[0128] Step S5.315: When the first conflict zone coincides with the first relay segment, search for upstream and downstream process nodes belonging to different production lines along the upstream and downstream of the first conflict zone, pull out a relay verification line across the first fence area in the workshop floor plan memory, and generate a breakpoint crossing verification result based on topological connectivity verification. In this application, overlapping relay segments mean that the spoken language transmission path intersects with the historical field's attribution range when crossing intermediate processes. At this point, it is necessary to search upstream and downstream along the vector direction of the first conflict zone for exclusive process nodes belonging to different production lines.

[0129] Furthermore, the coordinates of these two exclusive process nodes are extracted from memory, and a relay verification line is drawn across the first fenced area using linear interpolation. Based on topological connectivity verification, it is determined whether this verification line penetrates the boundary polygon of the first fenced area. The number of intersections between the line segment and the polygon edges is calculated. The decision logic is as follows: ; In formula (17), This represents the result of the breakpoint crossing verification.

[0130] If the number of intersections is odd, it indicates that the verification line passes from the inside to the outside, and the result is considered topological convergence. If the number is even, it indicates that no penetration or a return after penetration, and the result is considered topological divergence. For example, if there is a historical fence in the buffer zone between the stamping line and the final assembly line, and after pulling out the verification line, it is found that there is one intersection with the fence boundary, the result is considered topological convergence.

[0131] Step S5.316: When the breakpoint crossing verification result is topological convergence, in response to the triggering of multiple operators entering their employee ID and dynamic password to complete the dual electronic signature verification, the material category suffix of the form item corresponding to the spoken title text is marked and placed into the work order canvas based on natural language generation processing.

[0132] In this embodiment of the application, when the topology is determined to be converged, a two-person electronic signature verification process needs to be triggered to prevent data contamination caused by misoperation.

[0133] Furthermore, a signature collection window pops up on the industrial tablet screen, requiring two different operators to enter their respective employee ID strings and dynamic password strings.

[0134] Furthermore, the input string is compared with the hash value in the permissions database. If the comparison passes, the material category identifier corresponding to the colloquial terminology is extracted as a suffix using natural language processing. The suffix concatenation formula is as follows: ; In formula (18), The text representing the final placement field. This represents the suffix text for the material category.

[0135] Furthermore, the spliced ​​text is written into the work order canvas to complete the placement. For example, when the stamping line and the final assembly line have overlapping relay sections and the topology converges, after two operators enter their work ID passwords for verification, the loading text is annotated with the material category suffix, generating "Loading_Steel Coil" and placed into the canvas, thereby accurately distinguishing business objects with the same name but different meanings.

[0136] Example 2, this example also provides a low-code development interface AI intelligent generation system, including: a spoken semantic parsing module, used to obtain the spoken voice package input by the workshop team leader through an industrial tablet, perform natural language semantic parsing on the spoken voice package, extract the spoken title text, and extract the first production line identifier, the first access point identifier and the first process route identifier from the spoken conversation context based on semantic understanding; The propagation path generation module is used to obtain the first verbal anchor point by reverse lookup and positioning in the workshop floor plan coordinate system based on the first production line identifier and the first access point identifier, to determine the first intention landing point by retrieving the production line process layout description based on the first process route identifier, and to generate the first propagation path based on the first verbal anchor point and the first intention landing point. The fence area generation module is used to retrieve existing form item records with the same name as the spoken title text from the pre-made form item library, locate the field ownership center point and associated endpoint on the workshop floor plan according to the production line identification of each existing form item record, and generate the first fence area based on the field ownership center point and the associated endpoint. The conflict zone purification module is used to perform graphic overlay processing on the first propagation path and the first fenced area, and extract the overlapping line segments as the first conflict zone. The disambiguation strategy distribution module is used to distribute field disambiguation handling strategies according to the segment position of the first conflict zone on the first propagation path.

[0137] The above-mentioned unit modules can be embedded in the processor of the electronic device in hardware form or independent of it, or they can be stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of the above modules.

[0138] Figure 4 This paper illustrates the data flow and temporal interaction topology between the modules of the low-code development interface AI intelligent generation system in this embodiment. First, the spoken speech packet input from the industrial tablet is received by the spoken speech parsing module. The spoken address text and multi-dimensional environmental identifiers are extracted through acoustic feature mapping and semantic dependency analysis. Next, the propagation path generation module and the fenced area generation module operate in parallel. The former generates a first propagation path with physical truncation boundaries based on the environmental identifiers and the workshop coordinate system, while the latter constructs a first fenced area by retrieving existing records with the same name from the knowledge graph. Then, the conflict zone purification module performs graphical overlay and time validity verification on both modules to purify the first conflict zone. Finally, the disambiguation strategy distribution module distributes corresponding disambiguation handling instructions such as location interception, dual-person review, or prefix / suffix addition to the industrial tablet based on the overlap of the conflict zone in the path source segment, relay segment, or boundary segment. This completes topology-level pre-visualization and control in multi-production line intertwined scenarios before canvas rendering.

[0139] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0140] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A low-code development interface AI intelligent generation method, characterized in that, include: The system acquires the spoken voice packets input by the workshop team leader via an industrial tablet, performs natural language semantic analysis on the spoken voice packets, extracts the spoken address text, and extracts the first production line identifier, the first access point identifier, and the first process route identifier from the spoken conversation context based on semantic understanding. The first production line identifier is used to describe the production line that the narrator has logged into and bound to; The first access point identifier is used to describe the workshop wireless access point connected to the industrial tablet held by the speaker. The first process route identifier is used to describe the process route to which the work order opened by the speaker in this session belongs; Based on the first production line identifier and the first access point identifier, the first verbal anchor point is obtained by reverse lookup in the coordinate system of the workshop floor plan. The production line process layout description is retrieved based on the first process route identifier, and the process node responsible for receiving materials is determined from the production line process layout description as the first intention landing point. A first propagation path is generated based on the first spoken anchor point and the first intention landing point; Retrieve existing form item records with the same name as the spoken title text from the pre-made form item library, locate the field ownership center point and associated endpoint on the workshop floor plan according to the production line identification of each existing form item record, and generate a first fenced area based on the field ownership center point and the associated endpoint. The first propagation path and the first fenced area are graphically overlaid, and the overlapping line segments are extracted as the first conflict zone. Based on the segment position of the first conflict zone on the first propagation path, a field disambiguation handling strategy is distributed.

2. The low-code development interface AI intelligent generation method according to claim 1, characterized in that: The step of performing natural language semantic analysis on the spoken voice package to extract the spoken address text includes: The spoken voice packet is converted into a text sequence, entity recognition is performed on the text sequence, and business object descriptive words are extracted as the spoken title text; The text sequence is classified and identified to determine that the business action corresponding to the colloquial address text is a new field. Based on semantic dependency analysis, production line environment descriptors, network environment descriptors, and process environment descriptors are extracted from the context of the text sequence and mapped to the first production line identifier, the first access point identifier, and the first process route identifier, respectively.

3. The low-code development interface AI intelligent generation method according to claim 2, characterized in that: The generation of the first propagation path based on the first spoken anchor point and the first intention landing point includes: Retrieve the corresponding production line process layout description based on the first process route identifier; The production line process layout description includes embedded relative coordinate references of the workshop plane and material connection relationships for each process node from the raw material warehouse interface point to the finished product off-line station; Based on semantic similarity comparison, the first verbal anchor point is mapped to the coordinates of the process node in the production line process layout description and defined as a verbal process point. The coordinates of the process node closest to the raw material warehouse interface point and responsible for receiving materials in the production line process layout description are defined as the first intention landing point. In the coordinate system of the workshop floor plan, a vector connection is made with the verbal process point as the starting point and the first intention landing point as the ending point to generate the first propagation path with directionality.

4. The low-code development interface AI intelligent generation method according to claim 3, characterized in that: The generation of the first propagation path with directionality includes: The material connection lines on the workshop floor plan are traversed along the direction of the first propagation path. When the traversal reaches the graphic boundary of the adjacent production line section, the real-time material occupancy status of the adjacent production line section is read through the real-time process status database. Based on semantic matching, the real-time material occupancy status of the adjacent production line section is matched with the material injection direction corresponding to the spoken term text to establish a connection relationship. When a succession relationship is determined, the identifier and graphic outline of the adjacent production line section are connected to the first propagation path, and the next propagation path is regenerated from the process boundary on the other side of the adjacent production line section away from the first verbal anchor point. When it is determined that there is no connection, the extension of the first propagation path in the corresponding direction is terminated.

5. The low-code development interface AI intelligent generation method according to claim 4, characterized in that: The step of terminating the extension of the first propagation path in the corresponding direction when it is determined that no succession relationship exists includes: When traversing to an adjacent production line section, the production line identifier and material category identifier of the adjacent production line section are retrieved simultaneously. When the material category received by the adjacent production line section is not the same as the material category of the production line corresponding to the first production line identifier, the material category termination condition is triggered, and the propagation path extension in the corresponding direction is stopped. When the production line identifier of the adjacent production line section is inconsistent with the first production line identifier and there is no process route association with shared raw material supply, the production line isolation termination condition is triggered, and the propagation path extension in the corresponding direction is stopped. When the real-time status of the adjacent production line section is the material end consumption point, the material end termination condition is triggered, and the propagation path extension in the corresponding direction is stopped. The production line section that triggers the termination condition is defined as the intended cut-off section, and the boundary position of the intended cut-off section in the workshop plan is written into the end coordinate slot of the first propagation path.

6. The low-code development interface AI intelligent generation method according to claim 1, characterized in that: The process of generating a first fenced area based on the field's home center point and the associated endpoint includes: Based on knowledge graph association query, determine the field belonging center point and the associated endpoint of the existing form item record in the workshop floor plan coordinate system corresponding to the production line to which the field belongs; In the coordinate system of the workshop floor plan, the field ownership center point is connected to each of the associated endpoints by a straight line segment to generate the field ownership connection line; When the field belonging connection line satisfies the closed polygon enclosure condition and the associated endpoints are distributed in different quadrants of the field belonging center point, the adjacent associated endpoints are connected end to end according to the outer geometric arrangement order of the associated endpoints on the workshop floor plan to generate a closed first enclosure area as the first fence area. When the field belonging connection line does not meet the closed polygon enclosure condition, the field belonging connection line is used as the geometric center line, and the field is geometrically expanded outward at equal distances to both sides of the center line according to the preset belonging width parameter to generate a first strip area as the first fence area.

7. The low-code development interface AI intelligent generation method according to claim 1, characterized in that: The step of overlaying the first propagation path and the first fenced area into a graphic representation and extracting the overlapping line segments as the first conflict zone includes: The first fenced area is overlaid on the workshop topology layer where the first propagation path is located. The overlapping line segments in the first propagation path that fall completely inside the first fenced area are cut off and extracted as candidate conflict segments through line segment clipping. Parse the field creation time in the existing form item record, read the preset ownership retention time of the corresponding cross-production line same-name field from the system configuration table, and accumulate the field creation time and the preset ownership retention time on the time axis to generate the field ownership valid time period; Extract the current oral trigger time, determine whether the current oral trigger time is within the valid time period to which the field belongs, and read the effective activation identifier from the existing form item record; When the current verbal trigger time is within the valid time period of the field and the activation flag is still enabled, the candidate conflict segment is activated and written into the first conflict area; when the current verbal trigger time exceeds the valid time period of the field or the activation flag is disabled, the candidate conflict segment is erased.

8. The low-code development interface AI intelligent generation method according to claim 7, characterized in that: The step of distributing field disambiguation handling strategies based on the segment position of the first conflict zone on the first propagation path includes: The first dividing point is the geometric boundary of the production line section symbol where the first verbal anchor point is located on the workshop floor plan, and the second dividing point is the geometric boundary of the section symbol intended to be cut off on the workshop floor plan. Based on the first segmentation point and the second segmentation point, the first propagation path is divided into a first source segment, a first relay segment, and a first boundary segment; The first source segment encompasses the physical location of the current production line section of the narrator and the graphic area adjacent to the process pin; The first relay segment is sandwiched between the first source segment and the first boundary segment; The first boundary segment encloses the graphic area of ​​the section intended to be cut off. The first conflict zone is spatially matched with the first source segment, the first relay segment, and the first boundary segment, and the corresponding field disambiguation handling strategy is distributed according to the matching results.

9. The low-code development interface AI intelligent generation method according to claim 8, characterized in that: The method of distributing the corresponding field disambiguation processing strategy based on the matching result includes: When the first conflict zone overlaps with the first source segment, the writing channel from the form item corresponding to the spoken title text to the work order canvas is cut off. The conflict cause description text is generated based on natural language understanding processing. The conflict cause description text, along with the production line identifier, field creation time, and field material category of the existing form item record, is written into the conflict source data slot of the field interception area. The position to be placed on the work order canvas is flashed and highlighted on the industrial tablet screen, and a warning window pops up, displaying the conflict cause description text. In response to the team leader's confirmation operation, the production line prefix is ​​added to the form item corresponding to the colloquial title text based on natural language generation processing and placed into the work order canvas. When the first conflict zone coincides with the first relay segment, upstream and downstream process nodes belonging to different production lines are searched along the upstream and downstream of the first conflict zone. A relay verification line is pulled across the first fence area in the workshop floor plan memory. A breakpoint crossing verification result is generated based on topological connectivity verification. When the breakpoint crossing verification result is topological convergence, in response to the triggering of multiple operators entering their work numbers and dynamic passwords to complete the dual electronic signature verification, the material category suffix of the spoken title text is marked on the form item based on natural language generation processing and placed into the work order canvas.

10. A low-code development interface AI intelligent generation system, characterized in that, include: The oral speech semantic parsing module is used to acquire the oral speech package input by the workshop team leader through an industrial tablet, perform natural language semantic parsing on the oral speech package, extract the oral address text, and extract the first production line identifier, the first access point identifier, and the first process route identifier from the oral conversation context based on semantic understanding. The propagation path generation module is used to obtain the first verbal anchor point by reverse lookup and positioning in the workshop floor plan coordinate system based on the first production line identifier and the first access point identifier, to determine the first intention landing point by retrieving the production line process layout description based on the first process route identifier, and to generate the first propagation path based on the first verbal anchor point and the first intention landing point. The fence area generation module is used to retrieve existing form item records with the same name as the spoken title text from the pre-made form item library, locate the field ownership center point and associated endpoint on the workshop floor plan according to the production line identification of each existing form item record, and generate the first fence area based on the field ownership center point and the associated endpoint. The conflict zone purification module is used to perform graphic overlay processing on the first propagation path and the first fenced area, and extract the overlapping line segments as the first conflict zone. The disambiguation strategy distribution module is used to distribute field disambiguation handling strategies according to the segment position of the first conflict zone on the first propagation path.

Images