Hydraulic operating mechanism stroke switch intelligent adjustment and measurement method

Abstract

A kind of hydraulic operating mechanism stroke switch intelligent adjustment method, comprising the following steps: S1: obtaining channel mapping table and collection hardware list, generate measurement feature set;S2: based on model mapping table, equipment archives and the measurement feature set, generate comparison reference set and node deviation list;S3: obtain the node deviation list and the measurement feature set, generate suggestion instruction set;S4: based on the suggestion instruction set and the comparison reference set, generate strategy update entry structure containing new rule suggestion and reference version suggestion.A kind of hydraulic operating mechanism stroke switch intelligent adjustment method, the method realizes the automation, precision and intelligent of stroke switch adjustment process;Significantly improve test efficiency and reliability, while continuously optimizing system performance through closed-loop feedback mechanism.

Description

Technical Field

[0001] This invention relates to the field of intelligent adjustment and testing technology of limit switches, specifically to an intelligent adjustment and testing method for limit switches in hydraulic operating mechanisms. Background Technology

[0002] In the field of program control and data processing of hydraulic operating mechanisms, limit switches, as key state transition and position monitoring components, directly affect the reliability and timing accuracy of the hydraulic system's operation through accurate commissioning and performance verification. Currently, the commissioning schemes for limit switches generally adopt a semi-automated or manual operation mode dominated by human experience. A typical process includes: manually acquiring sensor data and aligning multi-source signals in the time domain; comparing and analyzing the data against preset standard threshold values ​​and historical trend curves; comparing the position logic and timing logic of the switch action based on manual interpretation results; and finally, verifying the execution process, confirming the results, and archiving the test records on-site.

[0003] However, the existing solutions have inherent limitations in terms of data processing flow consistency and system integration. Specifically: First, the acquisition of synchronous data frames is highly dependent on on-site personnel configuring the acquisition hardware parameters. The loose coupling between the comparison benchmark set and the benchmark index table makes it difficult for the generated suggested instruction set to maintain strict consistency with the execution linkage mapping table. Second, existing methods often load the channel mapping table and acquisition hardware list through scattered engineering steps and local scripts, while channel drift correction and time benchmark configuration still require manual intervention. After completing the above configuration, operators then perform feature retrieval based on historical curves and make a comprehensive judgment based on the standard threshold loading results. This workflow is prone to problems such as inconsistent node event extraction standards and blurred boundaries of working condition segment division when performing linkage operations and non-steady-state conditions such as slow separation and slow merging, making it difficult to meet the application requirements of intelligent and highly consistent adjustment of limit switches.

[0004] Furthermore, regarding the complete joint processing chain from channel mapping table, acquisition configuration set, synchronization data frame, measurement feature set, comparison benchmark set to benchmark index table, existing technologies exhibit significant functional fragmentation and inefficient data write-back at several key stages. Specifically, there is a lack of unified data models and process-oriented connection mechanisms between stages such as index mapping and tag binding, automatic generation of node deviation lists, inference application of rule base and strategy base, model adaptation and execution mapping, execution linkage and slow separation and slow merging process verification, node-level verification and secondary comparison, and automatic generation of test reports and structured extraction of knowledge items.

[0005] Therefore, in on-site commissioning and verification scenarios, existing solutions struggle to establish a closed-loop, continuous process from "data acquisition—alignment—judgment—execution—verification—archiving—strategy update." The consequences include: the commissioning process is susceptible to individual operator differences, resulting in insufficient stability; the identification and response to abnormal operating conditions lack timeliness; and due to the lack of structured process data, strategy update items are difficult to accumulate and reuse effectively, thus hindering the improvement of efficiency and reliability in the commissioning of hydraulic operating mechanism limit switches. Summary of the Invention:

[0006] This invention provides an intelligent adjustment and testing method for limit switches in hydraulic operating mechanisms. This method automates, refines, and intelligentizes the limit switch adjustment and testing process, significantly improving testing efficiency and reliability, while continuously optimizing system performance through a closed-loop feedback mechanism.

[0007] The technical solution adopted in this invention is as follows: A method for intelligent adjustment and testing of limit switches in a hydraulic operating mechanism includes the following steps: S1: Obtain the channel mapping table and acquisition hardware list, and generate the measurement feature set; S2: Based on the model mapping table, equipment files and the measurement feature set, generate a comparison benchmark set and a node deviation list; S3: Obtain the node deviation list and the measurement feature set, and generate a suggested instruction set; S4: Based on the suggested instruction set and the comparison benchmark set, generate a strategy update entry structure that includes suggestions for new rules and suggestions for reference versions.

[0008] In S1: the channel mapping table is used to describe the wiring relationship between the hydraulic operating mechanism and the limit switch status, hydraulic pressure, displacement stroke, opening and closing sequence, and environmental quantities, as well as the terminal number, range, and signal type; and establishes a one-to-one correspondence of identification items through the model dimension and the terminal dimension.

[0009]

[0010] Table 1 establishes a unique mapping relationship between physical wiring and software channels by using "model identifier" and "terminal number" as joint primary keys.

[0011] The hardware list includes the host model of the field acquisition unit, acquisition board specifications, number of channels and channel resolution, supported input types, calibration certificate number, firmware version, and synchronization bus type, as shown in Table 2:

[0012] In S1: Perform time base configuration, channel drift correction, sensor data acquisition, multi-source alignment, node event extraction, and working condition segmentation processing to generate a measurement feature set containing a node event table and a working condition segment table; The time base configuration includes setting a unified start time and sampling interval; specifically as follows: Let the globally unified timeline be... , in: To standardize the start time (unit: seconds) The sampling interval is specified in seconds. The master clock source time is calculated using a master-slave synchronization strategy. With the device's local time offset And will correct from device time to This ensures that all acquisition channels are started on the same time reference and that the time interval between adjacent sampling points is strictly equal, providing a unified time-scale basis for subsequent multi-source alignment.

[0013] The channel drift correction includes measuring zero-point offset and gain offset; specifically as follows: Using a linear correction model , ; in: (The input is a real physical quantity, such as voltage or pressure). These are the raw values ​​directly read by the acquisition device. This is the gain offset correction factor (dimensionless). This is the zero-point offset correction factor, in units of 1 / 2π. same.

[0014] Determine the zero-point offset by inputting a zero-value standard signal under static closed-loop or known steady-state operating conditions. Calculate the gain offset by inputting a full-scale standard signal. .

[0015] For channels exhibiting temperature drift, temperature drift compensation is introduced: , ; in: The current temperature. Reference temperature Gain coefficient under, Temperature drift compensation factor, unit: 1 / ℃ For reference temperature, 25℃ is typically used; the drift sensitivity level is also indicated. 0 indicates insensitive, and 2 indicates highly sensitive, used to dynamically adjust the correction frequency.

[0016] The sensor data acquisition includes the acquisition of key quantities such as the on / off state of the start limit switch and hydraulic pressure. The multi-source alignment refers to generating alignment points through interpolation resampling under a unified time reference; specifically as follows: Under a unified time reference, a high-frequency channel is set up. In the global timeline There are already sampled values ​​above. Low-frequency channel The original sampling time is The corresponding sample value is .

[0017] The low-frequency channel at the target alignment time is calculated using a linear interpolation formula. The estimated value: ; in: The alignment point value after interpolation; The slope between two points is expressed as the rate of change. For incomplete segments, such as consecutive missing samples exceeding 30% of the total sampling points, they are directly discarded and a missing flag is written, thereby generating a synchronous data frame in which all channels have complete samples at the same time.

[0018] The node event extraction refers to locating edge nodes on the on / off trajectory; specifically as follows: The limit switch on / off signal channel is a discrete switch signal, and its edge node is defined as the moment when the signal jumps from "0" to "1" or from "1" to "0".

[0019] Let the on / off signal be At the sampling time and Between, if If so, it is determined that an edge node exists, and the time when the node appears is taken as... This node is labeled as a "limit switch trigger node".

[0020] At the same time, combined with the rising edge (starting moment) of the coil current path It can identify the "action start node".

[0021] All nodes are accompanied by a quality label. In this context, 0 indicates low quality (requires manual review), 1 indicates average quality, and 2 indicates high quality (confirmed by multiple sources).

[0022] The operating condition segmentation process divides the circuit breaker tripping and closing execution segments using node events as anchor points. Specifically: Using key nodes from the extracted node events as anchor points, the working condition segments are segmented.

[0023] Define the action start node as The limit switch trigger node is The action termination node is .but: Preparing to trip the circuit breaker: ; Tripping execution segment: ; Tripping stabilization segment: ; The closing process is similar.

[0024] in: This indicates the end time of the previous action or the start time of data acquisition. Each segment is accompanied by a model identifier. Action direction indicator and operation method labels This is used for subsequent differentiation processing.

[0025] In S1, the process of generating the measurement feature set, which includes the node event table and the working condition segment table, also includes: a1: Obtain the channel mapping table and acquisition hardware list, perform time base configuration and channel drift correction processing, and obtain the acquisition configuration set; a2: Extract channel identifiers and timing identifiers from the acquisition configuration set, perform sensor data acquisition and multi-source alignment processing, and generate synchronous data frames; a3: Perform node event extraction and condition segmentation processing on the synchronous data frame to generate a measurement feature set.

[0026] The time reference configuration includes: setting a unified start time, sampling interval, and trigger priority, and establishing a single time reference through a master-slave synchronization strategy. Specifically: The master-slave synchronization strategy is implemented through a high-precision synchronization bus (such as PXIe Sync). The master clock source (such as a GPS timing module) broadcasts a synchronization pulse, and the slave devices (acquisition board 1, acquisition board 2, etc.) latch their local counters after receiving the pulse and calculate the offset. And adjust your local time.

[0027] Trigger priority is divided into three levels: Priority 1 is the on / off state of the limit switch and hydraulic pressure (critical quantity). Priority 2 is based on coil current and displacement stroke. Priority 3 is environmental quantities and vibration.

[0028] Set a unified start time The sampling interval is the first whole second after the arrival of the synchronization pulse. The unified configuration is 1ms or 0.1ms (depending on the signal bandwidth). This mechanism ensures the initial phase consistency of data acquired from multiple boards, laying the foundation for subsequent alignment.

[0029] The channel drift correction includes: measuring zero-point offset and gain offset, and marking the drift sensitivity level for variable-temperature drift channels. Specifically: In the aforementioned linear correction model Based on this, add a drift sensitivity level marker.

[0030] Define drift sensitivity level Historical drift rate Decide: like but ; like but ; like but .

[0031] in, and Temperature and Gain correction factor.

[0032] for The channel automatically performs dynamic zero-point calibration every 10 seconds; for The channel is calibrated every 60 seconds; The channel is calibrated only once before each test begins.

[0033] The sensor data acquisition includes: prioritizing the acquisition of high-frequency channels for key quantities such as limit switch on / off states and hydraulic pressure.

[0034] The multi-source alignment includes: generating alignment points according to the target sampling time under a unified time reference, and organizing the channel samples at the same time by interpolation resampling and discarding incomplete segments. Specifically: The original sampling times of the high-frequency channel (sampling rate 10kHz) and the low-frequency channel (sampling rate 1kHz) are different. A linear interpolation formula (same as above) is used to resample the low-frequency channel onto the time axis of the high-frequency channel. Simultaneously, a complete segment discrimination condition is defined: Let the total number of sampling points within the segment be... The number of missing markers is ,like (,in: A threshold value, typically 0.3, is used to classify a segment as incomplete and discard it. Otherwise, missing segments are filled using interpolation. Discarded segments trigger an exception log, are written to the collection log, and have their weight reduced in subsequent node event extractions.

[0035] The node event extraction includes: identifying the action start node and the limit switch trigger node, and locating the edge node on the on / off quantity track. Specifically: The identification of the start point of the action is based on the rising edge of the coil current path: Let the coil current signal be... Its first difference ,when ( A preset current change threshold, typically 0.5A, is used, and the duration exceeds two sampling points before a determination is made. This is the starting node for the action. The identification of the limit switch trigger node is based on the edge of the on / off channel: when... At that time, the judgment For trigger nodes, for multi-channel redundancy (such as two independent limit switches in series), an "OR" logic is used: the occurrence of an edge on any channel triggers a node event, and the numbers of all participating channels are recorded in the node table.

[0036] The operational segment segmentation includes: using the results of node event extraction as anchor points, dividing the circuit breaker preparation segment into a circuit breaker closing execution segment. Specifically: Define the general segmentation rules for action chains: Assume a complete action cycle includes: preparation phase Execution phase Stable phase .in: To control the timing of command issuance, whether from the PLC or HMI, The trigger time of the limit switch. This refers to the moment when the hydraulic pressure recovers to 95% of its steady-state range. The tripping preparation segment corresponds to this. The closing execution segment corresponds to Fragment length Used for subsequent deviation density calculation. Each fragment is written to the database with the following fields: fragment ID, action direction, start and end node index, and quality level, which are derived from the internal node quality weighted average.

[0037] In S1: the model mapping table is used to give the unified identification relationship of the hydraulic operating mechanism in different manufacturers, different specifications and different wiring terminals, including model identification, terminal number, measuring point name, signal type, range, action direction identification and adaptation status. The model mapping table is used to unify the identification relationships of hydraulic operating mechanisms from different manufacturers and with different specifications, ensuring the correctness of subsequent model adaptation and execution mapping. As shown in Table 3:

[0038] Among them, "Model Identifier" is the unique code of the device, "Terminal Number" corresponds to the actual terminal block position, "Measurement Point Name" is the physical quantity description, "Signal Standard" indicates the electrical type, "Measurement Range" is the sensor's measurement range, "Action Direction Identifier" distinguishes between opening / closing / general, and "Adaptation Status" indicates whether the mapping is currently valid. Table 3 provides a query basis for model adaptation in the suggested instruction set through the joint primary key of the Model Identifier and Terminal Number.

[0039] Equipment files provide basic information, standard threshold version records, historical curve storage location information, and operating environment descriptions for the same device during various maintenance and testing processes. They cover maintenance dates, executors, operating condition descriptions, device serial numbers, hardware change records, and data acquisition paths.

[0040] Equipment files record historical information about the same device during various maintenance and testing processes, used for standard threshold version selection and historical curve retrieval. See Table 4:

[0041] Among them, the "device serial number" is a unique identifier, the "standard threshold version" determines which set of criteria is loaded, the "historical curve ID" is used to retrieve the corresponding qualified curve from the storage system as a comparison benchmark, the "hardware change record" helps determine whether recalibration is needed, and the "data acquisition path" indicates the original data storage location. In step S210, the curve with the most recent acquisition time and marked as qualified is selected as the control sample, and the corresponding version of the criteria is loaded according to the standard threshold version.

[0042] In S2, standard threshold loading, historical curve retrieval, index mapping, label binding, position comparison and time series comparison processing are performed to generate a comparison benchmark set and a list of node deviations.

[0043] In S2, the process of generating the comparison benchmark set and the list of node deviations also includes: b1: Obtain the model mapping table and equipment file, perform standard threshold loading and historical curve retrieval processing, and obtain the comparison benchmark set; b2: Extract the standard node positions and standard time series from the benchmark set, perform index mapping and label binding processing, and generate a benchmark index table; b3: Perform position comparison and time series comparison processing on the measurement feature set and the benchmark index table to generate a list of node deviations.

[0044] The standard threshold loading includes: loading the limit switch on / off threshold and the hydraulic pressure reference threshold, and selecting the corresponding version according to the standard threshold version. Specifically: Standard threshold loading refers to loading the corresponding set of criteria from the standard library based on the model identifier in the model mapping table and the standard threshold version in the equipment file.

[0045] The standard threshold table includes the on / off threshold of limit switches, such as the closing position of the switch should be closed within the range of 98% to 102% of the displacement stroke, and the hydraulic pressure reference threshold, such as the pressure before opening should not be lower than 28 MPa, etc.

[0046] Let the standard threshold set for loading be... , in: For the first The threshold range of a limit switch, the lower limit upper limit ), For the first Reference threshold for each pressure measurement point This represents the threshold range for displacement travel. Each threshold is accompanied by a version number. The system automatically selects the corresponding version based on the version recorded in the device file (e.g., V2.1). Examples are shown in Table 5:

[0047] The historical curve retrieval includes retrieving opening and closing process curves, and prioritizing the curves with the most recent acquisition time and marked as qualified as the control sample. Specifically: Historical curve retrieval refers to retrieving qualified process curves of the same device under the same or similar operating conditions from equipment records as a reference sample.

[0048] The system displays three historical closing curves: Curve A (March 10, 2025, qualified), Curve B (February 18, 2024, qualified), and Curve C (November 20, 2024, unqualified).

[0049] Curve A, which was collected most recently and marked as qualified, was selected as the control sample.

[0050] Let the set of historical curves be... Each curve Includes timestamp Quality Label Operating conditions description .

[0051] The search rule is: Select That is, the one with the largest timestamp in the qualified curve.

[0052] If the search is unsuccessful, an alternative rule is triggered: select a representative curve with a higher quality level within the same model range, along with an alternative explanation.

[0053] The standard node positions and standard timing include: the candidate node locations of the action initiation node and the limit switch trigger node on the historical curve entries. Specifically: Standard node location and standard timing refer to the spatiotemporal coordinates of nodes extracted from historical curve entries in the comparison benchmark set and verified by standard thresholds. On a qualified closing curve, the system automatically locates the "action start node". (Rising edge of coil current), "Limit switch trigger node" (Switchover signal), "Pressure inflection point" (The pressure derivative crosses zero) etc.

[0054] Let the set of standard node locations be ,in, , The standard time (relative to the start of the action). For node type.

[0055] Standard timing is defined as the interval between adjacent nodes: . This represents the standard interval from the start of the action to the triggering of the switch. These standard nodes and timing sequences are written into a benchmark index table for subsequent comparison.

[0056] The tag binding includes binding an action direction tag and a quality level tag, and attaching a source tag and an adaptation tag. Specifically: Tag binding refers to attaching multiple tags to each index entry in the base index table to support multi-dimensional retrieval and prioritization decisions. Each node index entry includes: action direction tag. Quality grade label (2 is the highest), Source Tag Adaptive tags For example, the tag combination of a closing trigger node can be: After the tag is bound, the comparison and judgment unit can determine the comparison based on... Prioritize high-reliability nodes, based on Filter out index items that match the current operation method to avoid false matches.

[0057] The position comparison is used to describe the deviation of the actual node from the reference range and to compare the occurrence time with the channel context. Specifically: Position comparison describes the deviation between the actual node's position in terms of displacement stroke, hydraulic pressure, and other physical quantities and the standard reference range. The actual closing trigger node's position on the displacement curve is... The standard reference range is given by the benchmark index table. The reference median is .

[0058] The formula for calculating position deviation is: ; in: Position deviation, in millimeters (mm). A positive value indicates that the actual node position is too large (lagging), and a negative value indicates that it is too small (leading).

[0059] : The displacement value corresponding to the moment when the actual node appears on the displacement travel curve.

[0060] : The median of the standard reference range.

[0061] At the same time, compare with the channel context (such as pressure, current): If Exceeding the allowable deviation If the deviation is ±1mm, the node is marked as having an abnormal position, and the deviation amount and associated channel identifier are recorded. Actual node position. Located to the right of the reference range, If it is within the allowable range, it is considered acceptable.

[0062] The time-series comparison is used to describe the difference between the actual interval and the standard time sequence, and to perform neighborhood reinforcement on the actual nodes identified as low-quality. Specifically: Temporal comparison is used to describe the difference between the time interval between actual adjacent nodes and the standard temporal reference interval, and to perform neighborhood reinforcement on actual nodes with low-quality identification. Let the adjacent nodes in the actual node sequence be... and The timestamps are respectively and The actual interval is: ; The standard timing reference interval is: (Originally obtained from the baseline index table), the timing deviation is: ; in: : The actual time interval between adjacent nodes, in seconds (s).

[0063] : No. The and the first The time when each actual node appears.

[0064] Standard timing reference interval, derived from qualified historical curves.

[0065] Timing deviation: a positive value indicates that the actual interval is too large (the action is slower), and a negative value indicates that it is too small (the action is faster).

[0066] For low-quality labels ( For the actual node, perform neighborhood reinforcement: Take the nodes before and after this node sampling points ( ) constitutes a neighborhood window Within this window, the node position is re-estimated using other high-quality channels (such as the first derivative of pressure and peak current). If the estimated result deviates from the original node position by less than a threshold (e.g., 2ms), the corrected time is used. Replace the original Otherwise, retain the original node and add a low-quality warning. Low-quality node original interval deviation. After neighborhood reinforcement, it is corrected to It has entered an acceptable range.

[0067] In S3: abnormal fragment identification, working condition aggregation, rule base and strategy base reasoning, model adaptation and execution mapping processing are performed to generate a suggested instruction set.

[0068] The process of generating the suggested instruction set also includes: c1: Obtain the node deviation list and measurement feature set, perform abnormal segment identification and working condition aggregation processing to obtain the working condition context; c2: Extract node type and deviation direction from the node deviation list and operating condition context, perform rule base and strategy base reasoning processing, and generate a set of suggested candidates; c3: Performs model adaptation and execution mapping processing on the proposed candidate set to generate a proposed instruction set.

[0069] The abnormal segment identification includes: calculating the deviation density and verifying the consistency of the operating condition segment table. Specifically, abnormal segment identification first verifies the consistency of the operating condition segment table, checking for duplicate time markers, unreasonable spans, sudden drops in channel coverage, etc., and then calculates the deviation density of each segment based on the node deviation list. Let a certain operating condition segment... The duration is (Unit: seconds), the number of node offset entries contained in this segment is Then the deviation density is defined as: ; in: : fragment Deviation density, unit: particles / second.

[0070] : belongs to fragment The total number of entries in the node deviation list.

[0071] The duration of the segment is calculated from the time difference between the start and end points. .

[0072] like ( If a preset threshold is used (usually 5 segments / second), the segment is marked as an abnormal segment. The continuity check also checks for temporal reversal between the start and end nodes. The system calculates the concentration of missing markers (proportion of missing points > 30%), and if found, marks it as a candidate for suspicious segment. The abnormal segment identification unit outputs a set of abnormal segments. Each anomalous segment is accompanied by the dominant deviation direction. (Position ahead / behind, timing faster / slower) and combinations of key node types involved.

[0073] The operational condition aggregation includes: using the action direction identifier and operation mode identifier as the primary keys, grouping the abnormal segment set, node deviation list, and relevant entries in the node event table into the corresponding topics. Specifically: Working conditions are aggregated and identified by the direction of action. and operation method identification As a composite primary key, the set of abnormal fragments Node Deviation List and node event table The relevant entries are categorized into corresponding topics. The aggregation process generates a working condition context structure, and each topic contains: topic ID (e.g., "closing-slow closing"), a fragment list (arranged in chronological order), a summary of the dominant deviation, and a duplicate identifier. Table 6 shows an example of the aggregated working condition topics.

[0074]

[0075] Among them, the "repetition identifier" is used to mark the frequency of the abnormal pattern in the historical session. If it occurs more than 3 times, it is labeled as "high frequency repetition" to prompt subsequent strategy reasoning to pay priority.

[0076] The rule base is based on a structured rule set, and its rule antecedents consist of a combination of node types, the dominant deviation direction, and the fragment category. Specifically: The rule base is a knowledge base based on a set of structured rules. Each rule's antecedent consists of a combination of node types, the dominant deviation direction, and the fragment category, while the consequent is an executable adjustment suggestion primitive. Let the rule... The form is: ; in: Node type combination, such as {closing trigger node, pressure inflection point node}.

[0077] : A specific set of node types.

[0078] : Dominant deviation direction, value .

[0079] Fragment category, value .

[0080] Adjustment suggestions, such as "adjust the limit switch slightly in the closing direction by 0.5mm".

[0081] Specific action descriptions are shown in Table 7.

[0082] The rule base achieves rapid decision-making through antecedent matching, and directly outputs the corresponding consequent when there is a complete match.

[0083] The strategy library is an experience-based strategy library that generates candidate adjustment patterns based on similar topics and duplicate identifiers. Specifically: The strategy library is a collection of strategies based on historical experience. When there are no perfectly matching entries in the rule library, the strategy library generates candidate adjustment patterns based on similar topics and duplicate identifiers. Each strategy in the strategy library... Includes similarity measurement function And migration rules.

[0084] Let the current operating condition be described as follows: The historical strategy entry is Similarity calculation formula: ; in: : Indicator function, takes the value 1 if the condition is true, otherwise takes the value 0.

[0085] Weighting coefficient, usually taken as .

[0086] : The current set of abnormal fragments.

[0087] : A collection of anomalous fragments recorded in the historical strategy.

[0088] The historical strategy with the highest similarity is selected, and its adjustment pattern is transferred to the current scenario to generate candidate adjustment patterns. If a strategy is repeated, it is marked as "high-frequency repetition," and its priority is increased. The strategy base inference outputs a candidate set, with each candidate accompanied by a similarity score and supporting evidence.

[0089] The model adaptation includes mapping abstract objects to specific wiring terminals, and providing corresponding contact pairs and fixing methods according to the terminal mapping specifications. Specifically: Model matching maps abstract adjustment objects (such as "limit switch position adjustment") in the suggested candidate set to the specific wiring terminals, sensing channels and mechanical adjustment positions of the current device.

[0090] Let the abstract object in the suggested candidate be... The model mapping table contains a mapping function: ; in: The model identifier of the current device (e.g., “CY-H-01”).

[0091] : Specific terminal block number (e.g., "X1:1-2").

[0092] Fixing method (e.g., "M4 screw, torque 2N·m").

[0093] Adjust the step description (e.g., "rotating 1 / 4 turn corresponds to 0.2mm displacement").

[0094]

[0095] The model adaptation unit automatically performs mapping based on the model identifier of the current device. If the mapping is missing, the candidate is suspended and a prompt to update the model mapping table is displayed.

[0096] The execution mapping includes mapping to PLC automatic trigger instructions or HMI interactive prompt instructions, and splitting a candidate into a group of ordered execution entries. Specifically: Execution mapping translates the specific adjustment actions after model adaptation into instruction entries recognizable by the execution layer. This includes automatic trigger instructions for PLCs and interactive prompt instructions for HMIs, and breaks down a candidate into a set of ordered execution entries. Let's say a candidate... The specific action sequence is obtained after model adaptation. The execution mapping unit transforms it into: For automatically triggered types: Generate PLC instructions. .

[0097] For interactive prompt types: Generate HMI events .

[0098] Each execution entry includes: action object, action direction, step description, channel readback guidance, and a list of verification nodes. A candidate for "adjusting the position of the closing limit switch" is split into three ordered execution entries: Send a PLC command to put the mechanism into slow closing mode.

[0099] The HMI prompts the operator to loosen the limit switch fixing screws.

[0100] The PLC drives the slow closing action, the readback switch triggers the displacement, and the HMI displays the current deviation value.

[0101] The execution mapping also includes readback trigger conditions (such as "displacement change exceeds 0.1 mm") and abnormal interruption conditions (such as "pressure below 20 MPa") to ensure the execution process is safe and controllable. Table 9 shows the execution entries:

[0102] In S4: the execution linkage for establishing a handshake with the PLC command channel, slow splitting and slow merging verification, node-level verification, secondary comparison, report generation, and knowledge item extraction processing are performed to generate a strategy update item structure that includes suggestions for new rules and suggestions for reference versions.

[0103] The execution linkage is to establish a handshake with the PLC command channel; The slow separation and slow combination verification refers to setting the action rhythm and inserting a description of the dwell time; The node-level verification refers to verifying the node position against the neighborhood morphology on the actual fragment; The secondary comparison refers to performing chain-level verification by connecting adjacent entry verification records; The report is generated as a reconstruction of the execution chain grouped by action direction; specifically as follows: The report generation sub-units are grouped by action direction, reconstruct the execution chain in natural order of item numbers, and incorporate node-level verification records and reference curve comparison views. The report generation process consists of three layers: the first layer is based on action direction. Grouping; the second level involves numbering items within each group. The execution chain is reconstructed in a natural order; the third layer links node-level verification records with reference curves under each entry. Let the execution chain... The expression is: ; in: The reconstructed complete execution chain structure.

[0104] : Action direction indicator, with a value of "open" or "close".

[0105] : Belongs to the direction of action The total number of executed entries.

[0106] : No. One execution entry.

[0107] This entry contains the node-level verification record, including the actual node time, deviation, and verification conclusion.

[0108] : A link between the reference curve and the actual segment, such as a hyperlink or embedded image path.

[0109] The left side is the "Close" group, containing 14 entries; the right side is the "Open" group, containing 57 entries. Each entry expands to display verification records and curve comparison links, ultimately generating a structured report.

[0110] The knowledge item extraction process involves extracting knowledge items that are high-frequency and strongly correlated. Specifically: The knowledge entry extraction sub-unit traverses and verifies the input set, and extracts standardized knowledge entries from combinations that appear frequently, have clear causal relationships, or are strongly related to a specific model.

[0111] Define a knowledge entry The form is: ; in: : Trigger condition description, such as "the position of the trigger node in the closing execution segment lags by more than 1mm".

[0112] : Involves a collection of objects, such as .

[0113] Recommended execution order suggestions, such as "Adjust the switch position first, then retest the timing".

[0114] : A collection of tags, including easily confused node identifiers, common exception tags, etc.

[0115] : Evidence description index, pointing to the entry number in the original verification record.

[0116] Frequency calculations use sliding window statistics: Let the total number of historical sessions be A certain group The number of sessions that occurred was The frequency of occurrence .when ( The frequency threshold is typically set to 0.3; and when the causal relationship is clear and there is a consistent correspondence between the deviation direction and the adjustment effect, it is extracted as a standard knowledge item. As shown in Table 10.

[0117]

[0118] Table 10 illustrates the knowledge item extraction process: Input validation input set → Clustering similarity deviation combination → Statistical frequency → Filtering high frequency / strong causal combination → Output standardized knowledge item.

[0119] The generated strategy update entry structure includes new rule suggestions and reference version suggestions, which are used to achieve dynamic updates to the rule base and strategy base.

[0120] In S4, the process of generating a strategy update entry structure that includes new rule suggestions and reference version suggestions also includes: d1: Obtain the suggested instruction set and the comparison benchmark set, perform execution linkage and slow splitting and slow merging review processing, and obtain the execution linkage mapping table; d2: Extract node actions and real-time readback from the execution linkage mapping table, perform node-level verification and secondary comparison processing, and generate a verification input set; d3: Perform report generation and knowledge entry extraction processing on the validation input set, and update the entry structure using the generation strategy.

[0121] The execution linkage includes: establishing an execution session, establishing a handshake with the PLC command channel, and registering a placeholder for a prompt event with the HMI prompt channel.

[0122] The slow division and slow combination verification includes: setting the action rhythm, dividing the action steps into several small steps, and inserting a description of the dwell time.

[0123] The node-level verification includes: comparing the node positions on the actual fragment, and comparing the corresponding node positions and neighborhood morphology on the reference curve with those on the actual fragment.

[0124] The secondary comparison includes: concatenating the node-level verification records of multiple adjacent entries according to the operation method and action direction, and extracting the node chain for verification.

[0125] The report generation includes: grouping by action direction, reconstructing the execution chain in the natural order of item numbers, and importing node-level verification records and reference curve comparison views.

[0126] The knowledge entry extraction includes: extracting standardized knowledge entries, and extracting combinations that appear frequently, have clear causal relationships, or are strongly correlated with specific models.

[0127] In d2, node-level verification includes: The node readback acquisition subunit activates the readback channel one by one according to the mapping table, acquires multi-source data segments within the readback period, and attaches the entry number and operation mode identifier to ensure that the binding relationship between data and action is clear and stable. Under the slow separation and slow combination process, the node back-read acquisition subunit follows the dwell time description and back-read trigger conditions in the mapping table, and forms a corresponding small segment back-read set after each small step of action, covering limit switch on / off, hydraulic pressure, displacement stroke, coil current, environment and vibration. The time alignment and node extraction subunit performs unified time reference alignment on multi-source readback segments under the same session identifier, adopts the previously set timestamp format and sampling interval, and locates the node occurrence time in each segment based on the review node list; The node-level verification subunit performs verification at the item-level granularity. The verification input includes the actual node time, channel context, and reference curve link referenced by the item. The verification method is to compare the corresponding node positions and neighborhood morphology on the reference curve and the actual segment, and record the verification traces of item-node-comparison window; as detailed below: The node-level verification subunit performs verification at the item-level granularity. The verification method involves comparing the corresponding node positions and neighborhood morphology on the reference curve and the actual segment, and recording the verification traces of "item—node—comparison window". The verification subunit will then use the actual node time... Corresponding local waveform window ( (The window width is typically 5ms) and the corresponding window on the reference curve. Perform morphological comparison. The traces were verified and recorded as follows:

[0128] in: : Entry number.

[0129] Node type (e.g., "closing trigger node").

[0130] : The actual time when the node occurs (unit: seconds).

[0131] Position deviation or timing deviation value.

[0132] : Neighborhood morphological similarity, calculated as the correlation coefficient between the reference curve and the actual segment within the window: .

[0133] : List of channel identifiers involved in positioning.

[0134] When multiple channels provide the same node location information (such as the on / off edge and coil current peak), the sub-unit records the comparison results of each channel side-by-side and, based on the evidence embedded in the execution linkage mapping table, adopts a consistent priority strategy (such as prioritizing channels with higher quality levels) to provide entry-level verification conclusions. The upper part is the reference curve window, and the lower part is the actual segment window, highlighting the node neighborhood and labeling the deviation. and similarity .

[0135] When multiple channels provide the same node location information, the node-level verification subunit records the comparison results of each channel in parallel and gives the entry-level verification conclusion based on the evidence description embedded in the execution linkage mapping table using a consistent priority strategy.

[0136] In d2, the secondary contrast includes: The secondary comparison subunit performs a longitudinal check at the action chain level: Under the same session identifier, the node-level verification records of multiple adjacent entries are concatenated according to the operation method and action direction. The node chain of start-trigger-response-stable is extracted, and the temporal relationship and morphological trend between adjacent nodes are checked to see if they are consistent with the reference temporal sequence and reference morphology. When it is found that all item-level verifications are acceptable but the chain-level verification shows inconsistency, the secondary comparison sub-unit marks the corresponding item group with a chain-level inconsistency mark, and includes links involving item range, node type combination and reference curve in the output, prompting the subsequent report generation stage to display in a layered manner.

[0137] This also includes: the interface verification subunit presents a window view of node-level verification and secondary comparison in the HMI overlay view, indexed by item number. Engineers can view actual segments, reference curves, and evidence descriptions on the same screen, and add manual labels or notes when necessary. The interface verification subunit writes the manually added content back to the session layer record; specifically as follows: The interface verification subunit presents a window view of node-level verification and secondary comparison in the HMI overlay view, indexed by the item number. Engineers can view the actual segments, reference curves and evidence descriptions in the same screen, and add manual labels or notes when necessary. The interface verification subunit writes the manually added content back to the session layer record.

[0138] The HMI interface is divided into three areas: Left side: Displays the execution chain in a tree structure, with the root node as the session identifier, the next level as action direction grouping, the next level as item numbering, and a verification status icon displayed next to each item. ).

[0139] Top right area: Curve overlay display window, showing the reference curve (blue) and the actual readback curve (red) on the same coordinate axis, with node positions marked by green circles and neighboring windows labeled with dashed boxes. Engineers can zoom and pan the curves, and click on nodes to view detailed deviation data.

[0140] The bottom right area is for evidence description and manual annotation, displaying the verification details, deviation, similarity, and priority strategy description for the currently selected item. Engineers can add notes or select preset labels (such as "verified" or "retest required") in the text box, and then click "submit" to write it back to the session layer record.

[0141] The currently selected item number is "Close-03", and the verification status is " "Neighborhood morphological similarity" Positional deviation The engineers have added a note: "The switch fixing screws were slightly loose and have been tightened."

[0142] If an entry is not read back within the specified time window or the read back quality is marked as low, the node read back acquisition subunit marks the entry as read back abnormal and automatically excludes its dominant influence on the chain-level conclusion during the second comparison. At the same time, the cause of the abnormality and the channel status are written into the output.

[0143] This invention provides an intelligent adjustment method for the limit switch of a hydraulic operating mechanism, with the following technical advantages: 1) Construct a seamless link from the acquisition configuration set, synchronous data frame and measurement feature set to the comparison benchmark set and benchmark index table. Through node event extraction, working condition segmentation and index mapping, and tag binding, realize a standardized process for generating a node deviation list by comparing position and time series.

[0144] 2) Based on the node deviation list and working condition context, a set of suggestion candidates is formed through reasoning in the rule base and strategy base, and a set of suggestion instructions is generated through model adaptation and execution mapping. The linkage execution and slow separation and slow combination review processes are linked to form a suggestion-execution chain that can be continuously called.

[0145] 3) Obtain node actions and real-time readback from the execution linkage mapping table, perform node-level verification and secondary comparison to form a verification input set, and output the strategy update item structure after report generation and knowledge item extraction to establish a write-back update closed loop for the strategy base and rule base.

[0146] 4) It operates on the operational chain of channel mapping table - acquisition configuration set - synchronization data frame - measurement feature set and comparison benchmark set - benchmark index table to form verifiable node-level position comparison and time sequence comparison results, which is suitable for scenarios where there are multiple source acquisitions and historical curve retrieval processing in the field.

[0147] 5) It operates on the node deviation list - operating condition context - suggested candidate set - suggested instruction set - execution linkage and slow separation and slow combination review process, and outputs an execution-oriented callable instruction structure, which is suitable for multi-model commissioning needs caused by differences between model mapping table and equipment file.

[0148] 6) It operates on the execution chain of the linkage mapping table - verification input set - report generation and knowledge item extraction - policy update item structure, forming incremental update results for the rule base and policy base. It is suitable for application scenarios that need to accumulate experience in comparing the benchmark set and benchmark index table in continuous sessions. Attached Figure Description

[0149] The present invention will be further described below with reference to the accompanying drawings and examples; Figure 1 This is a flowchart illustrating an intelligent adjustment method for the limit switch of a hydraulic operating mechanism, provided in an embodiment of the present invention.

[0150] Figure 2 This is the displacement-stroke curve.

[0151] Figure 3 This is the hydraulic pressure curve. Detailed Implementation

[0152] Example 1: Reference Figure 1 This is a flowchart illustrating an intelligent adjustment method for the limit switch of a hydraulic operating mechanism provided in an embodiment of the present invention. The process may include at least steps S100-S400: S100: Obtain the channel mapping table and the hardware list for acquisition; perform time reference configuration including setting a unified start time and sampling interval; channel drift correction including measuring zero point offset and gain offset; prioritize the acquisition of sensor data for limit switch on / off and key hydraulic pressure quantities; generate multi-source alignment of alignment points by interpolation resampling under a unified time reference; extract node events to locate edge nodes on the on / off quantity track; and process the segmentation of operating conditions by dividing the opening and closing execution segments with node events as anchor points, generating a measurement feature set including a node event table and an operating condition segment table. S200, based on the model mapping table, equipment files and measurement feature set, performs standard threshold loading, historical curve retrieval, index mapping, label binding, location comparison and time series comparison processing to generate a comparison benchmark set and node deviation list; S300: Obtain the node deviation list and measurement feature set, perform abnormal segment identification, working condition aggregation, rule base and strategy base reasoning, model adaptation and execution mapping processing, and generate a suggested instruction set; S400: Based on the suggested instruction set and the comparison benchmark set, perform execution linkage to establish a handshake with the PLC command channel, set the action rhythm and insert a slow split and slow merge review with a description of the dwell time, perform node-level verification by comparing the node position and the neighborhood shape on the actual segment, perform a secondary comparison by connecting the verification records of adjacent entries for chain-level verification, generate a report by grouping and reconstructing the execution chain according to the action direction, and extract and process knowledge entries of high-frequency and strongly correlated combinations to generate a strategy update entry structure that includes suggestions for new rules and suggestions for reference versions.

[0153] Step S100 includes at least steps S110-S130: S110. Obtain the channel mapping table and acquisition hardware list, perform time base configuration and channel drift correction processing, and obtain the acquisition configuration set; Specifically, the channel mapping table generated during the preliminary project preparation phase is used as input. This table describes the wiring relationships, terminal numbers, measurement ranges, and signal formats between the hydraulic operating mechanism and detection points such as limit switch status, hydraulic pressure, displacement stroke, opening and closing sequence, and environmental quantities. A one-to-one correspondence of identifiers is established between the model and terminal dimensions. Simultaneously, a hardware acquisition list is obtained, containing the host model, acquisition board specifications, number of channels and channel resolution, supported input types, calibration certificate number, firmware version, and synchronization bus type of the field acquisition unit. The channel mapping table and the hardware acquisition list are then merged and loaded into the configuration management module, triggering a time reference configuration process. This time reference configuration sets the unified start time, sampling interval, trigger priority, and timestamp format for the entire acquisition link, and establishes a single time reference through a master-slave synchronization strategy when multiple cameras or boards are operating in parallel. Furthermore, channel drift correction processing is performed on the analog-to-digital conversion links of different channels. This channel drift correction processing is used to measure zero-point offset and gain offset under static closed-loop or known steady-state operating conditions, and writes the correction parameters into the channel parameter area according to the calibration record. At the same time, the drift sensitivity level is marked for variable temperature drift channels for dynamic correction in subsequent acquisition processes. In the processing link, the configuration management module generates a channel identifier, range identifier, and safety threshold identifier for each acquisition channel according to the channel mapping table, and associates them with the sampling interval and trigger priority in the time reference configuration to form an externally callable acquisition channel entry. In abnormal situations, if a missing channel mapping, unconnected terminal, or mismatched board firmware version is detected, an acquisition abnormality record is recorded, and the system reverts to the most recent approved configuration snapshot until manual review or hardware recovery. Understandably, this step is executed in the field of industrial automatic control and program control systems. The processing actions conform to the programmed configuration approach, and the data structure organization and field mapping adopt a data-oriented storage and retrieval process that conforms to data organization logic. During operation, the time reference configuration establishes a unified reference through high-priority clock source broadcasting. Channel drift correction is achieved through a dual-channel approach of periodic calibration and on-site steady-state retesting, and the correction status and validity period are attached to the channel entries. The final output is a data acquisition configuration set, which includes channel identifier, range identifier, safety threshold identifier, sampling interval, trigger priority, timestamp format, and drift correction parameters, used to drive subsequent sensor data access and synchronization. This output field is named "Data Acquisition Configuration Set" and is directly consumed during the generation of the synchronization data frame in S120. It also provides a time format and threshold alignment reference for loading standard and historical data in S200 in cross-main steps, and provides a channel status reference for the execution verification in the S400 stage.

[0154] S120. Extract the channel identifier and timing identifier from the acquisition configuration set, perform sensor data acquisition and multi-source alignment, and generate a synchronous data frame; Specifically, the acquisition configuration set output by S110 is used as input. The channel identifier is used to specify the object to be acquired by each acquisition channel (limit switch on / off, hydraulic pressure, displacement stroke, trip coil current, closing coil current, mechanism housing vibration, and ambient temperature, etc.). The timing identifier is used to specify the sampling interval, trigger priority, and timestamp format. At the start of operation, the acquisition execution unit establishes an acquisition queue according to the trigger priority, prioritizes the activation of high-frequency channels for key quantities such as limit switch on / off and hydraulic pressure, and performs slot-style polling of low-frequency quantities under the same time reference. For channels with start-up transients, a pre-buffering and debouncing strategy is first executed to ensure that the first batch of valid samples meets the consistency of the sampling interval and timestamp format. Furthermore, to address the temporal misalignment caused by multi-source data at different sampling rates, multi-source alignment processing is performed. This multi-source alignment processing generates alignment points based on the target sampling time under a unified time reference. By interpolating and resampling and discarding incomplete segments, samples from each channel at the same time are organized to form synchronization entries that can be directly compared and calculated. When a channel corresponding to a channel identifier is found to be disconnected or experiencing short-term packet loss, the acquisition execution unit marks that channel as temporarily unavailable and continues to acquire other channels. During the alignment phase, lost segments are placed in place and a missing flag is written. At the same time, the occurrence time, channel identifier, and duration are recorded in the acquisition anomaly record. During processing, the on / off state of the limit switch is initially determined at the sampling end using a level threshold method, with an edge timestamp attached. Hydraulic pressure and displacement stroke are sampled at equal intervals and quantized once at the hardware end. The opening and closing coil currents are continuously sampled within the trigger window to characterize the action timing. Environmental and vibration channels are included in the same alignment framework as background quantities. The alignment of all channels follows the timestamp format set in S110, ensuring that the same record has the same time location and sampling source identifier at the structural level after alignment. Understandably, the acquisition and execution unit operates in an industrial control network topology, supporting both single-machine direct acquisition and multi-machine collaborative modes. The former is aligned by a local high-precision clock, while the latter is aligned by a master-slave synchronous link. When the master-slave link is briefly interrupted, acquisition is maintained according to the local steady-state clock. After the link is restored, the relative offset caused by the interruption is corrected by a time calibration frame. Through the above acquisition and alignment, a synchronous data frame is constructed. The synchronous data frame is a structured time series, which contains the sample values ​​of each channel, channel identifier, alignment status identifier and missing flag at each moment. It is used to carry out subsequent event extraction and working condition segmentation. The output field is named synchronous data frame and is directly consumed in the measurement feature set generation process of S130. At the same time, in the cross-main step, the synchronous data frame provides the original alignment data for the abnormal segment identification in the S300 stage and also provides the playback basis for the slow division and slow combination review in the S400 stage.

[0155] S130. Extract node events and segment operating condition segments from the synchronous data frame to generate a measurement feature set; Specifically, the synchronous data frame generated by S120 is used as input. The node event extraction is used to identify key nodes related to the limit switch from the aligned multi-source samples. The key nodes include action start nodes, action end nodes, limit switch trigger nodes, opening response nodes, closing response nodes, pressure inflection point nodes, and displacement stagnation nodes. The definition of the nodes follows the rules of level change, timing trigger, and trend change. In the node event extraction process, the edge nodes are first located on the on / off quantity track, the occurrence time and the corresponding channel identifier are recorded, and the time is used as a timing reference. The local changes of hydraulic pressure, displacement stroke and coil current near the same time are associated and labeled to form node candidates. Then, trend change points are searched on the hydraulic pressure track and displacement stroke track. The trend change points are defined based on the rising, stable or falling pattern changes of continuous samples. Combined with the alignment state identifier, the samples before and after the critical change are robustly verified to eliminate false candidates caused by short-term missing time or noise. For the coil current track, the start time of driving coil energization and de-excitation are identified, and a causal sequence is established with the on / off quantity edge nodes to form an action chain. Furthermore, the operating condition segmentation uses the results of node event extraction as anchor points to divide the synchronous data frames into segments such as tripping preparation segment, tripping execution segment, tripping stabilization segment, closing preparation segment, closing execution segment, and closing stabilization segment. Segmentation follows node-to-node interval definition rules, with both ends of each interval originating from confirmed nodes. Within each interval, channel sampling remains complete and time signatures are continuous. Intervals with missing markers are marked as incomplete segments and have a missing percentage appended for weight reduction in subsequent comparison stages. To enhance the algorithm's adaptability to multiple models, the operating condition segmentation simultaneously includes a model identifier, action direction identifier, and operation mode identifier. The operation mode identifier distinguishes between slow tripping, slow closing, and constant-speed actions, facilitating subsequent replay using the same segmentation strategy during the execution verification phase. During operation, node event extraction sets a quality label for each type of node. The quality label is classified according to the node's location criteria (on / off edge, trend change, or multi-source consistency) and the integrity of the neighborhood data. Low-quality nodes are transferred to the anomaly review queue for manual or rule-based review before entering the segmentation process. The working condition segmentation performs sample consistency checks on the neighborhood data across nodes. If unreasonable time signatures or duplicate timestamps are found, the segment is marked as a time anomaly and a backtracking to the multi-source alignment state of S120 is triggered. The anomaly number and time range are recorded to ensure that the data entries output in this step are continuous in structure and time sequence.The final measurement feature set is generated, consisting of a node event table and a working condition segment table. The node event table records node type, occurrence time, participating channel, local context, and quality identifier. The working condition segment table records segment type, start and end nodes, segment length, channel coverage, model identifier, and operation mode identifier. This output field is named "Measurement Feature Set" and is consumed in the comparison benchmark set construction stage in S210. It is used for the alignment mapping of standard node positions and standard time sequences. Simultaneously, it is used for abnormal segment identification and working condition aggregation in stage S300, and for time anchors and playback organization during slow separation and slow combination verification in stage S400. The technical effect of this step can be summarized as follows: By completing acquisition configuration, aligned acquisition, and event segmentation under a unified time benchmark, a structured, comparable, and reusable measurement feature set is formed. This provides a stable data carrier for subsequent standard comparison and recommendation generation, and supports consistent judgment and verification across models and multiple working conditions.

[0156] Step S200 includes at least steps S210-S230: S210. Obtain the model mapping table and equipment file, perform standard threshold loading and historical curve retrieval processing, and obtain the comparison benchmark set; Specifically, the model mapping table formed during the previous tooling configuration phase and the equipment files accumulated during long-term operation and maintenance are used as inputs for this step. The model mapping table provides a unified identification relationship for hydraulic operating mechanisms across different manufacturers, specifications, and terminal blocks, including fields such as model identifier, terminal number, measurement point name, signal type, range, direction of action, and compatibility status. The equipment files provide basic information, standard threshold version records, historical curve storage location information, and operating environment descriptions for the same device during various maintenance and testing processes, covering maintenance date, executor, operating condition description, device serial number, hardware change records, and data acquisition paths. During operation, the standard data loading unit locates the standard threshold matching the model in the standard library based on the model identifier and device serial number in the model mapping table. The standard threshold is a standardized set of action criteria, including limit switch on / off thresholds, hydraulic pressure reference thresholds, displacement travel thresholds, opening / closing timing reference intervals, and environmental quantity safety intervals. The unit then selects and loads the corresponding version based on the standard threshold version recorded in the equipment files. Furthermore, the historical curve retrieval and processing unit retrieves typical curves of the same device under the same or similar operating conditions based on the historical curve storage location information in the equipment file. The typical curves cover the opening process curve, closing process curve, hydraulic pressure change curve over time, displacement stroke change curve over time, and coil current change curve over time. When there are multiple devices of the same model or multiple curves of the same device after multiple maintenance, the curve with the most recent acquisition time and marked as qualified is selected as the reference sample. The time start point and sampling interval of the curve are standardized to be consistent with the timestamp format and sampling interval set in S110 above. If there are missing or storage abnormalities during the retrieval process, the historical curve retrieval and processing unit writes the abnormal event into the standard data loading log and triggers the backup rule: select a representative curve with a higher quality level within the same model range as the replacement item, and attach the source identifier and replacement description in the output of this step. Understandably, in industrial program control scenarios, this step is completed by three types of actions: configuration-driven, data-driven, and retrieval-driven, which conforms to the automated loading and parameterized retrieval characteristics of program control devices. Simultaneously, it employs structured indexing and thematic storage strategies in data organization and field mapping, accommodating retrieval needs across time periods, models, and devices. Through the aforementioned loading and retrieval, the standard data loading unit and the historical curve retrieval processing unit aggregate and integrate the model mapping table, equipment files, standard thresholds, and historical curves, forming a comprehensive set containing standard threshold entries and historical curve entries. Each entry in the set is then written with a model identifier, version identifier, source identifier, and usage scope description to guide subsequent extraction of standard node positions and standard timing sequences.Finally, a comparison benchmark set is obtained. This comparison benchmark set is used as an output field of this step. In S220, it is used to extract standard node positions and standard time sequences, generate index mapping and tag binding. At the same time, in the cross-main step scenario, the comparison benchmark set is also called in the execution linkage and slow splitting and slow merging review process in S410 to provide a comparison reference during review playback.

[0157] S220. Extract the standard node positions and standard time series from the comparison benchmark set, perform index mapping and label binding, and generate a benchmark index table; Specifically, the comparison benchmark set output by S210 is used as input. The standard threshold entries within the comparison benchmark set provide the criterion threshold and action range, while the historical curve entries provide the timing pattern and curve trend under actual operating conditions. The standard element extraction unit first locates candidate nodes on the historical curve entries. Candidate nodes include categories such as action start nodes, action end nodes, limit switch trigger nodes, tripping response nodes, closing response nodes, pressure inflection point nodes, and displacement stagnation nodes. The location rules are jointly constrained by the threshold range given by the standard threshold entries and the timing reference range. Specifically, the unit searches for time points on the time axis of the curve that satisfy threshold crossing and trend change, and performs consistency checks on neighboring data, eliminating false candidates caused by short-term noise or sampling jitter. Further, the standard element extraction unit establishes node sequences in both tripping and closing action directions, and calculates the time interval and sequence relationship between adjacent nodes in the node sequence to form a standard timing sequence. The standard timing sequence is defined as a set of node sequences and intervals verified by historical curves under standard threshold constraints, including sequence order description, interval reference range, and operation mode adaptation description. To support cross-model and cross-device invocation, the index mapping unit writes the relationship between each node category and the corresponding channel into the index item according to the terminal number and measurement point name in the model mapping table. The index item includes the node category, channel positioning information, compatible model range, and operation mode range. The tag binding unit attaches multiple tags to the index item. The tags cover the action direction tag, quality level tag, source tag, and adaptation tag. The quality level tag is used to describe the reliability of the node positioning. The source tag is used to indicate that the node comes from standard threshold inference or historical curve verification. The adaptation tag is used to mark whether the node is universal for slow separation, slow combination, or constant speed processes. Understandably, to meet the collaboration requirements of different data acquisition systems and human-machine interfaces on-site, this step maintains two types of access paths in the index: one for automated process calls, loaded by the control station or programmable logic controller (PLC) during the automatic comparison phase; and the other for manual review and guidance, called by the human-machine interface (HMI) on the prompt page or curve overlay page. Both types of paths maintain a consistent identification system with the model mapping table in the index mapping unit. During operation, when multiple historical curve entries exist in the comparison benchmark set and the quality level labels are inconsistent, the label binding unit adopts a weighted merging strategy: without changing the standard threshold entries, a reference range is given for the time series interval, and the range source label is recorded on the index entry, prompting subsequent comparison phases to perform preference processing on the range boundaries; when a historical curve entry is missing a key node, the standard element extraction unit triggers a backtracking mechanism, returns to the historical curve retrieval processing unit of S210 to request the supplementation of the entry or replacement of the representative curve, and writes the reason for this extraction failure into the index generation log.Through the above extraction, mapping and binding, a baseline index table is generated that can be accessed by both the algorithm and the interface. The baseline index table is the output field of this step and is directly consumed in the position comparison and timing comparison process in S230. At the same time, in the cross-main step scenario, the baseline index table is used to constrain the node reference consistency of the suggested instruction set in the model adaptation and execution mapping process in S330, and provides an index entry for node playback and timing verification in the execution linkage and slow split and slow merge review process in S410.

[0158] S230. Compare the measurement feature set with the benchmark index table in terms of location and time sequence to generate a list of node deviations. Specifically, the measurement feature set generated in S130 and the reference index table generated in S220 are used as inputs. The measurement feature set contains structured data consisting of a node event table and a working condition segment table. The node event table records the node type, occurrence time, participating channel, local context, and quality identifier in the actual action. The working condition segment table records the segment type, start and end nodes, segment length, channel coverage, model identifier, and operation mode identifier. The reference index table contains index items for standard node positions and standard timing, channel positioning information, and multiple tags. The comparison and judgment unit first reads the segment category, model identifier, and operation mode identifier from the working condition segment table, selects the matching set of index items, and establishes a standard timing reference within the set according to the action direction and node sequence order. Subsequently, it locates the actual nodes one by one in the node event table according to the node category. For each actual node, it accesses the channel positioning information and reference range given in the reference index table, compares the occurrence time of the actual node with the channel context, and forms a position comparison result and a timing comparison result. The position comparison is used to describe the deviation between the actual node's landing point and the reference range in terms of displacement stroke, hydraulic pressure, and coil current, etc., and the timing comparison is used to describe the difference between the actual interval between the actual node and the adjacent nodes before and after it and the standard timing reference interval. Furthermore, the comparison and judgment unit performs neighborhood reinforcement on the actual nodes with low-quality identifiers: when a node has a low-level quality identifier in the node event table, it reads the adjacent nodes and channel context within the same working condition segment, and uses the neighborhood consistency rule to determine whether the node can be corrected. If it can be corrected, it outputs the corrected node time and channel context description and adds a correction label to the result of this step; if it cannot be corrected, it participates in the comparison with the original value and adds a low-quality label to the result. For segments with missing identifiers, the comparison and judgment unit uses a weighted processing method to mark the time sequence comparison results within the segment based on the alternative description written in stage S210 of the comparison benchmark set, prompting caution in subsequent abnormal segment identification steps. To address channel differences arising from multiple models and devices, the comparison and judgment unit filters index items consistent with the current model and operating method using adaptation tags in the benchmark index table, avoiding comparisons within incompatible index ranges. In actual operation, when the same node provides positioning basis on multiple channels, such as when the on / off edge and coil current rising edge coexist, the comparison and judgment unit records both types of basis simultaneously. If they are inconsistent, priority is selected according to the quality level tag provided by the tag binding unit, and a weight tag is written into the output of this step, prompting the subsequent suggestion generation stage to adopt different processing paths for the deviation interpretation of this node. Understandably, to support automated execution and manual verification during on-site comparison, this step also writes the comparison trace into a replayable record. The replayable record includes the node matching process, reference range call record, and neighborhood reinforcement record, which are displayed for traceability on the prompt page of the human-machine interface.By comparing location and time sequence, the deviation type, direction, and magnitude of each node are summarized, and associated with its corresponding segment, node category, channel positioning information, and tag description to form a node deviation list. This node deviation list, as an output field of this step, is directly consumed in the abnormal segment identification and condition aggregation processing in S310 to identify suspicious segments and aggregate condition contexts. Simultaneously, in cross-main step scenarios, the node deviation list is used to generate a suggestion candidate set in the rule base and strategy base inference in S320, and participates in playback and secondary comparison as a review comparison item in the execution linkage and slow separation / merging review processing in S410. In summary, the technical effect of this step is that by structurally comparing the standard elements of the comparison benchmark set with the actual elements of the measurement feature set, a traceable and callable node deviation list is formed, providing a clear and unified judgment basis for subsequent anomaly identification and commissioning suggestion generation.

[0159] Step S300 includes at least steps S310-S330: S310. Obtain the node deviation list and measurement feature set, perform abnormal segment identification and working condition aggregation processing, and obtain the working condition context. Specifically, the node deviation list output by S230 and the measurement feature set output by S130 in the previous main step are used as inputs for this step. The node deviation list is used to record the deviation type, deviation direction and deviation magnitude of each node in terms of position and timing, and includes the segment to which it belongs, node category, channel positioning information and label description. The measurement feature set consists of a node event table and a working condition segment table. The node event table is used to characterize the node occurrence time, participating channel, local context and quality identifier in the actual action process. The working condition segment table is used to characterize the start and end nodes, segment length, channel coverage, model identifier and operation mode identifier of segments such as opening preparation, opening execution, opening stabilization, closing preparation, closing execution and closing stabilization. During runtime, the abnormal segment identification unit first verifies the continuity of the operational segment table. Based on the start and end nodes and time extension relationship of the segments, it checks for duplicate time markers, unreasonable spans, sudden drops in channel coverage, and concentrated missing markers. When any of these situations occur, the segment is marked as a suspected segment candidate, with the triggering reason, involved channels, and time range appended to the candidate list for subsequent aggregation and interpretation. Subsequently, the abnormal segment identification unit calculates the node deviation density for each segment based on the node deviation list. Deviation density describes the distribution of deviation information within a unit segment length or unit number of nodes. When the deviation density of a segment is significantly higher than the historical distribution or standard reference range of similar segments, the segment is included in the abnormal segment set, and the segment category, dominant deviation direction, and involved key node categories are recorded in the set. Furthermore, the abnormal segment identification unit performs a robustness assessment of the abnormal conclusion of the segment based on the quality identifier in the node event table. If the proportion of low-quality identifier nodes in the abnormal segment is high, a neighborhood review is triggered: the original alignment samples of the synchronous data frame are read from the left and right neighborhoods of the segment's start and end nodes, the edge time and trend turning point are checked a second time, and the review results are written back to the evidence field of the abnormal segment set. Understandably, the abnormal segment identification unit operates in an industrial program control scenario, adopting a process that combines rule-oriented triggering with data-oriented evidence writing back. It follows the judgment boundaries of standard thresholds and timing references, while also taking into account the differences in data integrity caused by changes in on-site acquisition conditions. When a record of acquisition link interruption or clock state change occurs, the abnormal segment identification unit incorporates this information into the evidence field to avoid mistakenly attributing external causes to the device mechanism during subsequent strategy generation. After completing the abnormal segment identification, the operating condition aggregation processing unit starts the aggregation process. The operating condition aggregation is used to merge and thematically organize the discrete information composed of multiple segments and nodes in both the opening and closing directions during the same operation.Specifically, the operating condition aggregation processing unit establishes operating condition topics using action direction identifiers and operation mode identifiers as primary keys. It then categorizes abnormal fragment sets, node deviation lists, and relevant entries from the node event table into corresponding topics. Within each topic, it constructs a fragment linked list according to the fragment time sequence, attaching the dominant deviation direction, key node combinations, and channel coverage features to each fragment, forming semantic groups required for subsequent reasoning. To support consistent use across models and devices, the operating condition aggregation processing unit writes a model identifier and adaptation description to each operating condition topic, specifying the operation mode and action speed condition from which the topic originates. When the same action exhibits fragment structure differences in multiple executions, the operating condition aggregation processing unit merges them according to the similarity of the fragment linked lists, forming an aggregated topic with difference annotations. Within each topic, it distinguishes between stable and variable fragments, enabling different interpretation paths to be adopted based on the fragment nature during subsequent rule base and strategy base reasoning. During operation, the work condition aggregation processing unit also performs cross-topic cross-search on the abnormal segment set. If an abnormal segment repeatedly appears in multiple executions with similar deviation directions and combinations of similar key nodes, a duplicate identifier is written to the topic entry of that segment, prompting the subsequent strategy generation module to give priority to such recurring anomalies. Combining the above identification and aggregation, a work condition context is generated. The work condition context is used to centrally express the organizational relationship, anomaly distribution, and dominant deviation factors of each segment and node in one or more actions, and carries model identifiers, operation method identifiers, and evidence fields at the entry level, adapting to both automatic comparison and manual review dual-path calls. This output field, named work condition context, will be directly consumed in the rule base and strategy base inference of S320, used to extract node types and deviation directions from the node deviation list and work condition context and generate a suggested candidate set. At the same time, it serves as the thematic input for replay organization and on-site prompts in the execution verification of the S400 stage.

[0160] S320. Extract node type and deviation direction from the node deviation list and working condition context, perform rule base and strategy base reasoning, and generate a set of suggested candidates. Specifically, the node deviation list output by S230 and the operating condition context output by S310 are used as inputs. The node type is used to identify the role attribute of the node involved in the deviation in the action chain, covering categories such as action start node, action end node, limit switch trigger node, tripping response node, closing response node, pressure inflection point node, and displacement stagnation node. The deviation direction is used to identify the offset trend of the deviation at the position or time level, and there is a one-to-one correspondence with the segment, channel positioning information, and label description. When inference begins, the rule base and strategy base inference units first locate the direction and mode of the action to be processed based on the topic entries of the working condition context. They then read the set of abnormal segments and the segment list, collecting the associated node type combinations and dominant deviation directions for each segment to form a segment feature description. Subsequently, the rule matcher searches for matching entries in the rule base based on the segment feature description. The rule base is a structured set of rules refined through engineering and expert review. The antecedent of a rule consists of node type combinations, dominant deviation directions, and segment categories, while the consequent consists of executable adjustment suggestion primitives. These primitives describe the direction and step of adjustment actions such as limit switch positions, operating speeds, pre-filled oil levels, and test procedure sequences. Furthermore, when the rule base does not provide a completely matching entry, the strategy base inferrer initiates a strategy inference path. The strategy base is a collection of empirical and data-driven strategies collected over long-term operation, including adjustment patterns, operation sequences, and verification steps that can be transferred to similar segment structures. The strategy base inferrer generates candidate adjustment patterns for the current segment based on similar topics and duplicate identifiers in the operating context. It then sorts the candidates by combining these with weight tags from the node deviation list. These weight tags originate from the priority selection record in S230 when multiple channels have inconsistent criteria. Understandably, to support online closed-loop applications of automatic control systems, the rule base and strategy base inference units in this step are backward compatible with the structured call interfaces of control stations and programmable logic controllers (PLCs), and upward compatible with the interpretation and guidance views of human-machine interfaces (HMIs). Both share the same chain of evidence during the generation of the same suggestion. Simultaneously with candidate generation, the inference unit writes an evidence description for each candidate. This description describes the triggering rule number or strategy source, the operating topic and node deviation source used, the expected channels and operational objects involved, and provides alternative explanations when data gaps exist, prompting the execution phase to perform step-by-step verification of the candidate.During operation, if multiple mutually exclusive candidates exist within the same segment—for example, one candidate points to limit switch position adjustment while another points to operating speed adjustment—the inference unit determines the conflict of candidates based on the segment list and node causal relationships within the operating context. Candidates that may introduce secondary offsets are marked for delayed execution, and execution order hints are added to the candidates, allowing subsequent execution modules to be arranged in an order of ease to difficulty or root cause to manifestation. To adapt to differences in different models and wiring terminals, the inference unit associates a model identifier and terminal mapping description with each candidate. This description is derived from the relationship between terminal numbers and measurement point names in the preceding model mapping table, facilitating rapid mapping to the actual wiring and tooling of the execution layer in the next step. Through the above matching, inference, and evidence generation, a suggested candidate set is formed. This suggested candidate set is used to centrally express a set of adjustment suggestion primitives and review suggestions that can be invoked by the execution layer, including candidate identifiers, evidence descriptions, involved objects, execution order hints, model identifiers, and terminal mapping descriptions. This output field, named "Suggested Candidate Set," will be directly consumed in the S330 model adaptation and execution mapping process to convert candidates into a set of suggested instructions that can be issued. Simultaneously, it will be read by both the HMI and PLC during the execution verification in the S400 stage to demonstrate guidance and action triggering.

[0161] S330. Perform model adaptation and execution mapping on the proposed candidate set to generate a proposed instruction set; Specifically, the suggested candidate set output by S320 is used as input. The model adaptation is used to instantiate the adjustment objects in the suggested candidates under different manufacturers and different specifications, ensuring that the object names, terminal instructions, and tooling operation steps in the candidates are consistent with the model identifier of the current device. The execution mapping is used to translate the instantiated suggestions into instruction entries that can be recognized by the execution layer, covering two categories: automatic triggering instructions for control stations and programmable logic controllers (PLCs) and interactive prompt instructions for human-machine interfaces (HMIs). During operation, the model adaptation unit first reads the model identifier and terminal mapping description in the suggested candidate set, and accesses the relationship between terminal numbers and measurement point names in the model mapping table, mapping the abstract objects in the candidates to specific wiring terminals, sensor channels, and mechanical adjustment positions. When the candidate involves limit switch position adjustment, the model adaptation unit provides the corresponding contact pair and fixing method according to the terminal mapping description. When the candidate involves operating speed or pre-charge oil volume, the model adaptation unit provides the corresponding adjustment knob, valve position, and allowable adjustment step range according to the device parameters in the equipment file. Subsequently, the execution mapping unit performs item-based decomposition based on the execution order prompts attached to the suggested candidate set, splitting a candidate into a group of ordered execution items. Each execution item includes the action object, action direction, step description, channel readback guidance, and a list of verification nodes. In automatic triggering scenarios, the execution mapping unit maps the execution items to instruction formats that can be called by the PLC, and adds readback trigger conditions and abnormal interruption conditions to the instruction items, so that on-site execution can proceed to the next item when data readback and safety limits are met. In human-machine interaction scenarios, the execution mapping unit maps the execution items to HMI prompt events. Prompt events include three types of content: text guidance, graphical guidance, and curve overlay display. Graphical guidance marks the channels and nodes that need attention on the interface, and curve overlay display overlays the reference curve of the comparison benchmark set and the current readback curve on the same screen, providing operators with a reference for comparison. Furthermore, to accommodate both slow separation and slow merging verification processes, the execution mapping unit adds operation mode identifiers to entries involving slow separation and slow merging in the suggestion instruction set, and writes descriptions of triggering order and dwell time into the entries, ensuring that the verification process can fully cover every node and segment that needs to be observed. When there are entries with delayed execution in the suggestion candidate set, the execution mapping unit marks the entry as pending triggering and writes a precondition description into the suggestion instruction set. The precondition is then activated when it is met during the execution phase. Understandably, this step operates in a fusion scenario of industrial automation control and data organization, providing two types of outputs: structured instructions for the control system and structured prompts for the interface. Both are consistent through unified entry numbering and evidence descriptions to facilitate source tracing and secondary comparison during the execution verification phase.In terms of exception handling, when the model adaptation unit detects an inconsistency between the candidate and the model mapping table or a missing terminal mapping, it temporarily stores the candidate in a suspended state and writes a suspension explanation and supplementary requirement prompt into the suggested instruction set, prompting the operator or engineering configuration personnel to update the model mapping table before triggering remapping. Through the above adaptation and mapping, a suggested instruction set is generated. This suggested instruction set is used for issuing at the execution layer and prompting at the interface layer. It includes execution items, prompt events, readback trigger conditions, exception interruption conditions, operation mode identifiers, and evidence descriptions, and establishes a visual link with the reference curve of the comparison benchmark set for overlay display during the execution phase. This output field, named "Suggested Instruction Set," will be directly consumed in the execution linkage and slow separation and slow combination review process of S410, driving the execution linkage and readback review process. At the same time, in cross-main step scenarios, it provides item numbers and evidence description indexes for the node-level verification and secondary comparison of S420, and provides traceable execution and prompt records for the report generation and knowledge item extraction of S430.

[0162] In summary, the technical effects of this step are as follows: By reasoning from the node deviation list and operating context to the rule base and strategy base, and completing model adaptation and execution mapping, a set of suggested instructions that can directly drive on-site execution and readback verification is formed; the suggestion layer and execution layer form a closed loop through unified entries and evidence association, providing a stable call entry point for subsequent verification and knowledge accumulation.

[0163] In one specific embodiment, S300, the node deviation list and measurement feature set are obtained, and abnormal segment identification, working condition aggregation, rule base and strategy base reasoning, model adaptation and execution mapping processing are performed to generate a suggested instruction set.

[0164] Following the node deviation list output by S230 and the measurement feature set output by S130, this section first performs abnormal segment identification and operating condition aggregation. It reads the operating condition segment table from the measurement feature set and performs a continuity check on each segment: checking whether the start and end node times meet the requirements. And count the proportion of missing markers. ,in This represents the total number of missing markers within the segment. The theoretical number of sampling points; if If the time sequence is reversed, it is marked as a candidate for a suspicious segment. Then, the number of deviation entries belonging to each segment is extracted from the node deviation list. Calculate the deviation density. Formula ① defines the segment. Deviation density : ①; Variable and symbol definition: : fragment The deviation density, with the dimension "numbers / second" and a range of non-negative real numbers, is calculated from the "node deviation list" and the "measurement feature set".

[0165] : Index identifier for the working condition segment, with a value range of positive integers, pointing to a specific segment in the working condition segment table of the "Measurement Feature Set".

[0166] : belongs to fragment The total number of node deviation entries, measured in "number"; indexed by segment from the "Node Deviation List". The result is obtained by aggregation counting.

[0167] : fragment The duration, measured in seconds, is calculated from the start and end times of the working condition segment table in the "Measurement Feature Set". .

[0168] superscript : Indicates that the variable belongs to a fragment. , and index identifier They have the same meaning.

[0169] Data Source → Metrics → Variable Mapping: Extracting the segment belonging to the node deviation list The number of entries is obtained Extracted from the working condition segment table in the "Measurement Feature Set" And calculate .

[0170] Simple numerical example: A closing execution segment The corresponding entry for this segment in the node deviation list. ,but Preset abnormal threshold This fragment was included in the abnormal fragment set. .

[0171] Formula ① addresses the practical problem of quantifying the concentration of deviations within each operating condition segment, providing a calculable criterion for automatically identifying abnormal segments, and avoiding subjective human judgment.

[0172] After completing the deviation density calculation, indicate the direction of action. Operating method identifier Perform condition aggregation for composite primary keys. Include the set of abnormal fragments. Node Deviation List Node event table All of them have the same The entries are grouped into the same topic. To quantify the repetitiveness of anomalies within a topic, Formula ② defines a repetition identifier. : ②; Variable and symbol definition: :theme Repeating identifier, dimensionless, range of values ; indicates the frequency of the current abnormal pattern in the history of sessions.

[0173] Composite primary key This indicates the direction of operation (values: "open" or "close"). This is an operating mode identifier (values ​​include "normal speed", "slow disengagement", or "slow engagement"); it originates from the topic definition in the "operating condition context".

[0174] : Index of historical sessions From 1 to Sum.

[0175] : Index of historical sessions, with a value range of positive integers and a maximum value of . .

[0176] Total number of historical sessions, measured in "number"; obtained from statistics in "device archives".

[0177] : Indicator function, takes the value 1 if the condition in parentheses is true, otherwise takes the value 0.

[0178] : The condition of the indicator function, indicating "the first" Anomaly pattern description vector of the next historical session The exception pattern description vector of the current session resemblance".

[0179] : No. The vector describing the abnormal patterns of this topic in the previous historical session; derived from the historical verification input set stored in the "Device Profile".

[0180] : Anomaly pattern description vector for this topic in the current session; derived from the combination of the set of anomaly fragments and node deviation directions calculated in this step.

[0181] : Similarity relation symbol, indicating that the cosine similarity between two vectors is greater than a preset threshold of 0.7 (dimensionless).

[0182] Simple numerical example: Total number of historical sessions If the abnormal patterns of 6 historical sessions are similar to the current one, then... If the repetition threshold of 0.3 is exceeded, it will be labeled as "high frequency repetition".

[0183] Formula ② addresses the practical problem of identifying frequently occurring abnormal patterns, providing a quantitative basis for prioritizing these patterns during subsequent policy base inference. The output of this section is the operational context, containing fields such as topic ID, fragment list, dominant deviation summary, and duplicate identifier, which are directly consumed by the S320's rule base and policy base inference.

[0184] Following the operating context and node deviation list, this section performs rule base and strategy base reasoning. The rule base uses structured rules, with each rule's antecedent composed of node types. Dominant Deviation Direction Fragment categories The structure consists of a subordinate clause, which is an adjustment suggestion primitive. When no exact match is found in the rule base, policy base inference is initiated. Formula ③ defines the relationship between the current topic and the first... Similarity of historical strategies : ③; Variable and symbol definition: Current topic and the first The similarity of historical strategies, dimensionless, with a range of values. Used to select the best-matching historical experience.

[0185] Historical strategy index, with a value range of positive integers.

[0186] : Weighting coefficients, with values ​​of 0.3, 0.2, and 0.5 respectively, dimensionless.

[0187] : Indicator function, same as formula ②.

[0188] : Current action direction indicator (e.g., "closing"); derived from "operating condition context".

[0189] : No. The action direction indicator for each historical strategy; sourced from the "Strategy Library".

[0190] : Current operating mode identifier (e.g., "normal speed"); derived from "operating condition context".

[0191] : No. The operation method identifier for each historical strategy; sourced from the "Strategy Library".

[0192] : The current collection of abnormal fragments, where each element is a fragment index. ; This originates from the set of abnormal fragments output by formula ①.

[0193] : No. A collection of abnormal fragments recorded in the historical strategy; sourced from the "strategy library".

[0194] The cardinality (number of elements) of a set is dimensionless.

[0195] The intersection operation represents the elements common to two sets.

[0196] The set union operation represents all elements of two sets.

[0197] Subscript : Represents the value of the current session.

[0198] Subscript : Represents the value of the historical strategy.

[0199] superscript : indicates the first A historical strategy.

[0200] Simple numerical example: Current closing speed theme, , A certain historical strategy , The intersection size is 2, the union size is 3, and the similarity is... .

[0201] Formula ③ addresses the practical problem of quantifying the degree of matching between the current abnormal pattern and historical experience, thereby enabling automated selection for experience transfer.

[0202] If multiple mutually exclusive candidates exist within the same segment (such as position adjustment and speed adjustment), conflict determination and execution order marking are required. Formula ④ defines the execution order priority. : ④; Variable and symbol definition: : No. The execution order priority of each candidate is dimensionless and has a range of values. The larger the value, the higher the priority for execution.

[0203] : Candidate index, which is a positive integer and points to a candidate in the "suggested candidate set".

[0204] : Natural exponential function, with As the base.

[0205] Logistic regression coefficients, respectively taken as... , , , dimensionless.

[0206] : No. The conflict index of each candidate is 0 or 1. It is determined by the root cause relationship between candidates: if there is another candidate that conflicts with it (such as position adjustment and velocity adjustment being mutually exclusive), the value is 1; otherwise, the value is 0.

[0207] : No. The strength of causal relationship for each candidate, with a range of values. It is obtained by weighted calculation of the multi-channel priority records in the "Node Deviation List".

[0208] Simple numerical example: A candidate involves adjusting the position of a limit switch; conflict index. (Conflicts with another velocity regulation candidate), causal strength ,but It has a high priority and is marked as being executed first.

[0209] Formula ④ addresses the practical problem of automatically determining the execution order of conflicting candidates, avoiding secondary deviations caused by blind adjustments. The output of this section is a suggested candidate set, including candidate identifiers, evidence descriptions, involved objects, execution order hints, model identifiers, and terminal mapping descriptions, which are consumed by the S330's model adaptation and execution mapping.

[0210] Following the suggested candidate set, this section performs model adaptation and execution mapping. Model adaptation maps abstract objects to specific wiring terminals, fixing methods, and adjustment steps. Formula ⑤ defines the model adaptation mapping function. : ⑤; Variable and symbol definition: Model matching mapping function, which maps abstract objects and model identifiers to specific hardware operation parameters.

[0211]

[0212] : The model identifier of the current device, a string type; derived from the "Model Identifier" field in the "Model Mapping Table".

[0213] and

[0214]

[0215]

[0216] Simple numerical examples: , The corresponding entry in the model mapping table: , , ,but Output the above triplet.

[0217] Formula ⑤ addresses the practical problem of converting general recommendations into actionable instructions specific to a particular hardware model, thus eliminating ambiguity caused by model differences.

[0218] The execution mapping breaks down the adapted specific actions into ordered execution entries and generates PLC instructions and HMI events. Formula 6 defines the execution mapping function. : ⑥; Variable and symbol definition: : Execute the mapping function to break down the specific action sequence into ordered execution entries and their corresponding PLC instructions and HMI events.

[0219] The output of formula ⑤ is combined with the execution order prompt.

[0220] : Action sequence length (number of actions), dimensionless, and takes the value of a positive integer.

[0221] From 1 to The set union operation represents combining all entries into a single set.

[0222] : Execute the index of the entry, the value range is positive integers, and the maximum value is .

[0223] Each execution entry includes the action object, action direction, step description, channel readback guide, and review node list; sourced from One action

[0224]

[0225]

[0226]

[0227]

[0228] : Represents an unordered set containing three elements: execution entries, PLC instructions, and HMI events.

[0229] Simple numerical examples: After a candidate is adapted, an action sequence is obtained. , After mapping: , , .

[0230] Formula ⑥ addresses the practical problem of decomposing a high-level adjustment suggestion into a sequence of atomic operations that can be executed one by one and verified by rereading, while simultaneously generating two execution paths: automated and manual. The output of this section is a set of suggestion instructions, including execution entries, prompt events, rereading trigger conditions, exception interruption conditions, operation mode identifiers, and evidence descriptions, which are directly consumed by the S410's execution linkage and slow-divide-and-merge review process.

[0231] This section summarizes the technical effects: Through anomaly segment identification and operating condition aggregation, rule-and-strategy dual-engine reasoning, and model adaptation and execution mapping, the node deviation list is transformed into an ordered instruction chain that can directly drive the PLC and HMI, forming a closed-loop decision-making process from deviation to execution. The similarity and priority quantification methods in formulas ③ and ④ differ from conventional threshold comparisons, introducing historical experience transfer and conflict resolution mechanisms, significantly improving the adaptability and robustness of commissioning recommendations.

[0232] Step S400 includes at least steps S410-S430: S410. Obtain the suggested instruction set and the comparison benchmark set, perform execution linkage and slow splitting and slow merging review processing, and obtain the execution linkage mapping table. Specifically, the suggested instruction set output by the preceding step S330 and the comparison benchmark set output by S210 are used as inputs for this step. The suggested instruction set provides execution entries, prompt events, readback trigger conditions, abnormal interruption conditions, operation mode identifiers, and evidence descriptions for the control and interface layers. The comparison benchmark set provides a basis for comparing reference curves, standard node positions, and standard timing sequences, and includes model identifiers, version identifiers, and source identifiers at the field level. At the start of operation, the execution linkage and slow-separation / slow-merge review processing unit establishes an execution session, reads the sequence of execution entries from the suggested instruction set, and organizes them into an ordered queue that can be distributed according to the execution order prompts, operation mode identifiers, and preconditions within each entry. Understandably, the operation mode identifier is used to distinguish between slow separation, slow merging, and constant-speed processes, and the preconditions are used to activate subsequent entries after the established safety boundaries and dependencies are met. Subsequently, the linkage scheduling subunit establishes a handshake with the command channel of the Programmable Logic Controller (PLC) according to the queue order, and registers the corresponding prompt event placeholder with the prompt channel of the Human-Machine Interface (HMI). Both share the same entry number and evidence description, facilitating the reuse of the evidence chain under the same main line for subsequent review and verification. For each execution entry, the linkage scheduling subunit parses the action object, action direction, and step description from the entry content of the suggested instruction set, and accesses the channel review guidance and review node list given by the suggested instruction set to establish a closed ternary relationship of action trigger—review trigger—review node. Under the two types of processes of slow separation and slow merging, the slow execution control subunit sets the action rhythm according to the operation mode identifier, divides the action step into several small steps, and inserts a dwell time description between the action trigger and the review trigger of each small step, so as to facilitate step-by-step verification on the reference curve and node position of the comparison benchmark set. Specifically, for items that require the linkage of multiple components, the linkage scheduling subunit identifies the involved objects and their dependency order based on the evidence in the suggested instruction set. It first issues a trigger to the upstream object, and then releases the trigger to the downstream object after the readback trigger condition is met, forming a linkage chain of trigger-readback confirmation-re-trigger. For items that only require interface prompts and do not trigger automatic actions, the HMI prompt subunit presents the prompt events item by item using text guidance, graphical guidance, and curve overlay display. The graphical guidance is used to highlight the currently focused channels and nodes on the interface, and the curve overlay display is used to establish a comparison view between the current readback curve and the reference curve of the comparison benchmark set.Furthermore, to prevent inconsistent states during execution, the abnormal interruption and rollback mechanism is always effective at the session level: when the status indication returned by the PLC or the manual confirmation returned by the HMI is inconsistent with the readback trigger condition described in the entry, the abnormal management subunit immediately records the entry number, trigger time, readback channel, reference node, and inconsistent description, and marks the current entry as interrupted, executing a rollback or pause according to the abnormal interruption conditions in the suggested instruction set; in the paused branch, the HMI prompt subunit prompts the manual review of the key node layer in the graphical guide, and expands a local comparison area between the reference curve in the comparison benchmark set and the current curve to determine whether to continue. For the entire execution session, the execution log subunit records the action object, action direction, step description, readback trigger channel, review node, reference curve source, and label description in real time at the entry granularity, and generates a unique session identifier and entry sequence for each linkage chain, serving as the index entry for subsequent verification and report generation. Through the aforementioned queued distribution, slow-paced control, linkage sequence organization, and exception handling, an execution linkage mapping table is constructed. This table fully describes the mapping relationship between entries, action objects, triggering conditions, readback channels, review nodes, reference curve links, and exception branches, carrying entry numbers, session identifiers, operation mode identifiers, and evidence description indexes. The output field, named "Execution Linkage Mapping Table," is directly consumed in the node action and real-time readback stages of S420. It is used to extract node actions and readback data item by item and complete node-level verification and secondary comparison. Simultaneously, in cross-main steps, the execution linkage mapping table provides a structured entry point for playback, positioning, and tracing within the S400 stage, and maintains a visual link with the comparison benchmark set of S210 at the interface layer.

[0233] S420. Extract node actions and real-time readbacks from the execution linkage mapping table, perform node-level verification and secondary comparison, and generate a verification input set. Specifically, the execution linkage mapping table output by S410 is used as the input for this step. The execution linkage mapping table organizes the action triggering, readback channels, verification nodes and reference curve links at the item level with session identifiers and item numbers. During the execution session, the node readback acquisition subunit activates the readback channels one by one according to the mapping table, collects multi-source data segments within the readback period, and attaches the item number and operation mode identifier to ensure that the binding relationship between data and action is clear and stable. Under the slow split and slow merge process, the node readback acquisition subunit follows the dwell time description and readback trigger conditions in the mapping table to form a corresponding small segment readback set after each small step of action, covering channels such as limit switch on / off, hydraulic pressure, displacement stroke, coil current, environment and vibration. Subsequently, the time alignment and node extraction subunit performs unified time reference alignment on the multi-source readback segments under the same session identifier, using the timestamp format and sampling interval set in the preceding S110, and locating the node occurrence time within each segment based on the verification node list. Understandably, the verification node list comes from the entries in the execution linkage mapping table, and its node type and reference position definition are consistent with the standard node positions and standard timing extracted in S220. After alignment, the node-level verification subunit performs verification at the entry-level granularity. The verification input includes the actual node time, channel context, and reference curve links referenced by the entry. The verification method involves comparing the corresponding node positions and neighborhood morphology on the reference curve and the actual segment, and recording the verification traces of the entry-node-comparison window. When multiple channels provide the same node location basis, the node-level verification subunit records the comparison results of each channel side-by-side, and gives the entry-level verification conclusion based on the evidence embedded in the execution linkage mapping table using a consistent priority strategy. Furthermore, the secondary comparison subunit performs a vertical check at the action chain level: under the same session identifier, it concatenates the node-level verification records of multiple adjacent items according to the operation method and action direction, extracts the start-trigger-response-stable node chain, and checks whether the temporal relationship and morphological trend between adjacent nodes are consistent with the reference temporal sequence and reference morphology. When it is found that the item-level verifications are all acceptable but the chain-level verification shows inconsistency, the secondary comparison subunit marks the corresponding item group with a chain-level inconsistency mark and includes links involving the item range, node type combination, and reference curve in the output, prompting subsequent report generation to display them in a layered manner. To balance automatic verification and manual review, the interface review subunit presents a window view of node-level verification and secondary comparison in the HMI overlay view, indexed by item number. Engineers can view the actual segment, reference curve, and evidence description in the same screen, and add manual labels or notes when necessary. The interface review subunit writes the manually supplemented content back to the session layer record.Regarding anomalies, if an entry fails to be read back within a specified time window or its read back quality is identified as low, the node read back acquisition subunit marks the entry as a read back anomaly and automatically excludes its dominant influence on the chain-level conclusion during the secondary comparison. Simultaneously, the cause of the anomaly and the channel status are written into the output. Through the above extraction, alignment, node-level verification, and secondary comparison, a verification input set is formed. This verification input set is archived by session identifier and contains entry-level verification records, chain-level secondary comparison records, anomaly and manual annotation records, reference curve links, and evidence description indexes. This output field, named "Verification Input Set," is directly consumed in the report generation and knowledge entry extraction in S430 to form structured report content and knowledge entries. Furthermore, in cross-main-step scenarios, the verification input set can be fed back to the rule base and strategy base inference in S320 as fact samples to support incremental learning of the strategy base.

[0234] S430. Generate a report and extract knowledge items from the validation input set, and update the item structure using the generation strategy. Specifically, the verification input set output by S420 is used as the input for this step. This verification input set consists of item-level verification records, chain-level secondary comparison records, anomaly and manual annotation records, reference curve links, and evidence description indexes, and possesses a complete traceability path in terms of session identifier and item number. After reading the verification input set, the report generation subunit groups the items within the session according to the action direction and operation method, reconstructs the execution chain in the natural order of item numbers, imports node-level verification records and reference curve comparison view links under each item, and centrally presents the chain-level secondary comparison results. To conform to engineering archiving and auditing practices, the report generation subunit appends a session identifier, model identifier, and time range description before the paragraph of each group of items. Within the paragraph, the content is organized in the order of item—node—evidence, and, when necessary, references the manual annotations written back by the interface review subunit as side notes. In automated system integration scenarios, the report output is stored on disk in the form of a structured document, and the source relationship, version number, and generation time are recorded at the metadata layer to facilitate subsequent cross-indexing in the knowledge base. The knowledge entry extraction subunit traverses and verifies the input set while generating the report. It extracts fragment-node-deviation direction-execution entry combinations that occur frequently, have clear causal relationships, or are strongly related to a specific model into standardized knowledge entries. Each knowledge entry includes a description of the triggering condition, the objects involved, a suggested execution order hint, easily confused nodes, and common anomaly tags, and retains an evidence index for traceability. When a knowledge entry differs from an existing entry in terms of triggering conditions or recommended execution order, the knowledge entry extraction subunit does not overwrite it but generates a new entry and establishes mutually exclusive or complementary relationship markers between entries for subsequent use by the strategy library during inference. Furthermore, the strategy update orchestration subunit maps newly generated knowledge entries and key conclusions from the report to executable strategy update entries. These strategy update entries structurally express incremental update suggestions for the rule base and strategy base, covering four categories: new rules or strategies, tag correction, reference range adjustment, and entry execution order revision. Each suggestion is associated with a session identifier and evidence description index. For content involving changes in terminal blocks or channel parameters, the strategy update orchestration subunit attaches a channel change suggestion to the strategy update entry. This suggestion is used in cross-main step reflow to indicate whether the channel mapping table in S110 needs to be changed synchronously. For content involving reference curve selection or version drift, the strategy update orchestration subunit attaches a reference version suggestion to the strategy update entry. This suggestion provides a priority selection prompt for standard threshold loading and historical curve retrieval processing in S210 in cross-main step reflow.Understandably, report generation and knowledge entry extraction operate within a combined framework of program control and data organization. The former focuses on engineering archiving and audit presentation, while the latter addresses the continuous evolution of the rule base and strategy base. To ensure the system can directly reuse update results in subsequent sessions, the strategy update entry structure establishes foreign key relationships with the rule base, strategy base, and model mapping table when it is stored in the database. Simultaneously, a pending review status is written to the interface layer for engineers to review on the HMI. Through the aforementioned report generation, knowledge extraction, and strategy orchestration, a strategy update entry structure is generated. This structure is the output field of this step and will be consumed in S110 of the next process, used to guide the versioning adjustment of the channel mapping table and the parameter revision of the hardware acquisition list. Simultaneously, it is directly invoked as incremental knowledge in the rule base and strategy base inference in S320, forming a closed-loop feedback from suggestion, execution, verification to knowledge accumulation.

[0235] In summary, the technical effects of this step are as follows: by transforming the item-level and chain-level conclusions of the verification input set into archiveable reports and reusable knowledge items, and arranging them into a policy update item structure, a procedural closed loop from execution review to policy evolution is achieved; this closed loop establishes a stable loop between the control layer and the data layer, supporting configuration updates and inference improvements in subsequent sessions.

[0236] In one specific embodiment, S400 involves: establishing a handshake with the PLC command channel based on the suggested instruction set and the comparison benchmark set; setting the action rhythm and inserting a slow splitting and merging verification with a dwell time description; performing node-level verification by comparing the node position and neighborhood shape on the actual segment; performing a secondary comparison by chaining adjacent item verification records; generating a report by grouping and reconstructing the execution chain according to the action direction; and extracting knowledge items of high-frequency, strongly correlated combinations to generate a strategy update item structure that includes new rule suggestions and reference version suggestions. Received from the suggested instruction set output by S330 and the comparison benchmark set output by S210, this section first performs the execution linkage and slow split / slow merge verification processing. The execution item sequence is read from the suggested instruction set and organized into an ordered queue according to execution order prompts, operation mode identifiers, and preconditions. The linkage scheduling subunit establishes a handshake with the PLC command channel and registers a prompt event placeholder with the HMI prompt channel. For each execution item, the action object, action direction, and step description are parsed, and the channel readback guidance and verification node list are accessed to establish a ternary relationship of "action trigger—readback trigger—verification node". During slow split or slow merge, the slow execution control subunit divides the action step into several small steps and inserts a dwell time description between the action trigger and readback trigger of each small step. Let the first step be... There are a total of 10 execution entries. Each small step has a dwell time of [number] seconds. To quantify the quality of readback data during execution, a readback integrity score is defined. Formula ⑦ calculates the first... Completeness of readback for each execution entry: ⑦; Variable and symbol definition: : No. Completeness of readback of each execution entry, dimensionless, range. It is calculated from the "suggested instruction set" and the data returned by the "PLC".

[0237] : Execution entry index, which takes the value of a positive integer and comes from the entry number in the "Execution Linkage Mapping Table".

[0238] : No. The total number of steps contained in an item is dimensionless and is obtained from the dwell time description and action rhythm setting of the item in the "Suggested Instruction Set".

[0239] :right From 1 to Summation, This is the small step index, and its value is a positive integer.

[0240] : Indicator function, takes the value 1 if the condition in parentheses is true, otherwise takes the value 0.

[0241] : No. Article No. The actual readback time of each small step is measured in seconds and originates from the timestamp transmitted by the PLC.

[0242] : No. Article No. The readback deadline for each small step, measured in seconds, is calculated by superimposing the trigger time with the dwell time description in the "suggested instruction set".

[0243] superscript : indicates that it belongs to the first Execution entry.

[0244] superscript : indicates that it belongs to the first A small step.

[0245] Data Source → Metrics → Variable Mapping: Extract the total number of small steps from the "Dwell Time Description" and "Action Rhythm" of each execution entry in the "Suggested Instruction Set". and deadline ; obtained from the readback timestamp returned by the "PLC command channel" .

[0246] Simple numerical example: A slow-execution entry has There are 4 small steps, with the actual readback time being 4 of them. Less than or equal to the deadline ,but .

[0247] Formula ⑦ addresses the practical problem of quantifying the integrity of the readback data in each execution step during the slow separation and merging process, providing a basis for data quality in subsequent verification stages, and avoiding misjudgments due to missing readback data.

[0248] During the queue deployment process, for items requiring the linkage of multiple components, the dependency order is identified based on evidence. Let there be an ordered set of pairs with upstream and downstream dependencies. Define dependency satisfaction Formula ⑧ calculates the dependency satisfaction of the current session: ⑧; Variable and symbol definition: Dependency satisfaction, dimensionless, range It is calculated from the dependencies in the "Suggested Instruction Set" and the "PLC" feedback status.

[0249] : An ordered set of upstream and downstream dependencies, where each element is an ordered pair , indicating an entry The entry can only be triggered after completion. The evidence and preconditions are derived from the "Recommended Instructions Set".

[0250] : Execute the entry index, with a positive integer value representing the upstream and downstream entries respectively.

[0251] : Boolean value, representing an entry The readback trigger condition has been met, originating from the status confirmation signal transmitted back by the "PLC".

[0252] : Boolean value, representing an entry It has been successfully triggered, originating from the execution log sub-cell record.

[0253] : Logical implication symbol, representing "if...then..."; here it means if If true, then It must be true.

[0254] The cardinality (number of elements) of a set is dimensionless.

[0255] curly braces : Represents a set.

[0256] Data Source → Metrics → Variable Mapping: Dependency pairs are extracted from the "Precondition Description" and "Evidence Statement" of each execution entry in the "Suggested Instruction Set". constitute The status indication is obtained from the feedback of the "PLC command channel". ; obtained from the execution log subunit .

[0257] Simple numerical example: Dependency set There are 3 pairs in total. Item 2 is triggered after item 1 is confirmed, and item 3 is triggered after item 2 is confirmed. However, item 5 is not triggered after item 4 is confirmed (due to readback timeout). Therefore, the number of satisfied pairs is 2. .

[0258] Formula ⑧ addresses the practical problem of quantifying the degree of coordination between upstream and downstream dependencies in the execution chain, identifying execution interruptions caused by dependency breakage, and providing a basis for judgment in exception handling. The output of this section is an execution linkage mapping table, containing the mapping relationships of entries—action objects—trigger conditions—readback channels—review nodes—reference curve links—abnormal branches, which are directly consumed by the node actions and real-time readback process of S420.

[0259] Following the execution of the linkage mapping table, this section performs node-level verification and secondary comparison. The node readback acquisition subunit activates the readback channel line by line according to the mapping table, acquiring multi-source data segments within the readback period. The time alignment and node extraction subunit uses the previously set timestamp format and sampling interval to locate the node occurrence time within each segment based on the review node list. The node-level verification subunit performs a verification at the item-level granularity, recording the actual node time. The corresponding local waveform window is compared with the reference curve window in terms of shape. Neighborhood shape similarity is defined. Formula 9 calculates the morphological similarity between the actual segment and the reference curve within the node neighborhood: ⑨; Variable and symbol definition: Neighborhood morphological similarity (Pearson correlation coefficient), dimensionless, range of values. It is calculated from the data of the "comparison benchmark set" and the "node readback acquisition subunit".

[0260] The number of sampling points within the window, dimensionless, determined by the window's half-width in the "Execution Linkage Mapping Table". With sampling interval calculate: .

[0261] :right From 1 to Summation, This is the index of the sampling point within the window, and its value is a positive integer.

[0262] : Reference curve window The amplitude of each sampling point, with dimensions depending on the physical quantity (displacement in mm, pressure in MPa), is derived from the reference curve in the "comparison benchmark set".

[0263] The mean value of the sampling points within the reference curve window, with the same dimensions. The calculation formula is: .

[0264] The actual reread segment window contains the first segment. The amplitude at each sampling point, with the same dimensions The data originates from multi-source data fragments collected by the "node back-read acquisition subunit".

[0265] The mean of the sampling points within the actual segment window, with the same dimensions. The calculation formula is: .

[0266] : Square root function.

[0267] superscript : indicates the first in the window One sampling point.

[0268] Data Source → Indicators → Variable Mapping: Obtained from the reference curve window data extracted from the "Benchmark Set". The actual read-back segments collected by the "node read-back acquisition subunit" are obtained. .

[0269] Simple numerical example: sampling points within the window Reference curve mean Actual segment mean The numerator is calculated to be 5.2, and the denominator is... ,but .

[0270] Formula 9 addresses the practical problem of quantifying the morphological consistency between the actual waveform and the reference waveform in the node neighborhood, providing objective numerical basis for node-level verification, and avoiding the limitations of relying solely on threshold judgments.

[0271] When multiple channels provide the same node positioning basis (such as the on / off edge and coil current peak), the node-level verification subunit records the comparison results of each channel side-by-side and, based on the evidence, adopts a consistent priority strategy to give an item-level verification conclusion. Further, the secondary comparison subunit performs vertical verification at the action chain level: it concatenates the node-level verification records of multiple adjacent items according to the operation mode and action direction, extracting the "start-trigger-response-stability" node chain. Chain-level coordination is defined. Formula 10 calculates the timing coordination degree of the node chain: ⑩; Variable and symbol definition: Chain-level coordination degree, dimensionless, range of values. The actual node time output by the "node-level verification subunit" is calculated from the standard time sequence in the "comparison benchmark set".

[0272] The total number of nodes in the node chain, dimensionless, derived from the concatenation result of the verification node list of adjacent entries in the "Execution Linkage Mapping Table".

[0273] :right From 1 to Summation, This is the node interval index, and its value is a positive integer.

[0274] : Absolute value symbol.

[0275] : The first in the node chain The actual duration of each interval, measured in seconds, is calculated from the difference in the actual occurrence times of the nodes: ,in For the first The actual occurrence time of each node.

[0276] : No. The standard duration of each interval, measured in seconds, is derived from the standard time series in the "comparison benchmark set".

[0277] : A very small positive number, with the dimension of "seconds". To prevent division by zero, it is usually taken as . .

[0278] superscript : indicates the first Each node interval.

[0279] Data source → Metrics → Variable mapping: Obtained from the actual node time output by the "node-level verification subunit". Extracted from the standard time series in the "comparison benchmark set" .

[0280] Simple numerical example: Node chain contains There are 1 node, with a total of 3 intervals. The actual intervals are as follows: Standard interval Then calculate the absolute value of each relative error: , , , average , .

[0281] Formula 10 addresses the practical problem of quantifying the overall coordination of the temporal relationships between multiple nodes in an action chain, achieving an upgraded judgment from single-point verification to chain-wide collaboration. Values ​​below a threshold (e.g., 0.8) are marked with a chain-level inconsistency flag. The output of this section is the verification input set, containing item-level verification records, chain-level secondary comparison records, anomaly and manually annotated records, reference curve links, and evidence explanation indexes. This set is directly consumed by S430's report generation and knowledge entry extraction.

[0282] Following the validation input set, this section performs report generation and knowledge item extraction. The report generation subunit groups items within the session according to action direction and operation method, reconstructs the execution chain in natural order of item numbers, and incorporates the node-level validation records and reference curve comparison view links. The knowledge item extraction subunit traverses the validation input set, extracting high-frequency combinations with clear causal relationships as standardized knowledge items. The effective confidence level of each knowledge item is defined. ,formula Calculate candidate knowledge entries Confidence level: ; Variable and symbol definition: Candidate knowledge entries Confidence level, dimensionless, range It is calculated from the "Equipment File" and the "Node Deviation List".

[0283] Candidate knowledge entry identifiers are derived from the combination of "fragment-node-deviation direction-execution entry" in the "verification input set".

[0284] :combination The number of times it appears in the history of sessions, measured in "times", is obtained from the history verification input set in the "Device Archive".

[0285] Total number of historical sessions, measured in "number", sourced from "device archives".

[0286] Causal attenuation coefficient, dimensionless, with a value of 2.

[0287] Clarity of causal relationship, dimensionless, range of values. It is calculated based on the consistency of the weight labels and the interpretability of the deviation direction in the "Node Deviation List".

[0288] Multiplication operator.

[0289] superscript : Indicates that it belongs to the candidate knowledge item .

[0290] Data Source → Metrics → Variable Mapping: Statistics from "Equipment Files" and The consistency calculation is based on the multi-channel priority records and deviation direction in the "Node Deviation List". .

[0291] Simple numerical example: a certain combination Number of occurrences Total number of historical sessions The frequency is 0.8; the clarity of causal relationships. ,but .

[0292] formula The practical problem addressed is to quantify the reliability of candidate knowledge items, filter out low-confidence items, and ensure that the extracted knowledge has statistical significance and causal interpretability.

[0293] The strategy update orchestration subunit maps newly generated knowledge entries and key conclusions from the report to executable strategy update entries. Let the strategy update entry structure be... This includes suggestions for new rules and suggested versions. (Formula) Define the priority weights of policy update entries. : ; Variable and symbol definition: Priority weights for policy update entries, dimensionless, with a range of values. ; From the formula The confidence level is calculated using the duplicate identifier from formula ②.

[0294] :Same formula definition.

[0295] :theme Repeating identifier, dimensionless, range of values This is derived from the output of formula ② in step S300 (i.e., the repeat identifier field in the operating context).

[0296] : The maximum value of all topic repetition identifiers in the current system, dimensionless, derived from all steps in S300. The maximum value.

[0297] Action direction indicator and operation mode indicator are defined in the same way as formula ②.

[0298] superscript : Indicates belonging to the topic .

[0299] Simple numerical example: Confidence level of a knowledge item Duplicate tags for corresponding topics The system's maximum duplicate identifier ,but .

[0300] formula The practical problem addressed is to dynamically calculate the priority of policy update items by combining the confidence level of knowledge items with the historical repetition frequency of abnormal patterns, so that high-frequency, high-confidence knowledge is preferentially fed back into the rule base and policy base. The output of this section is the policy update item structure, including new rule suggestions and reference version suggestions, which are directly consumed by the next round of processes S110 (channel mapping table versioning adjustment) and S320 (rule base and policy base inference).

[0301] This section summarizes the technical effects: By implementing coordinated execution and slow decomposition / merging verification, node-level verification and secondary comparison, report generation and knowledge item extraction, a closed-loop feedback loop is formed from execution and verification to knowledge accumulation. Formulas ⑨ and ⑩ achieve dual verification from two dimensions: node neighborhood morphology and chain-level temporal coordination, respectively. and By introducing a confidence and priority quantification mechanism, which differs from the conventional single threshold judgment, the robustness of verification and the adaptability of policy updates are significantly improved.

[0302] Figure 2 This is a displacement-stroke curve used for comparing the position of the limit switch trigger point. The displacement-stroke curve describes the linear displacement of the moving contact (or piston rod) over time during the opening or closing process of the hydraulic operating mechanism. This curve is the direct basis for determining whether the limit switch is triggered in the correct mechanical position. Figure 2As shown, the displacement-stroke curve illustrates the linear displacement of the moving contact of the hydraulic operating mechanism over time during the opening and closing process. The standard trigger node (from the comparison reference set) is marked with "★", and the actual trigger node (from the current measurement feature set) is marked with "●". By comparing the horizontal coordinate positions of the two, it is possible to intuitively determine whether there is a positional deviation Δ of the limit switch. L = L actual L mid. If the actual node is to the right of the standard node and Δ L Exceeding the permitted range ( δ If the maximum value is ±0.5mm, it can be determined as positional lag; otherwise, it is positional lead. Furthermore, the local waveform window within the node's neighborhood is compared with the actual readback curve in terms of morphological similarity. r Calculate (Formula 9), if r A low value (e.g., <0.7) indicates possible mechanical jamming, sensor loosening, or signal interference. Figure 2 This invention embodies the dual judgment mechanism of "position comparison + neighborhood morphology verification", which makes the mechanical position adjustment of the limit switch based on evidence and avoids misoperation based solely on threshold judgment.

[0303] Figure 3 This is a hydraulic pressure curve used to compare the action sequence with the pressure inflection point. The hydraulic pressure curve reflects the change in oil pressure within the accumulator or working cylinder over time and is a key indicator for judging the energy reserve, smoothness of action, and opening / closing capability of the operating mechanism. The inflection point, rate of descent, and recovery time of the pressure curve are directly related to the rationality of the limit switch timing. For example... Figure 3 As shown, this hydraulic pressure curve reflects the change in oil pressure within the accumulator or working cylinder over time, serving as a crucial basis for determining the energy reserve and action sequence of the operating mechanism. The pressure inflection point (the point where the first derivative crosses zero) in the curve corresponds to the transition moment when the hydraulic system changes from pressure build-up to pressure release. This is achieved by calculating the actual inflection point and adjacent nodes (such as the action initiation node). t The time interval Δ between cmd and the limit switch trigger node trip) t actual, and compared with the standard timing Δ t By comparing the std values, the timing deviation Δ is obtained. t dev. If Δ t If dev is positive and exceeds the limit, it indicates that the action is slowing down, which may be due to insufficient nitrogen pressure or valve core sticking. If the curve recovery segment shows oscillation, it indicates a sealing or venting problem. Figure 3 It is also used for segmenting operating conditions, at the moment when the pressure recovers to 95% of its steady state. t The end point serves as the endpoint of the segment and a chain-level coordination check, thereby linking hydraulic dynamics with mechanical stroke analysis to achieve intelligent diagnosis from a single parameter to multi-source collaboration.

Claims

1. A method for intelligent adjustment and testing of limit switches in a hydraulic operating mechanism, characterized in that... Includes the following steps: S1: Obtain the channel mapping table and acquisition hardware list, and generate the measurement feature set; S2: Based on the model mapping table, equipment files and the measurement feature set, generate a comparison benchmark set and a node deviation list; S3: Obtain the node deviation list and the measurement feature set, and generate a suggested instruction set; S4: Based on the suggested instruction set and the comparison benchmark set, generate a strategy update entry structure that includes suggestions for new rules and suggestions for reference versions.

2. The method of claim 1, wherein: In S1: the channel mapping table is used to describe the wiring relationship between the hydraulic operating mechanism and the limit switch status, hydraulic pressure, displacement stroke, opening and closing sequence, and environmental quantities, as well as the terminal number, range and signal type; and establishes a one-to-one correspondence of identification items through the model dimension and the terminal dimension. The hardware list includes the host model of the field acquisition unit, acquisition board specifications, number of channels and channel resolution, supported input types, calibration certificate number, firmware version and synchronization bus type.

3. The method of claim 2, wherein: In S1: Perform time base configuration, channel drift correction, sensor data acquisition, multi-source alignment, node event extraction, and working condition segmentation processing to generate a measurement feature set containing a node event table and a working condition segment table; The time base configuration includes setting a unified start time and sampling interval; specifically as follows: Let the globally unified timeline be... The master clock source time is calculated using a master-slave synchronization strategy. With the device's local time offset And will correct from device time to This ensures that all acquisition channels are started on the same time reference and that the time interval between adjacent sampling points is strictly equal, providing a unified time scale basis for subsequent multi-source alignment; The channel drift correction includes measuring zero-point offset and gain offset; specifically as follows: Adopting linear correction model ; in: For the actual physical quantity input, These are the raw values ​​directly read by the acquisition device. This is the gain offset correction factor. This is the zero-point offset correction factor; Determine the zero-point offset by inputting a zero-value standard signal under static closed-loop or known steady-state operating conditions. Calculate the gain offset by inputting a full-scale standard signal. ; For channels exhibiting temperature drift, temperature drift compensation is introduced: ; in: The current temperature. Reference temperature Gain coefficient under, As a temperature drift compensation factor, Reference temperature; also marked with drift sensitivity level. 0 indicates insensitive, and 2 indicates highly sensitive, used to dynamically adjust the correction frequency; The sensor data acquisition includes the acquisition of key quantities such as the on / off state of the start limit switch and hydraulic pressure. The multi-source alignment refers to generating alignment points through interpolation resampling under a unified time reference; specifically as follows: Under a unified time reference, a high-frequency channel is set up. In the global timeline There are already sampled values ​​above. Low-frequency channel The original sampling time is The corresponding sample value is ; The linear interpolation formula is used to calculate the estimated value of the low frequency channel at the target alignment time . ； wherein: is the interpolated aligned point value; represents the slope between two points; For incomplete segments, discard them directly and write a missing flag to generate a synchronous data frame in which all channel samples are complete at the same time. The node event extraction refers to locating edge nodes on the on / off trajectory; specifically as follows: The limit switch on / off signal channel is a discrete switch signal, and its edge node is defined as the moment when the signal jumps from "0" to "1" or from "1" to "0". Let the on-off quantity signal be At the sampling time and , if , it is determined that there is an edge node, and the node occurrence time is taken as ; the node is marked as "travel switch trigger node"; At the same time, the rising edge of the coil current channel, i.e. the starting moment , can be identified as the "action starting node"; All nodes are accompanied by a quality label. Where 0 represents low quality, 1 represents average quality, and 2 represents high quality; The operating condition segmentation process divides the circuit breaker tripping and closing execution segments using node events as anchor points; specifically as follows: Using key nodes in the node event extraction results as anchor points, the working condition segments are segmented. Define the action start node as The limit switch trigger node is The action termination node is ;but: Opening preparation fragment: ; Switching execution fragment: ; Stable segment of opening and closing ; The closing process is similar; Wherein: is the end time of the previous action or the acquisition start time; each segment is attached with a model identifier , an action direction identifier and an operation mode identifier for subsequent differential processing.

4. The method of claim 3, wherein: In S1, the process of generating the measurement feature set, which includes the node event table and the working condition segment table, also includes: a1: Obtain the channel mapping table and acquisition hardware list, perform time base configuration and channel drift correction processing, and obtain the acquisition configuration set; a2: Extract channel identifiers and timing identifiers from the acquisition configuration set, perform sensor data acquisition and multi-source alignment processing, and generate synchronous data frames; a3: Perform node event extraction and condition segmentation processing on the synchronous data frame to generate a measurement feature set.

5. The intelligent adjustment and testing method for the limit switch of a hydraulic operating mechanism according to claim 4, characterized in that: The time reference configuration includes: setting a unified start time, sampling interval, and trigger priority, and establishing a single time reference through a master-slave synchronization strategy; specifically as follows: The master-slave synchronization strategy is implemented through a high-precision synchronization bus; the master clock source broadcasts a synchronization pulse, and the slave device latches its local counter upon receiving the pulse to calculate the offset. And adjust your local time; Trigger priority is divided into three levels: Priority 1 is the on / off state of the limit switch and hydraulic pressure. Priority 2 is based on coil current and displacement stroke. Priority 3 is environmental quantities and vibration; Set a unified start time The sampling interval is the first whole second after the arrival of the synchronization pulse. Based on the signal bandwidth setting, this mechanism ensures the initial phase consistency of data acquired from multiple boards, laying the foundation for subsequent alignment. The channel drift correction includes: measuring zero-point offset and gain offset, and marking the drift sensitivity level of variable-temperature drift channels; specifically as follows: In a linear correction model On the basis of the drift-sensitive level markings: Define drift sensitivity level Historical drift rate Decide: like ,but ; If then ; If then ; wherein, and are gain correction coefficients at temperatures and respectively. The sensor data acquisition includes: prioritizing the acquisition of high-frequency channels for key quantities such as limit switch on / off states and hydraulic pressure. The multi-source alignment includes: generating alignment points according to the target sampling time under a unified time reference, and organizing the channel samples at the same time by interpolation resampling and discarding incomplete segments; specifically as follows: The original sampling times of the high-frequency channel and the low-frequency channel are different; a linear interpolation formula is used to resample the low-frequency channel to the time axis of the high-frequency channel. At the same time, the complete segment judgment condition is defined: assuming that the total sampling point number in the segment is , and the missing flag number is , If Where: is a threshold value, then it is determined as an incomplete segment and discarded; otherwise, the missing points are filled by interpolation; the discarded segment will trigger an exception record, written into the collection log, and reduced weight in the subsequent node event extraction; The node event extraction includes: identifying the action initiation node and the limit switch trigger node, and locating the edge node on the on / off quantity track; specifically as follows: The identification of the start point of the action is based on the rising edge of the coil current path: Let the coil current signal be... Its first difference ,when , If the preset current change threshold is met and the duration exceeds two sampling points, then a determination is made. This is the starting node of the action; The identification of the limit switch trigger node is based on the edge of the on / off channel: when At that time, the judgment For triggering nodes; for multi-channel redundancy, "OR" logic is used: the occurrence of an edge on any channel triggers a node event, and the numbers of all participating channels are recorded in the node table; The operational segment segmentation includes: using the results of node event extraction as anchor points, dividing the circuit breaker preparation segment into a circuit breaker closing execution segment; specifically as follows: Define the general segmentation rules for action chains: Assume a complete action cycle includes: preparation phase Execution phase Stable phase ;in: To control the timing of command issuance, The trigger time of the limit switch. This refers to the moment when the hydraulic pressure recovers to 95% of its steady-state range; the corresponding tripping preparation segment. The closing execution segment corresponds to Fragment length Used for subsequent deviation density calculation; each segment is written to the database with the following fields: segment ID, action direction, start and end node index, and quality level.

6. The intelligent adjustment method for the limit switch of a hydraulic operating mechanism according to claim 5, characterized in that: In S1: the model mapping table is used to give the unified identification relationship of the hydraulic operating mechanism in different manufacturers, different specifications and different wiring terminals, including model identification, terminal number, measuring point name, signal type, range, action direction identification and adaptation status. Among them: "Model Identifier" is the unique code of the device, "Terminal Number" corresponds to the actual terminal block position, "Measurement Point Name" is the physical quantity description, "Signal Standard" indicates the electrical type, "Measurement Range" is the sensor measurement range, "Action Direction Identifier" distinguishes between opening / closing / general, and "Adaptation Status" indicates whether the mapping is currently valid; Equipment files provide basic information, standard threshold version records, historical curve storage location information, and operating environment descriptions for the same device during various maintenance and testing processes. They cover maintenance dates, executors, operating condition descriptions, device serial numbers, hardware change records, and data acquisition paths.

7. The method of claim 6, wherein: In S2, standard threshold loading, historical curve retrieval, index mapping, label binding, position comparison and time series comparison processing are performed to generate a comparison benchmark set and a list of node deviations. The process of generating the benchmark set and the list of node deviations includes: b1: Obtain the model mapping table and equipment file, perform standard threshold loading and historical curve retrieval processing, and obtain the comparison benchmark set; b2: Extract the standard node positions and standard time series from the comparison benchmark set, perform index mapping and label binding processing, and generate a benchmark index table; b3: Perform position comparison and time series comparison processing on the measurement feature set and the benchmark index table to generate a list of node deviations; The standard threshold loading includes: loading the limit switch on / off threshold and the hydraulic pressure reference threshold, and selecting the corresponding version according to the standard threshold version; The historical curve retrieval includes: retrieving the opening process curve and the closing process curve, and prioritizing the curves with the most recent acquisition time and marked as qualified as the control sample; specifically as follows: Historical curve retrieval refers to retrieving qualified process curves of the same device under the same or similar operating conditions from equipment records as a reference sample; The system displays three historical closing curves: Curve A is qualified; Curve B is qualified; Curve C is unqualified. Curve A, which was collected most recently and marked as qualified, was selected as the control sample. Let the set of historical curves be Each curve with a time stamp , a quality label , a working condition description ; The search rule is: select i.e. the one with the largest timestamp in the eligible curves; If the search is unsuccessful, an alternative rule is triggered: select a representative curve with a higher quality grade within the same model range, and provide an alternative explanation; The standard node positions and standard timing sequences include: the candidate node locations of the action initiation node and the limit switch trigger node on the historical curve entries; specifically as follows: Standard node position and standard timing refer to the spatiotemporal coordinates of nodes extracted from historical curve entries in the comparison benchmark set and verified by standard thresholds; on a qualified closing curve, the system automatically locates the "action start node". (Rising edge of coil current), "Limit switch trigger node" (Switchover), "Pressure Inflection Point" (The pressure derivative crosses zero, etc.); Let the set of standard node positions be wherein, , is a standard time, is a node type; Standard timing is defined as the interval between adjacent nodes: ; represents the standard interval from the start of an action to the triggering of a switch; these standard nodes are written into a reference index table for later comparison; The tag binding includes binding action direction tags and quality level tags, and attaching source tags and adaptation tags; specifically as follows: Tag binding refers to attaching multiple tags to each index entry in the base index table to support multi-dimensional retrieval and priority decision-making; each node index entry includes: action direction tags. Quality grade label (2 is the highest), Source Tag Adaptive tags After the tag is bound, the comparison and judgment unit can determine the comparison based on... Prioritize high-reliability nodes, based on Filter out index items that match the current operation method to avoid false matches; The position comparison is used to describe the deviation of the actual node from the reference range, and to compare the occurrence time with the channel context; specifically as follows: Position comparison is used to describe the deviation between the actual node's landing point in physical quantities such as displacement stroke and hydraulic pressure and the standard reference range; the position of the actual closing trigger node on the displacement curve is... The standard reference range is given by the benchmark index table. The reference median is ; The formula for calculating position deviation is: ; in: : position deviation, positive value means actual node position is larger, negative value means smaller; : the displacement value corresponding to the moment when the actual node appears on the displacement stroke curve; : median of the standard reference range; At the same time, the channel context is compared: exceeds the allowable deviation , the node is marked as position abnormal, and the deviation and the associated channel identifier are recorded; The time-series comparison is used to describe the difference between the actual interval and the standard time sequence, and to perform neighborhood reinforcement on the actual nodes identified as low-quality nodes; specifically as follows: Temporal comparison is used to describe the difference between the time interval between actual adjacent nodes and the standard temporal reference interval, and to perform neighborhood reinforcement on actual nodes with low-quality identification; let the adjacent nodes in the actual node sequence be... and The timestamps are respectively and The actual interval is: ； The standard timing reference interval is From the reference index table, the timing offset is: ； in: : actual neighbor time interval; : the first and second actual nodes; and : the time of appearance of the first and second actual nodes; and : the time of appearance of the first and second actual nodes. : standard timing reference interval derived from qualified history curve; : timing deviation, positive value means actual interval is larger, action is slower; negative value means smaller, action is faster; For low-quality labels For actual nodes, perform neighborhood reinforcement: Take the nodes before and after this node Each sampling point constitutes a neighborhood window. Within this window, the node position is re-estimated using other high-quality channels. If the deviation between the estimated result and the original node position is less than a threshold, the corrected time is used. Replace the original Otherwise, retain the original node and add a low-quality warning; low-quality nodes have a deviation from the original interval. After neighborhood reinforcement, it is corrected to It has entered an acceptable range.

8. The method of claim 7, wherein: In S3: abnormal fragment identification, working condition aggregation, rule base and strategy base reasoning, model adaptation and execution mapping processing are performed to generate a suggested instruction set; The process of generating a suggested instruction set includes: c1: Obtain the node deviation list and measurement feature set, perform abnormal segment identification and working condition aggregation processing to obtain the working condition context; c2: Extract node type and deviation direction from the node deviation list and operating condition context, perform rule base and strategy base reasoning processing, and generate a set of suggested candidates; c3: Performs model adaptation and execution mapping processing on the proposed candidate set to generate a proposed instruction set.

9. The intelligent adjustment method for the limit switch of a hydraulic operating mechanism according to claim 8, characterized in that: The abnormal segment identification includes: calculating the deviation density and verifying the continuity of the working condition segment table; specifically as follows: Abnormal segment identification first verifies the continuity of the working condition segment table, checking for duplicate time markers, unreasonable spans, sudden drops in channel coverage, etc., and then calculates the deviation density of each segment based on the node deviation list; assuming a certain working condition segment The duration is The number of node deviation entries contained in this segment is Then the deviation density is defined as: ； in: : fragment Deviation density; : belongs to segment : total number of entries in the node deviation list : Duration of the fragment, computed from the start-stop node time difference: ; like , If a preset threshold is met, the segment is marked as an abnormal segment; the continuity check also checks whether there is a time reversal at the start and end nodes. The concentration of missing markers is analyzed, and if any are found, they are marked as suspicious fragment candidates; the abnormal fragment identification unit outputs a set of abnormal fragments. Each anomalous segment is accompanied by the dominant deviation direction. and the combination of key node types involved; The operational condition aggregation includes: using the action direction identifier and operation mode identifier as primary keys, grouping the abnormal segment set, node deviation list, and relevant entries in the node event table into corresponding topics; specifically as follows: Working conditions are aggregated and identified by the direction of action. and operation method identification As a composite primary key, the set of abnormal fragments Node Deviation List and node event table The relevant entries are assigned to the corresponding topics; the aggregation process generates a working condition context structure, and each topic contains: topic ID, fragment list, dominant deviation summary, duplicate identifier, etc. The rule base is based on a set of structured rules, and its rule antecedents consist of a combination of node types, the dominant deviation direction, and the fragment category; specifically as follows: The rule base is a knowledge base based on a set of structured rules, the antecedent of each rule is composed of node type combination, dominant bias direction and fragment category, and the consequent is an executable adjustment suggestion primitive; the form of the rule is: ​ ； in: : node type combination; : set of specific node types; : Dominant deviation direction, value ; : fragment category, value ; : adjustment suggestion primitive; The strategy library is an experience-based strategy library that generates candidate adjustment patterns based on similar topics and duplicate identifiers; specifically as follows: The policy library is a collection of historical experience policies. When the rule library has no complete matching entry, the policy library generates a candidate adjustment mode according to a similar topic and a repeated identification. Each policy in the policy library includes a similarity measure function and a migration rule; Let the current working condition theme description be , the historical policy entry be , and the similarity calculation formula be: ； in: : indicator function, taking 1 if the condition is true, 0 otherwise; : weight coefficient; : current set of abnormal segments; The set of anomalous fragments recorded in historical strategies; The historical strategy with the highest similarity is selected, and its adjustment mode is transferred to the current scenario to generate candidate adjustment modes. If a repetition is marked as "high-frequency repetition", the priority of the strategy is increased. The strategy base inference outputs a candidate set, with each candidate accompanied by a similarity score and evidence. The model adaptation includes mapping abstract objects to specific wiring terminals, and providing corresponding contact pairs and fixing methods according to the terminal mapping specifications; specifically as follows: Model adaptation maps the abstract adjustment objects in the suggested candidate set to the specific wiring terminals, sensing channels and mechanical adjustment positions of the current device. Let the abstract object in the suggestion candidate be , and there is a mapping function in the model mapping table: ； in: : model identifier of the current device; : specific terminal number; : fixed manner; : Adjusted step description; The model adaptation unit automatically performs mapping based on the current device's model identifier. If a mapping is missing, it suspends the candidate and prompts for an update to the model mapping table. The execution mapping includes mapping to PLC automatic trigger instructions or HMI interactive prompt instructions, and splitting a candidate into a set of ordered execution entries; specifically as follows: Execution mapping translates the specific adjustment actions after model adaptation into instruction entries that can be recognized by the execution layer, including automatic trigger instructions for PLCs and interactive prompt instructions for HMIs, and splits a candidate into a set of ordered execution entries; let a candidate... The specific action sequence is obtained after model adaptation. The execution mapping unit transforms it into: For auto-triggering class: generate PLC instruction ; For interaction prompt type: generate HMI event ; Each execution entry includes: action object, action direction, step description, channel readback guidance, and a list of verification nodes; a candidate for "adjusting the position of the closing limit switch" is split into three ordered execution entries: Send a PLC command to put the mechanism into slow closing mode; The HMI prompts the operator to loosen the limit switch fixing screws; The PLC drives the slow closing action, the readback switch triggers the displacement, and the HMI displays the current deviation value; The execution mapping also includes readback trigger conditions and exception interruption conditions to ensure that the execution process is safe and controllable.

10. The method of claim 9, wherein: In S4: the execution linkage for establishing a handshake with the PLC command channel, slow splitting and slow merging review, node-level verification, secondary comparison, report generation, and knowledge item extraction processing are performed to generate a strategy update item structure that includes suggestions for new rules and suggestions for reference versions. The execution linkage is to establish a handshake with the PLC command channel; The slow separation and slow combination verification refers to setting the action rhythm and inserting a description of the dwell time; The node-level verification refers to verifying the node position against the neighborhood morphology on the actual fragment; The secondary comparison refers to performing chain-level verification by connecting adjacent entry verification records; The report is generated as a reconstruction of the execution chain grouped by action direction; specifically as follows: The report generation sub-units are grouped by action direction, reconstruct the execution chain in natural order of item numbers, and incorporate node-level verification records and reference curve comparison views; the report generation process is divided into three layers: the first layer is based on action direction. Grouping; the second level involves numbering items within each group. The execution chain is reconstructed in a natural order; the third layer imports a link between node-level verification records and reference curves under each entry; the execution chain is then defined. The expression is: ； in: : the complete execution chain structure after reconstruction; : action direction identifier, taking value of "open" or "close"; : number of execution entries belonging to the action direction ; : the first execution entry; : the second execution entry;​ : The node level verification record of the entry, containing actual node time, deviation amount, and verification conclusion; : reference curve and actual segment comparison view link; The left side is the "Close" group, which contains multiple entries; the right side is the "Open" group, which contains multiple entries; each entry expands to display verification records and curve comparison links, ultimately generating a structured report; The knowledge item extraction process involves extracting knowledge items that are high-frequency and strongly correlated; specifically as follows: The knowledge item extraction sub-unit traverses and verifies the input set, and extracts standardized knowledge items from combinations that appear frequently, have clear causal relationships, or are strongly related to a specific model. Set a knowledge item in the form of: ； in: : trigger condition description; : involves a set of objects; : recommended execution order hints; : Tag set, including easy-mix node identification, common exception tag, etc. : evidence explanation index, points to entry number in original validation record; Frequency calculations use sliding window statistics: Let the total number of historical sessions be A certain group The number of sessions that occurred was The frequency of occurrence ;when , When the frequency threshold is defined, and the causal relationship is clear and there is a consistent correspondence between the deviation direction and the adjustment effect, it is extracted as a standard knowledge item. The generated strategy update entry structure includes new rule suggestions and reference version suggestions, which are used to achieve dynamic updates to the rule base and strategy base.

11. The method of claim 10, wherein: In S4, the process of generating a strategy update entry structure that includes new rule suggestions and reference version suggestions also includes: d1: Obtain the suggested instruction set and the comparison benchmark set, perform execution linkage and slow splitting and slow merging review processing, and obtain the execution linkage mapping table; d2: Extract node actions and real-time readback from the execution linkage mapping table, perform node-level verification and secondary comparison processing, and generate a verification input set; d3: Perform report generation and knowledge entry extraction processing on the validation input set, and update the entry structure using the generation strategy; The execution linkage includes: establishing an execution session, establishing a handshake with the PLC command channel, and registering a prompt event placeholder with the HMI prompt channel; The slow division and slow combination verification includes: setting the action rhythm, dividing the action steps into several small steps, and inserting a description of the dwell time; The node-level verification includes: comparing the node positions on the actual fragment, and comparing the corresponding node positions and neighborhood morphology on the reference curve with those on the actual fragment; The secondary comparison includes: concatenating the node-level verification records of multiple adjacent entries according to the operation method and action direction, and extracting the node chain for verification; The report generation includes: grouping by action direction, reconstructing the execution chain in the natural order of item numbers, and importing node-level verification records and reference curve comparison view links; The knowledge item extraction includes: extracting standardized knowledge items, and extracting combinations that appear frequently, have clear causal relationships, or are strongly correlated with specific models; In d2, node-level verification includes: The node readback acquisition subunit activates the readback channel one by one according to the mapping table, acquires multi-source data segments within the readback period, and attaches the entry number and operation mode identifier to ensure that the binding relationship between data and action is clear and stable. Under the slow separation and slow combination process, the node back-read acquisition subunit follows the dwell time description and back-read trigger conditions in the mapping table, and forms a corresponding small segment back-read set after each small step of action, covering limit switch on / off, hydraulic pressure, displacement stroke, coil current, environment and vibration. The time alignment and node extraction subunit performs unified time reference alignment on multi-source readback segments under the same session identifier, adopts the previously set timestamp format and sampling interval, and locates the node occurrence time in each segment based on the review node list; The node-level verification subunit performs verification at the item-level granularity. The verification input includes the actual node time, channel context, and reference curve link referenced by the item. The verification method is to compare the corresponding node positions and neighborhood morphology on the reference curve and the actual segment, and record the verification traces of item-node-comparison window; as detailed below: The node-level verification subunit performs verification at the item-level granularity. The verification method involves comparing the corresponding node positions and neighborhood morphology on the reference curve and the actual segment, and recording the verification traces of "item-node-comparison window". The verification subunit will record the actual node time. Corresponding local waveform window , The window width is half the width of the reference curve. Perform morphological comparison; verify the trace records as follows: ； in: : item number; : node type; : actual node appearance time; : position or timing offset value; : neighborhood morphological similarity, computed as the correlation coefficient of the reference curve and the actual segment within the window: ； : list of channel identities participating in positioning; When multiple channels provide the same node location information, the subunit records the comparison results of each channel in parallel, and adopts a consistent priority strategy based on the evidence embedded in the execution linkage mapping table to give an item-level verification conclusion. When multiple channels provide the same node location information, the node-level verification subunit records the comparison results of each channel in parallel and gives the entry-level verification conclusion based on the evidence description embedded in the execution linkage mapping table using a consistent priority strategy.

12. The intelligent adjustment method for the limit switch of a hydraulic operating mechanism according to claim 11, characterized in that: In d2, the secondary contrast includes: The secondary comparison subunit performs a longitudinal check at the action chain level: Under the same session identifier, the node-level verification records of multiple adjacent entries are concatenated according to the operation method and action direction. The node chain of start-trigger-response-stable is extracted, and the temporal relationship and morphological trend between adjacent nodes are checked to see if they are consistent with the reference temporal sequence and reference morphology. When it is found that all item-level verifications are acceptable but the chain-level verification shows inconsistency, the secondary comparison sub-unit marks the corresponding item group with a chain-level inconsistency mark, and includes links involving item range, node type combination and reference curve in the output, prompting the subsequent report generation stage to display in a layered manner.

13. The method of claim 12, wherein: It also includes an interface verification subunit. The interface verification subunit presents a window view of node-level verification and secondary comparison in the HMI overlay view with the item number as the index. Engineers can view the actual segments, reference curves and evidence descriptions in the same screen, and add manual labels or notes when necessary. The interface verification subunit writes the manually supplemented content back to the session layer record. If an entry is not read back within the specified time window or the read back quality is marked as low, the node read back acquisition subunit marks the entry as read back abnormal and automatically excludes its dominant influence on the chain-level conclusion during the second comparison. At the same time, the cause of the abnormality and the channel status are written into the output.

Images