A method and system for condition-based and layered status evaluation and self-calibration of port cranes

Abstract

This invention discloses a hierarchical status evaluation and self-calibration method and system for port cranes under different operating conditions, relating to the field of equipment health monitoring. The method includes: mapping the evaluation object to hierarchical evaluation units according to equipment, mechanism, measuring points, and operating condition sub-intervals; performing dual-channel judgment by concurrently calling standard reference thresholds and historical statistical baselines; identifying conflict types when the results of the two channels contradict each other or the initial judgment difference of multiple features exceeds the threshold, and selecting additional diagnostics, cross-modal consistency, and abnormal duration as evidence for weighted resolution, outputting the status level and abnormal confidence; establishing a new configuration version and constructing candidate baselines when component replacement or continuous drift occurs, switching models after backtracking verification and a recalibration observation period; for operating conditions with insufficient samples, using adjacent bin interpolation, similar operating condition migration, and threshold backtracking to progressively complete the evaluation, ensuring the continuity of the evaluation. This invention can effectively reduce the risk of misjudgment and omission under complex operating conditions, improving the accuracy and long-term effectiveness of status evaluation.

Description

Technical Field

[0001] This invention relates to the field of port crane equipment health monitoring and intelligent operation and maintenance technology, specifically to a method and system for port crane condition-based hierarchical status evaluation and self-calibration. Background Technology

[0002] Port cranes are core operational equipment in port logistics hubs, and their operational reliability directly affects the overall throughput capacity and production safety of the terminal. Typical port cranes, such as quay container cranes and gantry cranes, frequently perform heavy-load start-stop, high-speed speed changes, and multi-mechanism linkage actions during the coordinated operation of hoisting, trolley, crane, slewing, and luffing mechanisms. They generally exhibit technical characteristics such as drastic load changes, variable operating speeds, large spatial spans, and strong coupling of input conditions. These characteristics lead to highly differentiated dynamic responses of the equipment under different operating scenarios, and the distribution of monitoring signal characteristics corresponding to their normal state itself shows significant condition dependence. Therefore, the need for condition-specific and hierarchical condition evaluation has become a general consensus in this field.

[0003] However, existing methods for assessing the condition of port cranes still suffer from the following technical shortcomings. First, the evaluation criteria lack adaptability to different operating conditions. They generally employ uniform fixed thresholds or mixed statistical baselines across all operating conditions, failing to reflect the objective differences in the normal output distribution of the same equipment under different load, speed, and position ranges. This leads to false alarms under high operating conditions and missed alarms under low operating conditions. Second, the robustness of the judgment logic is insufficient. Even with the introduction of operating condition signals, the evaluation process typically relies on a single judgment channel, lacking a parallel judgment mechanism between the standard threshold and the equipment's individual historical baseline. Furthermore, it lacks a structured conflict identification and resolution process when two results are inconsistent or when the initial judgment levels of multiple features are clearly contradictory, resulting in an inability to stably output a reliable state level in contradictory scenarios. Third, the system lacks self-calibration capabilities for configuration changes and gradual drift. After key component replacement, sensor replacement, or long-term wear, the baseline of the monitoring signal distribution may shift abruptly or continuously. Existing methods lack configuration version management and corresponding recalibration and backtracking verification mechanisms, causing the original threshold and baseline models to quickly become invalid. Fourth, the robustness of evaluation under unbalanced sample conditions is insufficient. High-risk or extreme condition sub-intervals naturally have scarce samples, and existing technologies lack structured strategies for backing up and completing insufficient samples, which leads to the interruption of the evaluation link under some conditions, affecting the continuity of monitoring coverage and the traceability of results.

[0004] In summary, the key technical problems that urgently need to be solved in this field are: how to construct a hierarchical and refined evaluation system for different mechanisms, measurement points, and operating condition sub-intervals under complex and ever-changing operating conditions; how to establish a dual-channel parallel judgment and evidence weight configuration and conflict resolution mechanism based on conflict type to improve the credibility of the judgment results; how to introduce configuration version management and self-calibration mechanism to cope with changes in equipment configuration and long-period signal drift; and how to achieve traceable robust evaluation in operating condition sub-intervals with insufficient samples. Summary of the Invention

[0005] To address the technical problems in existing port crane condition assessments, such as the inability of fixed thresholds and single baselines to accommodate complex and variable operating conditions and individual equipment differences, and the lack of structured conflict resolution, versioned recalibration, and mechanisms to ensure evaluation continuity when samples are insufficient, leading to misjudgments, omissions, and long-term instability of the assessment system, this invention proposes a layered condition assessment and self-calibration method and system for port cranes. This method constructs the current assessment object as an assessment request data object and matches layered assessment units corresponding to equipment, mechanisms, measuring points, and operating condition sub-intervals based on lookup keys. It introduces a dual-channel parallel judgment using standard thresholds and historical baselines, supplemented by a closed-loop mechanism based on conflict type-based evidence weight configuration and conflict resolution, configuration versioned recalibration, and progressive supplementation for insufficient samples. This allows for stable output of condition levels under complex operating conditions and drift conditions, achieving a comprehensive effect that balances the universality of initial screening with individual adaptability, reduces misjudgments and omissions, and ensures evaluation continuity.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for evaluating and self-calibrating the working conditions of port cranes, the method comprising: Step S1: Obtain the operating condition signals and condition monitoring characteristic values ​​of the port crane, and construct an evaluation request data object from the current evaluation object, which consists of equipment type identifier, target mechanism identifier, measuring point group identifier, operating condition sub-interval label, and condition monitoring characteristic value vector. ;by The equipment type identifier, target organization identifier, measurement point group identifier, and working condition sub-interval label are used to generate the search key. And match the corresponding hierarchical evaluation unit in the preset hierarchical evaluation unit index library. ; Step S2: For the hierarchical evaluation unit, retrieve its bound standard reference threshold set and historical statistical baseline model in parallel, compare the state monitoring feature value with the standard reference threshold and the historical statistical baseline respectively, obtain the threshold channel judgment result and the baseline channel judgment result, and form the initial state level of each state monitoring feature; Step S3: Determine the judgment results of the status monitoring feature values ​​in the threshold channel and baseline channel respectively, and identify the conflict type; Step S4: For the identified conflict type, select evidence items to participate in the calculation from the preset evidence item set, and configure corresponding weight vectors for the selected evidence items; calculate the conflict resolution score according to the selected evidence items and their weights, map the conflict resolution score to the anomaly confidence level, and output the final state level according to the anomaly confidence level and the initial state level. Step S5: When a configuration change event or a persistent drift event is detected, a new configuration version is established and the original threshold and baseline model are frozen; steady-state samples of the new configuration in each operating condition sub-interval are collected to establish a candidate new baseline model, and a recalibration observation period is started; during the observation period, the evaluation results of the new and old models are output in parallel and backtracking verification is performed; when the candidate new baseline model meets the model switching conditions, the model switching is performed.

[0007] On the other hand, the present invention also provides a port crane condition-based hierarchical status evaluation and self-calibration system, the system being used to perform the above-described method, the system comprising: The hierarchical evaluation unit construction module is used to acquire the operating condition signals and status monitoring characteristic values ​​of port cranes. It constructs an evaluation request data object from the current evaluation object, which consists of equipment type identifier, target mechanism identifier, measuring point group identifier, operating condition sub-interval label, and status monitoring characteristic value vector. ;by The equipment type identifier, target organization identifier, measurement point group identifier, and working condition sub-interval label are used to generate the search key. And match the corresponding hierarchical evaluation unit in the preset hierarchical evaluation unit index library. ; The dual-channel parallel judgment module is used to retrieve the standard reference threshold set and historical statistical baseline model bound to the hierarchical evaluation unit in parallel, compare the state monitoring feature value with the standard reference threshold and the historical statistical baseline respectively, obtain the threshold channel judgment result and the baseline channel judgment result, and form the initial judgment state level of each state monitoring feature. The conflict type identification module is used to determine the judgment results of the status monitoring feature values ​​under the threshold channel and the baseline channel, and to identify the conflict type. The evidence weight configuration and conflict resolution module is used to select evidence items for calculation from a preset set of evidence items for the identified conflict type, and configure corresponding weight vectors for the selected evidence items; calculate the conflict resolution score based on the selected evidence items and their weights, map the conflict resolution score to anomaly confidence, and output the final state level based on the anomaly confidence and the initial state level. The configuration version management and self-calibration module is used to establish a new configuration version and freeze the original threshold and baseline model when a configuration change event or a continuous drift event is detected; to collect steady-state samples of the new configuration in each operating condition sub-interval to establish a candidate new baseline model and start the recalibration observation period; to output the evaluation results of the new and old models in parallel during the observation period and to perform backtracking verification; and to perform model switching when the candidate new baseline model meets the model switching conditions.

[0008] On the other hand, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-described method for evaluating and self-calibrating the working conditions of port cranes.

[0009] On the other hand, the present invention also provides a computer program product, including program instructions, which, when executed by a processor, are used to implement the above-mentioned port crane working condition layered status evaluation and self-calibration method.

[0010] Compared with the prior art, the beneficial effects of the present invention are: (1) By defining the current evaluation object as an evaluation request data object and with (1) Accurately match the hierarchical evaluation unit u, so that the status evaluation is transformed from a global threshold to a fine evaluation by equipment, organization, measurement point and working condition sub-interval, and improve the traceability and operability of the evaluation results; (2) Make judgments in parallel by standard reference threshold channel and historical statistical baseline channel, while taking into account the standardization screening capability and the adaptability of individual equipment differences; (3) By setting the conflict type identification rules with evidence weight vector and conflict resolution score formula, the status level and abnormal confidence can be output in scenarios such as threshold overrun, baseline drift and multi-feature level contradiction, reducing misjudgment and omission. (4) By managing configuration versions, backtracking verification, recalibrating observation period and verifying consistency between new and old models, the risk of threshold failure after equipment modification, key component replacement and sensor replacement is reduced; (5) By using the insufficient sample rollback mechanism and source identification mechanism, the engineering availability and result traceability of the evaluation system under complex working conditions are improved; (6) By using the completed time-aligned state monitoring feature values ​​with clear working condition attribution as evaluation input, the decoupling of the preceding data acquisition and subsequent state evaluation is realized, which is convenient for collaborative integration with the steady-state window extraction, fault diagnosis and trend analysis modules. The above effects are based on the premise that the evaluation request data object, hierarchical evaluation unit, configuration version, evidence weight vector and insufficient sample identifier are all recorded and traceable.

[0011] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0012] Figure 1This is a schematic diagram of the overall system structure provided by the present invention; Figure 2 This is a schematic diagram of the hierarchical evaluation unit mapping provided by the present invention; Figure 3 A schematic diagram of the dual-channel parallel determination and conflict type identification process provided by the present invention; Figure 4 A schematic diagram of the conflict resolution process based on evidence weight configuration provided by the present invention; Figure 5 A schematic diagram illustrating the configuration version management, backtracking verification, recalibration observation period, and model switching process provided by this invention; Figure 6 This is a schematic diagram of the sample insufficiency rollback and migration source labeling process provided by the present invention; Figure 7 This is a schematic diagram illustrating the closed-loop process of result classification, manual review, version association, and baseline update provided by the present invention. Detailed Implementation

[0013] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings, so as to more clearly understand the purpose, features and advantages of this invention. It should be understood that the embodiments shown in the drawings are not intended to limit the scope of this invention, but are only for illustrating the essential spirit of the technical solutions of this invention. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0014] Unless the context requires otherwise, throughout the specification and claims, the word “comprising” and its variations, such as “including” and “having”, shall be understood to have an open, inclusive meaning, that is, to be interpreted as “including, but not limited to”.

[0015] Throughout this specification, references to "an embodiment" or "an embodiment" indicate that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Therefore, the appearance of "in an embodiment" or "an embodiment" in various places throughout the specification does not necessarily refer to the same embodiment. Furthermore, a particular feature, structure, or characteristic may be combined in any manner in one or more embodiments.

[0016] The singular forms “a” and “the” used in this specification and the appended claims include plural references unless otherwise expressly stated herein. It should be noted that the term “or” is generally used to mean “and / or” unless otherwise expressly stated herein.

[0017] In the following description, in order to clearly demonstrate the structure and working method of the present invention, a number of directional terms will be used. However, terms such as "front", "back", "left", "right", "outside", "inside", "outward", "inward", "up", and "down" should be understood as convenient terms and not as limiting terms.

[0018] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention. All other embodiments obtained by those skilled in the art based on the embodiments described herein without inventive effort are within the scope of protection of this invention.

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the following description will be provided in conjunction with the appendix. Figure 1 To be continued Figure 7 The present invention will be further described in detail below with reference to specific implementation methods. In the following embodiments, the condition monitoring characteristic values ​​and operating condition labels can be provided in real time by the preceding data acquisition, unified time base alignment, and steady-state window extraction modules, or they can be directly provided by the field acquisition system synchronized with the control system. The evaluation objects involved include at least one or more of the following port crane types: quay container cranes, gantry cranes, ship loaders, and ship unloaders; the target mechanisms include at least one or more of the following: hoisting mechanisms, trolley traveling mechanisms, crane traveling mechanisms, slewing mechanisms, and luffing mechanisms.

[0020] like Figure 1As shown, the overall system structure of this invention includes: a data access module 100, which includes a status feature input 110 and a working condition label input 111; a unit mapping module 120; a dual-channel judgment section, consisting of a standard reference threshold channel 131 and a historical statistical baseline channel 132; a consistency judgment module 130; a conflict identification module 140; an evidence weight configuration and conflict resolution module 160; a result output module 150; an evaluation result library 170a; and a threshold / baseline / version library 170b. Working condition signals are accessed through the working condition label input 111, and status monitoring feature values ​​are accessed through the status feature input 110. Both are sent to the unit mapping module 120 to construct hierarchical evaluation units. The output of the unit mapping module 120 is simultaneously and in parallel sent to the standard reference threshold channel 131 and the historical statistical baseline channel 132. The initial judgment results of both channels are sent to the consistency judgment module 130; if the results of the two channels are consistent, the initial judgment result is directly sent to the result output module 150; if they are inconsistent, they are sent to the conflict identification module 140 to identify the conflict type. The conflict type and required evidence output by the conflict identification module 140 are transmitted to the evidence weight configuration and conflict resolution module 160. After resolution calculation, the final state level, anomaly confidence level, and maintenance suggestions are output to the result output module 150. The result output module 150 stores the evaluation results in the evaluation result library 170a. After review and confirmation, sample accumulation, and baseline update, the data is fed back to the threshold / baseline / version library 170b. The threshold / baseline / version library 170b feeds back to the standard reference threshold channel 131 and the historical statistical baseline channel 132 through the "baseline model / version configuration" path shown by the dotted line, and feeds back to the evidence weight configuration and conflict resolution module 160 through the "weight template / resolution parameter / version configuration" path, thus forming a full-process self-calibration mechanism.

[0021] It is important to understand that, in a specific implementation of this invention, the current evaluation object is not simply the name of the port crane type, but an evaluation request data object R that can be processed by a computer, and its operable definition is a quintuple. The field values ​​of this evaluation request data object R should be fixed before entering the subsequent evaluation process. Subsequent threshold invocation, baseline retrieval, evidence weight configuration, and evaluation result storage all use this field combination as an index to avoid the evaluation results becoming unreproducible due to changes in the index range within the same evaluation period.

[0022] ; in: : Equipment type identifier, which is an enumerated data type, whose value space includes at least {shore container crane, gantry crane, ship loader, ship unloader}; : Target mechanism identifier, whose value space includes at least {hoisting mechanism, trolley traveling mechanism, gantry traveling mechanism, slewing mechanism, luffing mechanism}; : Measurement point grouping identifier, whose value space includes at least {motor mounting position, gearbox housing, bearing housing, coupling neighborhood, mechanism transmission support part}; A multidimensional feature vector of operating conditions obtained by analyzing real-time acquired operating condition signals, its mathematical representation is as follows: ,in For load range labels, For speed range labels, Labels for position or angle ranges. Labels for task stages. For braking status label, For linkage status labels; in, The generation process involves binning and discretizing the multidimensional continuous operating condition signal. Taking the load dimension as an example, let the continuous physical quantity collected by the load sensor be... The preset box boundaries, i.e., the load interval boundary points, are... Then the load range label The determination rule is: if ,but The processing of speed and position dimensions is similar. Operation phases, braking states, and linkage states are directly mapped to discrete labels based on the corresponding status words provided by the control system, such as the braking state label. The value can be 0 (non-braking) or 1 (braking), and the linkage status label is displayed. It can be 0 (non-linked) or 1 (linked). This discretization process ensures that the originally continuous physical conditions can be classified into a finite range of sub-conditions, making the mapping operation deterministic and reproducible in engineering.

[0023] : Acquired and extracted by multi-source sensors The dimensional state monitoring eigenvalue vector, mathematically represented as: .

[0024] Accordingly, the pre-defined hierarchical evaluation unit is a pre-calculated and stored data index structure U that can be queried and managed through a database system. Each hierarchical evaluation unit u is assigned a unique identifier key. The key is composed of the non-feature values ​​from the aforementioned object properties: ; Based on the above definition, the current evaluation object is mapped to a preset hierarchical evaluation unit. This operation specifically includes the following sub-steps: Constructing an evaluation request data object: After receiving the input data, the system parses the above fields from the data source and instantiates an evaluation request data object R.

[0025] Calculate lookup key: Extract In , , and These are combined to form a search key to be matched.

[0026] Perform a mapping lookup: Match the lookup key with the unique identifier key of all preset units u in the hierarchical evaluation unit index. Perform an exact match.

[0027] If the match is successful, that is If the mapping is successful, then the evaluation object has been successfully mapped to the hierarchical evaluation unit. .

[0028] In a preferred embodiment, exact matching takes precedence over invoking similar operating conditions; similar operating condition rules are only activated when exact matching fails or the number of valid samples in the target hierarchical evaluation unit is insufficient. The similar operating condition rules refer to rules that determine the similarity of labels for operating condition sub-intervals under the premise of consistent equipment type, target mechanism, and measurement point grouping; where continuous ordered operating condition dimensions such as load, speed, position, height, or angle can be judged using adjacent bins or normalized distance, and discrete operating condition dimensions such as operation stage, braking state, and linkage state use consistency judgment. When the similarity between a candidate operating condition sub-interval and a target operating condition sub-interval is not less than a preset similarity threshold, it is determined to be a similar operating condition. The triggering reason, candidate operating condition source, similarity or bin distance, and completion method of the similar operating condition rules should be recorded along with the evaluation results to reproduce the call path from the evaluation request data object to the actual evaluation baseline.

[0029] Then, the system extracts In To use it as the unit The set of feature values ​​to be evaluated is then sent to the subsequent dual-channel parallel judgment process.

[0030] Example 1 In this embodiment, the hoisting mechanism of a quay container crane is used as a specific application object to describe the complete evaluation process of the present invention in detail. First, the system receives time-aligned vibration velocity RMS values, vibration acceleration root mean square values, peak values, kurtosis, spectral peak values, and envelope spectrum indices, along with corresponding operating condition labels, including the hoisting uniform speed stage, heavy load range, high speed range, high position range, braking state, and non-linkage state. Based on these operating condition labels, the current evaluation object is mapped to a hierarchical evaluation unit defined by the equipment type being a quay container crane, the target mechanism being the hoisting mechanism, the measurement point grouping being bearing housing measurement points, and the operating condition sub-range being heavy load-high speed-high position-non-linkage. The establishment of this hierarchical evaluation unit determines the precise data retrieval range for subsequent calls to the corresponding standard thresholds and historical baselines. The system then retrieves the standard reference threshold set and historical statistical baseline model bound to this hierarchical evaluation unit from the database and initiates a dual-channel parallel judgment process.

[0031] The standard reference threshold may be derived from national standards, industry standards, equipment manufacturer recommended limits, port enterprise operation and maintenance procedures, historical maintenance experience thresholds, or an initial threshold set confirmed by experts.

[0032] like Figure 2 As shown, the mapping process of the hierarchical evaluation unit is as follows: Input the evaluation request data object R 201 and extract the lookup key Key(R); then, based on the key, perform equipment type layer matching 202, target mechanism layer matching 203, measurement point grouping layer matching 204, and working condition sub-interval layer determination 205 in sequence; after the above layer-by-layer matching, finally output a uniquely determined hierarchical evaluation unit u 206. This hierarchical evaluation unit u is bound to the corresponding standard reference threshold, historical statistical baseline model, and evidence weight configuration.

[0033] To further define the engineering meaning of the hierarchical evaluation unit, the equipment type, target mechanism, measurement point group, and working condition sub-interval are not only used for display or statistical classification, but also serve as common index fields for calling the standard reference threshold set, historical statistical baseline model, evidence weight vector, and configuration version record. The same state monitoring feature value is only compared horizontally, sample merging, or migration completion when the above index fields are consistent or meet the preset similar working condition rules, so as to avoid mixing non-stationary responses under different loads, speeds, positions, or operating stages for unified judgment.

[0034] Preferably, the measurement point grouping layer is divided according to the location of the measurement points, and the location includes at least one or more of the following: motor mounting position, gearbox housing, bearing housing, coupling neighborhood, and mechanism transmission support location. Accordingly, the condition monitoring characteristic value is extracted from the physical signals collected in real time by vibration sensors, temperature sensors, current transformers, and oil sensors deployed at the above-mentioned locations, so that the evaluation object of the present invention directly corresponds to the actual operating state of the target mechanism of the port crane, rather than a pure data processing result detached from the equipment object.

[0035] In some implementations, within the hierarchical evaluation unit, the standard reference threshold channel reads the reference threshold set corresponding to the target mechanism and measurement point group, and the historical statistical baseline channel reads the steady-state sample distribution in multiple work cycles under the same or similar working condition labels. The steady-state sample refers to the state monitoring feature value sample extracted from the steady-state output segment of each work cycle. If the effective value of vibration velocity exceeds the reference threshold but does not exceed the historical statistical baseline boundary, it is identified as a threshold overshooting conflict, and the weight of additional diagnostic feature evidence is given priority. If the effective value of vibration velocity does not exceed the reference threshold, but the envelope spectrum index and kurtosis under the same working condition continuously exceed the historical statistical baseline boundary and show a unidirectional shift in multiple consecutive evaluation cycles, it is identified as a baseline drift conflict, and the weight of cross-modal consistency evidence and continuous drift evidence is given priority. If, within the same hierarchical evaluation unit, the set of state levels obtained after the initial judgment of multiple features such as vibration intensity features, envelope spectrum index, and kurtosis through their respective channels is {1,2,3}, then the difference between the maximum level value and the minimum level value is 2. When the preset level difference threshold is 2, it is determined to be a multi-feature level contradiction conflict. Repeated occurrences within multiple steady-state segments are no longer used as defining conditions for this conflict type, but are instead used as evidence of subsequent recurrence frequency in conflict resolution score calculation. After evidence weight configuration and conflict resolution are completed, the output status level, anomaly confidence level, and maintenance recommendations such as continued observation, shortening the detection cycle, or scheduling planned maintenance are provided. If multiple high-risk preliminary judgments originate from highly correlated derived indicators of the same original signal, the same frequency band, or the same intermediate calculation result, they should be treated as related evidence when configuring evidence weights, reducing the weight of some evidence items, or selecting only one representative evidence item for positive weighting to avoid double-weighting the same anomaly source; if they originate from different sensors, different frequency bands, or different physical mechanisms, they can be used as independent evidence items for resolution.

[0036] like Figure 3As shown, the dual-channel parallel judgment and conflict type identification process specifically includes the following steps: First, the state monitoring feature value x 301 of the hierarchical evaluation unit u is input and simultaneously sent to the standard threshold channel 302 and the historical baseline channel 303 composed of the historical statistical baseline model for initial threshold judgment and baseline judgment, respectively; then, the initial judgment results of the two channels are sent to the conflict condition judgment step 304; if it is determined that there is no conflict, the initial judgment state level 305 is directly output; if it is determined that there is a conflict, it enters the conflict type identification step 306, and the identified conflict types include threshold overrun type, baseline drift type or multi-feature level contradiction type, and the identified conflict type is output to the subsequent evidence resolution process 307.

[0037] like Figure 4 As shown, the conflict resolution process based on evidence weight configuration specifically includes the following steps: First, step 401 is executed, receiving information on the evaluation object, including conflict type C, hierarchical evaluation unit u, current status monitoring feature value, and sample identifier, as input to the conflict resolution process. Subsequently, based on conflict type C, the system enters one of three differentiated processing paths. Step 410 corresponds to the threshold exceeding the limit conflict path, triggered when the standard threshold channel is judged as abnormal while the historical baseline channel is judged as normal. This path prioritizes calling additional diagnostic feature evidence. Evidence of abnormal duration Step 420 corresponds to a baseline drift-type conflict path, which is triggered when the standard threshold channel is judged as normal while the historical baseline channel is judged as abnormal. This path prioritizes calling cross-modal consistency evidence. and evidence of persistent drift Step 430 corresponds to a multi-feature level contradictory conflict path. Its trigger condition is that the difference between the maximum and minimum values ​​of the initial judgment levels of multiple features within the same hierarchical evaluation unit is not less than a preset level difference threshold. This path is based on evidence of abnormal duration. Evidence of recurrence frequency The primary focus is on configuring additional diagnostic feature evidence based on the independence of the additional diagnostic features. The weight.

[0038] Step 440 involves evidence screening and weight allocation, applying the positive evidence selected from the aforementioned paths. , , , , Perform screening and determine the penalty for insufficient samples. And assign a corresponding weight vector to each piece of evidence according to the conflict type C. , , , , and If a piece of evidence is unavailable, its weight is reset to zero, and the remaining positive evidence items are renormalized so that the sum of their weights is 1. Step 450 is the conflict resolution score calculation. The configured evidence items and weights are substituted into a unified conflict resolution score formula to calculate the conflict resolution score. Furthermore, the score is mapped to anomaly confidence level using a mapping function. Finally, step 460 outputs the conflict resolution, summarizing the anomaly confidence level. Based on the initial state level assessment, output the final state level after resolution, the anomaly confidence level, and corresponding maintenance recommendations.

[0039] In some specific implementations, for any hierarchical evaluation unit u and its corresponding k-th state monitoring feature, let the set of historical steady-state samples in multiple work cycles under the same or similar working condition labels be: To enhance adaptability to non-Gaussian distributions, outliers, and operating condition fluctuations, the historical statistical baseline center value and robust scale are defined as follows: Furthermore, the quantile boundaries and the deviation from the historical statistical baseline are defined as follows: where, + + =1.

[0040] in, Represents a hierarchical evaluation unit Next A set of historical steady-state samples for each state monitoring feature; Represents the first in the sample set Each sample value; Indicates the central value of the historical statistical baseline; A robust scale representing the historical statistical baseline; This represents median operations; Represents absolute value; This indicates a small positive number that prevents the denominator from being zero. This is a robust scaling correction constant used to convert the absolute deviation of the median to a scale equal to the standard deviation; its value is derived from the condition of a normal distribution. ,Right now After adopting this correction constant, It can be used as a robust scale estimator with dimensions comparable to standard deviation, and can be used to reduce the impact of outliers, non-Gaussian fluctuations and a small number of outliers on the calculation of deviation from historical statistical baselines. and These represent the lower and upper quantile boundaries of the historical sample distribution, respectively. and They represent quantiles and quantiles; This indicates the degree of deviation of the current feature value from the historical statistical baseline; , , This represents the weighting coefficient corresponding to each deviation term.

[0041] Based on the results of the threshold channel and baseline channel determinations, different types of conflicts can be identified, including: When a certain status monitoring feature is determined to be abnormal by the threshold channel but normal by the baseline channel, it is identified as a threshold overrun conflict; the determination rule for the threshold channel abnormality is determined by the upper limit threshold, lower limit threshold or normal range threshold configured for the feature. When a certain state monitoring feature is judged to be normal by the threshold channel but abnormal by the baseline channel, it is identified as a baseline drift type conflict; where, let The baseline anomaly detection threshold is set, and the criteria for determining baseline channel anomalies are the historical statistical baseline deviation. Not less than ; When the same stratified evaluation unit Inside The status level is initially determined by each status monitoring feature through its respective channel. to satisfy Not less than the preset grade difference threshold At that time, it was identified as a multi-feature hierarchical contradictory conflict; The frequency of recurrence and the duration of abnormal occurrences are used for subsequent conflict resolution score calculations, but are not considered as conditions for the establishment of multi-feature level conflict types.

[0042] The rule for determining threshold channel anomalies is determined based on the configured threshold type: if the feature is configured with an upper limit threshold... ,but This is abnormal; if a lower threshold is configured. ,but This is abnormal; if a range threshold is configured [ , ], then the current eigenvalue If the value falls outside this interval, it is considered abnormal. The criteria for determining an anomaly in the baseline channel are as follows: Not less than ,in The baseline anomaly detection threshold, This is the preset threshold for grade difference.

[0043] In other words, the standard reference threshold is compared with the current state monitoring feature value. Baseline anomaly detection threshold The comparison object is the deviation from the historical statistical baseline. The two serve the standard reference threshold channel and the historical statistical baseline channel respectively, and are not compared with each other.

[0044] To enhance the transparency of project implementation, the above parameters can be initialized based on experience in monitoring the condition of port crane mechanisms, and then verified and corrected using historical samples during the trial operation period. Preferably, the quantile parameters... The value can be between 0.01 and 0.10, with an initial value of 0.05; when the number of valid samples is less than 50, A value of 0.05 to 0.10 is recommended to avoid using excessively extreme quantiles when the sample size is insufficient, which could lead to boundary instability. When the effective sample size reaches more than 100, α can be taken as 0.01 to 0.05 to improve the ability to identify the distribution of outliers. This prevents the denominator from being zero. After feature normalization, the following can be taken: ~ When not normalized, the corresponding characteristic engineering dimension reference value can be taken. ~ Multiples; Baseline anomaly detection threshold The threshold can be 2.0–4.0, with an initial value of 2.5–3.0. Values ​​of 2.0–2.5 are suitable for early warning, while values ​​of 3.0–4.0 are suitable for sub-intervals with high noise, high fluctuations, or small sample sizes. For the four-level state level coding, the level difference threshold... The preferred value is 2. A value of 1 can be used when a more sensitive warning is needed, but when 1 is chosen, it should only trigger observation, review, or conflict resolution, and should not be directly used as the basis for status escalation. The weights of each deviation item used for historical statistical baseline deviation are... , , The values ​​can be set to 0.4–0.6, 0.2–0.4, or 0.1–0.3. The initial combination can be 0.5, 0.3, or 0.2, and the sum of all weights should be 1. After verification during the trial operation period, the above parameters should be stored together with the equipment number, target mechanism, measurement point grouping, operating condition sub-interval, and configuration version. Subsequent parameter adjustments should form a version record and should not overwrite the original evaluation results.

[0045] The above parameter range represents a preferred initialization range suitable for port crane mechanism condition monitoring scenarios, and is not the only limitation on all implementation methods. Quantile parameters Used to control the coverage range of historical normal sample quantile boundaries; when the sample size is small, a larger value is used. This avoids the instability of boundary values ​​caused by extreme quantiles being affected by a small number of samples. When the sample size is sufficient, a smaller quantile is used. This allows for a more complete characterization of the tail distribution of historical normal samples. Parameters As a numerical stability term, it is used to prevent division by zero or abnormal amplification of deviation when the scale term or quantile interval width is close to zero. Baseline anomaly detection threshold. Based on engineering experience in judging robust scale normalization deviations, this setting aims to strike a balance between sensitive early warning and false alarm control. (Level difference threshold) To identify significant contradictions between initial judgments of multiple features, a value of 2 can prevent excessive alarms caused by fluctuations in adjacent levels; a value of 1 should only trigger observation, review, or conflict resolution processes, and should not be directly used as the basis for status escalation. Weighting , , This is used to adjust the contribution of each deviation term to the deviation of the historical statistical baseline. The initial setting is based on center-scale deviation as the main factor, quantile boundary deviation as the secondary factor, and other deviation terms as supplementary factors. During the trial operation period, it is checked and corrected based on the historical sample distribution, manual review results, and maintenance verification records.

[0046] Furthermore, the values ​​of the above parameters can be determined by combining robust statistical methods, experience with non-stationary operating conditions of port cranes, and historical sample verification during the trial operation period. Specifically, robust statistical methods are used to determine basic calculation methods such as the median, robust scale, quantile boundaries, and normalized deviation. Experience with non-stationary operating conditions of port cranes is used to constrain threshold sensitivity, level jump identification, and false alarm control requirements under variations in load, speed, position, and operating phase. Historical sample verification during the trial operation period is used to verify the parameters based on historical normal samples, manual review results, maintenance verification records, false alarm and missed alarm situations, and the number of samples in each operating condition sub-interval. , , , as well as , , The above parameters, after verification, should be bound and saved with the corresponding configuration version, hierarchical evaluation unit, and operating condition sub-interval for subsequent state evaluation, model switching, and result traceability.

[0047] To facilitate the calculation of risk level differences, consistency rate statistics, and stability verification, this implementation method encodes the status levels as integer values ​​in ascending order of risk. For example, normal, low risk, medium risk, and high risk can be encoded as 1, 2, 3, and 4, respectively, or... The status levels are mapped to 1, 2, 3 and 4 respectively; in different equipment or different institutional scenarios, other numerical coding methods that maintain a monotonically increasing relationship of risk can also be used.

[0048] Among them, the robust sizing formula This is a correction constant for converting the absolute deviation of the median to a scale similar to the standard deviation, derived from the condition of a normal distribution. Approximately ,Right now After adopting this correction constant, the robust sizing... While preserving the median absolute deviation's ability to resist outliers, it remains comparable to the traditional standard deviation scale, facilitating subsequent baseline deviation analysis. and baseline anomaly detection threshold The engineering explanation.

[0049] The above quantile parameters To prevent the denominator from being zero Baseline anomaly detection threshold Grade difference threshold and the weight of deviation from historical statistical baseline , , The range of values ​​should preferably be determined using a combination of robust statistical methods, experience with non-stationary operating conditions of port cranes, and verification using historical samples from the trial operation period. Robust statistical methods are used to define the mathematical meaning of quantile boundaries, robust centers, and robust scales. Experience with non-stationary operating conditions of port cranes is used to account for normal fluctuations caused by heavy loads, speed changes, braking, linkage, and changes in spatial position. Verification using historical samples from the trial operation period is used to correct the initial parameters based on false alarms, missed alarms, manual verification, and maintenance validation results. The verified parameters should be bound and saved with the configuration version, hierarchical evaluation unit, and operating condition sub-interval for subsequent state evaluation, model switching, and result traceability.

[0050] In the event of a conflict, the target institution is given additional diagnostic evidence. Evidence of cross-modal consistency is Evidence for abnormal duration is Evidence of recurrence frequency is Evidence of continuous drift is The penalty for insufficient samples is The conflict type is The system is based on Call the preset evidence weight vector Then calculate the conflict resolution score. and abnormal confidence They are defined as follows: Preferably, the relevant evidence items can be represented as follows: in, For bias terms, , , , , These are the positive weighting coefficients for additional diagnostic feature evidence, cross-modal consistency evidence, abnormal duration evidence, recurrence frequency evidence, and persistent drift evidence, respectively. The deduction weight corresponding to the insufficient sample penalty item; Indicates the duration of the anomaly. Indicates the normalized baseline for duration; Indicates the number of times the anomaly occurred. Indicates the normalization benchmark for the number of repetitions; Indicates the first evaluation period in the most recent multiple evaluation periods The drift slope of each state monitoring feature, This represents the normalized baseline for the drift slope; This indicates the number of valid samples in the current stratified evaluation unit; This indicates the preset minimum sample size. The optimal value is 20-50, with an initial value of 30. For sub-intervals of naturally sparse but high-risk conditions, the value can be 10-20, with the addition of a sample shortage indicator and requirements for manual verification or subsequent supplementary sampling.

[0051] in, The preferred method is to determine the additional diagnostic quantity of the target mechanism relative to the normalized deviation of the historical normal sample center under the same working conditions. For hoisting mechanism, running mechanism or slewing / luffing mechanism, one or more of the following can be selected: envelope spectrum fault frequency band amplitude deviation, bearing fault frequency amplitude deviation, gear meshing frequency and its sideband amplitude deviation, and motor current characteristic frequency band deviation. The preferred value is obtained by weighting the normalized deviation of the temperature rise rate, the normalized deviation of the current, the normalized deviation of the power, and / or the normalized deviation of the oil particle index. For equipment without an oil sensor, temperature sensor, or power measurement channel, the value can be calculated based on two or more of the available cross-modal signals. The above-mentioned normalized deviations are normalized deviations relative to the center of historical normal samples under the same working conditions. They can be obtained by (current value - historical normal sample center value) / historical normal sample robust scale or similar normalized mapping, where the historical normal sample center value and robust scale are calculated with reference to the median and MAD in the historical statistical baseline model. Preferably, the most recent ones arranged in chronological order are those within the same configuration version, the same hierarchical evaluation unit, and under the same or similar operating condition labels. The results are obtained by linear fitting on the feature value sequence of each evaluation cycle to avoid pseudo-drift introduced by the switching of operating conditions. This indicates the number of most recent evaluation periods selected for drift slope fitting, with a value ranging from 5 to 20 evaluation periods, preferably 8 to 12. When the evaluation period is set according to the job cycle... It is advisable to select a higher value from the above range to obtain more comprehensive trend data; when the evaluation period is set by hour or shift, The lower end of the above range can be selected to capture recent drift dynamics in a timely manner.

[0052] The above-mentioned configuration of corresponding evidence weight vectors based on conflict type is specifically manifested in this embodiment as follows: the system pre-configures different evidence weight vectors for different conflict types. The weight vector is used to determine the evidence items that actually participate in the score calculation and their relative contributions. The specific configuration principle is as follows: the configuration result of the weight vector should be saved along with each conflict resolution result, including the conflict type, participating evidence items, non-participating evidence items and their reasons for non-participation, for subsequent review.

[0053] For threshold-out-of-bounds conflicts, the characteristic is that the feature value exceeds the standard reference threshold but does not deviate from the historical statistical baseline. In this case, the focus is on checking whether this feature is accompanied by additional diagnostic symptoms related to the fault. Set to a higher value, Set to zero or a small value; For baseline drift-type conflicts, the characteristic is that the eigenvalues ​​do not exceed the standard reference threshold but have significantly deviated from the historical statistical baseline. In this case, the focus is on verifying whether this deviation is reflected in multiple cross-modal monitoring quantities and whether it exhibits a unidirectional, continuous drift trend. Therefore, and Set to a higher value, Set to zero or a small value; For multi-feature level contradictory conflicts, the characteristic is that the initial judgment levels given by different state monitoring features within the same hierarchical evaluation unit are contradictory. In this case, the focus should be on whether the contradiction persists throughout the continuous evaluation period and whether it repeatedly occurs in the same or similar operating condition sub-intervals; additional diagnostic features are used to determine whether the contradiction has a clear fault location or fault mechanism, and should not be unconditionally used as the highest weighted evidence in all scenarios. Therefore, under this type of conflict, it is advisable to use... and As the dominant weighting term; when the additional diagnostic features can point to the same fault location or fault mechanism as the high-level initial judgment features, and are not completely derived from the high-level initial judgment features, they can be used as the dominant weighting term; Set to medium-high or high values; when additional diagnostic features are missing, lack confidence, are inconsistent with the direction of conflict, or are highly homologous to high-level preliminary diagnostic features, the values ​​should be reduced. The value of or not to be used as the primary positive evidence item.

[0054] Furthermore, if a certain piece of evidence is not selected, the corresponding sensor is not configured, or it is unavailable in the current configuration version, hierarchical evaluation unit, and operating condition sub-interval, then the weight corresponding to that piece of evidence can be set to 0; the remaining positive evidence items participating in the calculation should be re-normalized so that the sum of the weights of the positive evidence items participating in the calculation is 1, and the insufficient sample penalty item... It will still be used as a deduction weight in the calculation of conflict resolution score.

[0055] For multi-feature level contradictions, if there is only a single level difference without evidence of duration, evidence of recurrence, or additional diagnostic features pointing to an independent failure mechanism, the conflict should be triggered by observation, manual review, or a shortened detection cycle. It is not advisable to directly upgrade the final state level based solely on a single multi-feature level difference. When upgrading the state level accordingly, the system should at least record the evidence items involved in the upgrade, the weight vector, the source of the evidence, and the conflict resolution score for subsequent auditing and manual review.

[0056] The weights of each positive evidence item are normalized for each conflict type so that their sum equals 1; the corresponding weights of the insufficient sample penalty items are... As a deduction weight, it is set separately and does not participate in the normalization of positive evidence item weights. Through the above configuration, the same conflict resolution scoring formula calls different sets of evidence items under different conflict types, thus making the correspondence between "conflict type—evidence item—weight vector—conflict resolution score" clear, calculable, and consistent with... Figure 4 The process shown is consistent.

[0057] Furthermore, each piece of evidence is determined based on the current evaluation input, historical normal samples, historical abnormal samples, cross-modal monitoring data, and / or maintenance review records under the same configuration version, the same hierarchical evaluation unit, and the same or similar operating condition sub-intervals; the cross-modal consistency evidence is corrected based on whether two or more cross-modal monitoring data point to the same risk direction under the same evaluation cycle, the same hierarchical evaluation unit, and the same or similar operating condition sub-intervals; different conflict types correspond to different priority or weight configurations for evidence items, and for each conflict type, , , , and The weights of the positive evidence terms are normalized to sum to 1; the insufficient sample penalty term... by Form is included as a deduction item in the conflict resolution score calculation. It does not participate in the weight normalization of positive evidence items.

[0058] In the parameter settings for conflict resolution scores, the weights for additional diagnostic feature evidence can be set to 0.25–0.35, cross-modal consistency evidence to 0.20–0.30, abnormal duration evidence to 0.10–0.20, recurrence frequency evidence to 0.10–0.20, persistent drift evidence to 0.15–0.25, and insufficient sample penalty to 0.10–0.25. The weights of each positive evidence item can be initialized within the above ranges and normalized before actual calculation so that the sum of the positive evidence item weights is 1. The insufficient sample penalty is set separately as a deduction item and does not participate in the normalization of positive evidence weights. Abnormal duration normalization benchmark. The evaluation period can be 3 to 10 cycles, with 5 cycles being preferred; the normalization benchmark for the number of repetitions. Two to five times can be taken, with three times being preferred; the drift slope is normalized to a benchmark. The evaluation cycle can be set to 0.05–0.20 times the robustness scale of normal samples under the same working conditions. When the evaluation cycle is set according to the work cycle, It is advisable to select a higher value within the above range to avoid misjudging single-cycle disturbances as drift; when the evaluation period is set by hour, shift, or daily statistical window, Lower values ​​within the above range can be used to improve sensitivity to slow degradation trends. When the anomaly confidence level is less than 0.40, a recommendation to continue observation can be issued; when the anomaly confidence level is between 0.40 and 0.70, a recommendation for manual review or shortening the testing cycle can be issued; when the anomaly confidence level is greater than or equal to 0.70, a recommendation for special inspection or planned maintenance can be issued. The above values ​​are the preferred initial range; in actual engineering, calibration can be performed based on equipment type, target mechanism, sensor configuration, sampling cycle, and maintenance verification records. For multi-feature level contradictory conflicts, the above... The value range should only be used when the additional diagnostic features clearly indicate a fault mechanism and do not have completely identical weighting with the high-level initial judgment features; otherwise, it should be adjusted downwards. , and by , It bears the main weight. The maintenance recommendations are auxiliary decision-making information generated based on the condition assessment results, and the final maintenance and disposal can be determined by combining on-site inspection, manual verification, work plan and safety management requirements.

[0059] Therefore, different conflict types can be mapped to different available evidence items and weight vectors, and the state level and anomaly confidence after conflict resolution can be output in a unified quantitative way through the correspondence of "conflict type - available evidence items - normalized weight vector - conflict resolution score".

[0060] The dual-channel parallel judgment and conflict resolution mechanism described above can effectively solve the evaluation problem when the standard threshold and historical baseline judgment conclusions are inconsistent under the same configuration version and stable operating condition distribution. However, during the long-term service of port cranes, the equipment inevitably undergoes changes in key components such as reducer replacement, bearing replacement, and sensor replacement, or irreversible and continuous drift in the sample distribution under the same operating condition due to long-term wear. In such cases, the standard threshold system corresponding to the original hierarchical evaluation unit and the historical statistical baseline model will gradually become mismatched. If the original model is still used for judgment, even if the dual-channel mechanism operates normally, the accuracy of its output results cannot be guaranteed. To this end, this invention further introduces a configuration version management and recalibration mechanism to ensure the long-term effectiveness of the evaluation system after changes in equipment configuration. This will be explained in detail below.

[0061] In some implementations, when the hoisting mechanism undergoes a gearbox replacement, bearing replacement, sensor replacement, or a continuous drift in the distribution of samples under the same operating conditions, the system establishes a new configuration version. This new configuration version can be associated with the equipment number, target mechanism identifier, configuration change event type, and effective time. The system freezes the original threshold system and the original baseline model, retaining the original version identifier and corresponding evaluation records. Subsequently, steady-state samples of the new configuration are collected within each target operating condition sub-interval to establish a candidate new baseline model and / or revise the thresholds, and a recalibration observation period is initiated. During the observation period, the original model and the candidate new baseline model output evaluation results in parallel; simultaneously, backtracking verification of the candidate new baseline model is performed using historically confirmed samples and steady-state samples before the switchover. The model switching conditions can be adjusted according to the equipment type, safety level, and amount of verification data. In this specific implementation, the model switching conditions include the consistency rate of state level, the coverage rate of sub-intervals of operating conditions, the stability of evaluation results, and the comprehensive score of consistency. When the candidate new baseline model meets the model switching conditions, the model switching is performed. If only some sub-intervals of operating conditions meet the conditions, the switching is performed on the sub-intervals of operating conditions that have passed the verification, and the original version is retained or migration calibration is performed on the sub-intervals of operating conditions that have not passed the verification.

[0062] like Figure 5 As shown, the configuration version management, backtracking verification, recalibration observation period, and model switching process specifically includes the following steps: Step 501 is the trigger event detection. The system continuously monitors the operating status of the port crane. When it detects key component change events such as reducer replacement, bearing replacement, sensor replacement, or irreversible continuous drift events in the distribution of samples under the same working conditions, it initiates the self-calibration process.

[0063] Step 502 involves establishing a new configuration version. The system generates a new configuration version identifier and associates it with the equipment number, target organization identifier, configuration change event type, and effective time, thus isolating the old and new states from the management level.

[0064] Step 503 involves freezing the original threshold system and the original baseline model, while retaining the original version identifier and all corresponding evaluation records to ensure the traceability of historical evaluation results and provide a basis for subsequent backtracking verification and version comparison.

[0065] Step 504 involves collecting steady-state samples. The system begins to collect steady-state operating data of the new configuration in each target operating condition sub-interval. These data are collected under the same or similar operating condition labels and are used to train candidate new baseline models.

[0066] Step 505 involves establishing a candidate new baseline model or revising the threshold. Using the collected steady-state samples, the system calculates the new historical statistical baseline center value, robust scale, and quantile boundary, or adjusts the standard reference threshold accordingly to form a candidate new baseline model.

[0067] Step 506 is backtesting. The system uses historically confirmed samples and steady-state samples before the switch to backtest the candidate new baseline model and preliminarily assess the accuracy of its evaluation results.

[0068] Step 507 is to start the recalibration observation period, which is a preset observation time window. During this window, sufficient and diverse operating data covering each target operating condition sub-range is accumulated.

[0069] Step 508 is to output the evaluation results in parallel. During the recalibration observation period, the original model and the candidate new baseline model run simultaneously, and output their respective state levels for the same evaluation object to intuitively compare the behavioral differences between the old and new models.

[0070] Step 509 is the consistency verification. The system quantitatively evaluates the candidate new baseline model based on three key indicators: calculating the state level consistency rate. It measures the consistency between the old and new models for the same input-output state level; and calculates the coverage rate of the target working condition sub-interval. It measures the validation coverage of candidate new baseline models across various operating condition sub-intervals; and calculates the stability of the evaluation results. The consistency score J is calculated by measuring whether the candidate new baseline model exhibits unnecessary level jumps during the validation period and combining the above indicators.

[0071] Step 510 is the overall switching condition judgment, and the system judges the consistency rate of the status level. Target operating condition sub-interval coverage Stability of evaluation results And whether the overall consistency score J fully meets their respective preset thresholds. , , , If all conditions are met, proceed to step 511 to perform a global model switch, i.e., switch the entire target mechanism or all target operating condition sub-intervals to the candidate new baseline model. If the judgment result of step 510 is negative, proceed to step 512 to determine whether any operating condition sub-intervals meet the switching conditions. If at least one operating condition sub-interval passes verification, proceed to step 513 to perform a partial switch on the verified operating condition sub-intervals, and simultaneously proceed to step 514 to retain the original version or perform migration calibration on the operating condition sub-intervals that have not passed verification.

[0072] If the judgment result of step 512 is negative, that is, all working condition sub-intervals do not meet the switching conditions, then proceed to step 515, keep the old and new models evaluated in parallel and continue to observe, and then evaluate again after accumulating more sample data.

[0073] Finally, step 516 outputs the processing results and version identifier. Regardless of the switching decision made, the system outputs the final processing results along with detailed version identifier information, including the currently effective configuration version, the model version used in each operating condition sub-interval, the criteria items that failed verification, and the corresponding sample quantity and observation period range, to ensure the high traceability of the evaluation results.

[0074] In some specific implementations, to quantify the substitutability of candidate new baseline models during the recalibration observation period, it is assumed that there are a total of [number missing] models during the observation period. The validation sample, the candidate new baseline model for the first validation sample, is compared with the first validation sample. The state level output by each verification sample is: The original model output or the reference state level confirmed by manual review is: The consistency rate of state levels is defined as follows: Let the total number of target working condition sub-intervals be . The number of operating condition sub-intervals that passed the minimum sample requirement and completed the verification was . Then the target operating condition sub-interval coverage rate and the stability of the evaluation results They are defined as follows: The total number of evaluation periods included in stability statistics during the observation period When the value is greater than 1, the stability of the evaluation results is calculated according to the above formula; when... If the value is less than or equal to 1, it will not be included in the consistency comprehensive score calculation, or it will be marked as insufficient verification period.

[0075] Furthermore, the consistency score and model switching conditions are defined as follows: in, Indicates the consistency rate of status levels; This represents the total number of paired samples of the old and new models used for validation during the observation period; Indicates the candidate new baseline model for the first The status level output by each verification sample; This indicates the old model output or the reference status level confirmed by manual review; This indicates an indicator function that takes the value 1 when the condition inside the parentheses is true, and 0 otherwise. Indicates the coverage rate of the target operating condition sub-interval; Indicates the total number of sub-intervals for the target operating condition; This represents the number of operating condition sub-intervals that have passed the minimum sample requirement and completed the verification. This indicates the stability of the evaluation results; This indicates the number of unnecessary grade jumps that occurred between adjacent evaluation periods during the observation period; This indicates the total number of evaluation periods included in the stability statistics during the observation period; This indicates the overall consistency score; , , This represents the weighting coefficient of each evaluation indicator in the overall consistency score, and the sum of the three is 1; , , , These represent the preset thresholds corresponding to the model switching conditions, namely the state level consistency rate threshold, the target working condition sub-interval coverage threshold, the evaluation result stability threshold, and the consistency comprehensive score threshold.

[0076] For the weights in the overall consistency score, the weight for the state level consistency rate can be 0.40–0.60, the weight for the target working condition sub-interval coverage rate can be 0.20–0.30, and the weight for the evaluation result stability can be 0.20–0.30. The initial combination can be 0.50, 0.25, and 0.25, and each weight should be normalized so that its sum is 1. In the model switching conditions, the threshold for the state level consistency rate can be 0.88–0.94, the threshold for the target working condition sub-interval coverage rate can be 0.70–0.90, the threshold for the evaluation result stability can be 0.85–0.95, and the threshold for the overall consistency score can be 0.85–0.93. When the port crane is in a high-risk mechanism or critical operation window such as hoisting, braking, slewing, or high load and high speed, the higher values ​​in the above range can be used; when it is in the early stage of sample accumulation or only used for trend observation, the lower values ​​in the above range can be used and the manual review mark should be retained.

[0077] Model switching conditions should be used as a combination of necessary conditions for a candidate new baseline model to replace the original model. It is not advisable to substitute a single indicator meeting the standard or a high overall consistency score for fundamental conditions such as state level consistency rate, target operating condition sub-interval coverage, and evaluation result stability. When any key condition is not met, the original evaluation baseline model should remain unchanged, and the unmet criterion, corresponding operating condition sub-interval, sample size, observation period, and candidate model version identifier should be output. For high-risk institutions or critical operational windows, the observation period may be extended or manual review may be required before model switching. Evaluation results with insufficient sample indicators, migration completion indicators, or those obtained only through threshold backoff should not be used as the sole basis for model switching approval.

[0078] In this embodiment, The statistical criteria are further defined as follows: Only between adjacent evaluation periods of the same configuration version, the same hierarchical evaluation unit, the same or similar operating condition sub-intervals, and without records of maintenance, component replacement, inspection confirmation of abnormalities, sudden changes in operating conditions, or sensor anomalies, is it determined whether an unnecessary level jump has occurred. If the status level changes by only one level and remains at the new level for a subsequent short period, and is supported by evidence of cross-modal anomalies, evidence of anomaly duration, evidence of recurrence frequency, or manual verification conclusions, it will not be counted. If the state level undergoes a sudden change to level 2 or higher, or if... , , Short-cycle recurring jumps that are not supported by evidence of cross-modal anomalies, anomaly duration, recurrence frequency, maintenance records, inspection records, or manual review conclusions are counted as one unnecessary level jump. The short retention period is a preset number of evaluation cycles, preferably 2 to 3 evaluation cycles; if the level change is supported by confirmed fault development, post-repair condition improvement, replacement of key components, sensor replacement, abnormal oil detection, or expert review conclusions, it is not counted.

[0079] in, The reference label can be the original model output, the results of manual verification and confirmation, the results of maintenance and disassembly confirmation, or a comprehensive reference label formed by maintenance records, inspection records, oil test results, and expert review conclusions. When manual verification or maintenance confirmation results exist, they should be preferred as the reference label. To ensure consistency across different time periods, and The same status level coding rules should be used.

[0080] By establishing a new configuration version and freezing the original threshold system and baseline model, the new and old models are output in parallel and backtracked for verification during the recalibration observation period. After comprehensive verification by indicators such as the consistency rate of state level, the coverage rate of target working condition sub-intervals, and the stability of evaluation results, model switching or migration calibration is implemented. This effectively avoids misjudgments and omissions caused by threshold failure and baseline mismatch after equipment configuration changes, key component replacements, or sensor replacements. It ensures the long-term effectiveness and traceability of the evaluation system throughout the entire equipment life cycle. The overall consistency score before and after the switch can reach 0.85 to 0.93, which significantly reduces the risk of evaluation inaccuracy after configuration changes.

[0081] In some embodiments, when the number of samples in a certain sub-interval of a working condition is insufficient, the system first determines whether the evaluation baseline can be supplemented by binning interpolation of adjacent working conditions. If binning of adjacent working conditions still cannot meet the requirements, the system further determines whether there are similar working condition samples that can be used for migration. If similar working condition samples exist, the evaluation baseline is supplemented by migrating similar working conditions, and the migration source information is retained. If the sample requirements are still not met, the system reverts to the standard reference threshold channel to perform initial screening, and the result is marked as insufficient sample. The insufficient sample indicator includes at least one or more of the following: insufficient sample type, the supplementation method used, and migration source information, for subsequent manual review, sample accumulation, and result traceability.

[0082] like Figure 6 As shown, the insufficient sample rollback and migration source labeling process adopts a progressive safeguard strategy, which specifically includes the following steps: Step 601 involves detecting the number of valid samples. The system checks the number of valid samples in the current stratified evaluation unit u. .

[0083] Step 602 involves determining whether there are insufficient samples; the system then determines the current number of valid samples. Is it less than the preset minimum number of samples? If the sample size is sufficient (i.e., the judgment result is negative), proceed to step 610 and directly use the current evaluation baseline to output the result without any completion operation. If the sample size is insufficient (i.e., the judgment result is positive), proceed to step 603 and start the progressive completion process.

[0084] Step 603 determines whether adjacent load cells meet the interpolation conditions. The system checks whether there are adjacent load cells in the target load cell u across physical dimensions with continuous transitional relationships, such as load, speed, and position, and whether these adjacent load cells have sufficient valid samples. If the interpolation conditions are met, proceed to step 604 to perform adjacent load cell interpolation completion and determine whether the equivalent sample number after interpolation meets the minimum sample requirement. If the requirement is met, proceed to step 608; otherwise, proceed to step 605. If the result of step 603 indicates that the interpolation conditions are not met, proceed directly to step 605.

[0085] Step 605 involves searching for similar working condition samples and performing migration completion. The system searches from the candidate similar working condition set... The similarity to the target working condition u is selected from the data. Not less than the preset similarity threshold Candidate working condition set The baseline center value, scale value, and equivalent sample number are calculated after migration completion using a similarity-weighted method.

[0086] Step 606 determines whether the sample requirements are met after migration completion. The system checks whether the equivalent sample number meets the minimum sample requirement. If the requirement is met, proceed to step 608; if the requirement is not met, proceed to step 607, revert to the standard reference threshold channel for initial screening, abandon the use of the statistical baseline model, and rely solely on the standard reference threshold for judgment to ensure the continuity of evaluation in extreme data loss situations.

[0087] Step 608 generates sample status, completion method and source information. Regardless of whether the result is obtained through the current baseline, adjacent bin interpolation, similar working condition migration or threshold backoff, the system synchronously records metadata such as sample status, completion method used, migration source identifier, similarity and equivalent sample number.

[0088] Step 609 outputs the labeled evaluation results. The system outputs the final status level, along with the sample status, completion method, and source identifier. It clearly indicates whether the result is derived from the current baseline, interpolation completion, migration completion, or threshold rollback, in order to ensure the operability of subsequent manual review and traceability analysis.

[0089] In this embodiment, when the target operating condition sub-interval Valid sample size Less than the preset minimum number of samples When performing this operation, first, interpolation completion of adjacent load conditions is performed. Let the adjacent bins in a certain continuous load dimension be... and The center parameters of its sub-boxes are as follows: and The center parameter of the target operating condition sub-interval is The corresponding baseline center values ​​are respectively and The scale values ​​are respectively and Then the interpolated baseline center value and scale value can be expressed as follows: When interpolation of adjacent load cases is still insufficient to meet the minimum sample requirement, the candidate similar load case set is used. Search for transferable operating conditions. Define the target operating condition sub-interval. Similar working conditions to candidates The similarity is: Based on the aforementioned similarity, the baseline center value, scale value, and equivalent sample number after migration completion can be defined as follows: when and When the migration-completed historical statistical baseline is used for evaluation, if any condition is not met, the process reverts to the standard reference threshold channel for initial screening, and adds a sample insufficiency indicator and migration source information.

[0090] In practice, Only those meeting the preset similarity threshold are included. Candidate similar working conditions, i.e. This is the set after filtering by a similarity threshold; therefore, the equivalent number of samples... The contribution of migrated samples in the process is only accumulated for candidate similar conditions that meet the similarity threshold.

[0091] in, Represents the target operating condition sub-interval The set of candidate similar working conditions; Indicates satisfaction The set of candidate similar working conditions; Represents the target operating condition sub-interval Similar working conditions to candidates Similarity; to Represents the weight coefficients of each component in the similarity model; , , , , , These respectively represent the operation stage, load range, speed range, position or angle range, linkage status, and target mechanism or measuring point group identification; , , Normalized scaling parameters representing differences in load, velocity, and position or angle; and These represent the baseline center value and scale value after migration and completion, respectively; This represents the equivalent number of samples after migration completion; This represents the migration contribution reduction factor; This represents the number of valid samples corresponding to the candidate similar working conditions.

[0092] Specifically, adjacent working condition interpolation is preferably only applicable to working condition dimensions such as load, speed, position, height, or angle, which can be divided according to numerical value or spatial order and have a continuous transition relationship between adjacent sub-divisions. For discrete category fields such as braking state, linkage state, mechanism category, and measuring point group, linear interpolation is not performed; instead, consistency judgment or similarity weighting is used for migration completion. The similarity calculation is preferably based on the joint determination of three or more working condition fields that are actually available on the target equipment. When a certain field is not available on the target equipment, the similarity can be recalculated by normalization on the remaining available fields.

[0093] In the migration parameter settings for similar working conditions, a preset similarity threshold is used. The value can be taken as 0.70 to 0.85, with 0.75 being preferred; migration contribution reduction factor. A value of 0.30–0.80 is acceptable, with 0.50–0.60 being preferred. For sub-sections involving hoisting mechanisms, braking-related components, high loads, high speeds, or other high-risk operating conditions, A value of 0.80–0.85 is recommended. A value of 0.30 to 0.50 is recommended to improve the conservatism of migration completion; for general trend evaluation or low-risk observation scenarios, A value of 0.70 to 0.75 is acceptable. A weight of 0.50 to 0.80 can be used to improve the continuity of evaluation in scenarios with insufficient samples. The weights of the operation stage, load range, speed range, position or angle range, linkage state, target mechanism or measuring point group, etc. in the similarity model can be set to 0.15 to 0.25, 0.20 to 0.30, 0.15 to 0.25, 0.10 to 0.20, 0.10 to 0.15 and 0.10 to 0.20 respectively, and the sum of each weight is 1; when a certain field cannot be obtained from the target device, the weights can be renormalized in the remaining fields.

[0094] When using adjacent bin interpolation or similar working condition migration for completion, the completion results are used to maintain the continuity of the evaluation and should not be interpreted as the target working condition sub-interval having sufficient original samples. The system should simultaneously retain the sample insufficiency indicator, completion method, similarity, migration contribution reduction factor, and equivalent sample number. Evaluation results with sample insufficiency indicators should be statistically analyzed separately when used for model switching, baseline updates, or effect statistics; without manual review or subsequent sample verification, they should not be used as high-confidence baseline update samples.

[0095] In some embodiments, for evaluation results that have insufficient sample flags or remain at a medium-to-high risk level after conflict resolution, the system can send them for manual review or subsequent verification processes and adjust maintenance recommendations accordingly. For steady-state output segments in the current configuration version that do not have insufficient sample flags and whose status level is normal or low risk after conflict resolution, the system can prioritize including historical statistical baseline update samples to improve the quality of baseline updates. The system can also perform long-term accumulation of evaluation results for different stratified evaluation units, forming a correlation between threshold versions, baseline versions, and evaluation records, providing support for subsequent backtracking analysis, version comparison, and maintenance decisions.

[0096] like Figure 7 As shown, the closed-loop process of result classification, manual review, version association, and baseline update specifically includes the following steps: Step 701 involves receiving the evaluation results and related identifiers. The system receives the final status level, anomaly confidence level, and metadata information such as insufficient sample identifier and configuration version identifier after processing through the aforementioned dual-channel judgment and conflict resolution process.

[0097] Step 702 involves result classification and attribute organization. The system performs unified classification and preprocessing on the received evaluation results, extracting key attributes related to baseline update decisions to prepare for subsequent multi-level screening and judgment.

[0098] Step 703 is the first-level screening, determining whether there is an insufficient sample indicator. If the evaluation result has an insufficient sample indicator, it indicates that the result is obtained based on methods such as adjacent bin interpolation, similar working condition migration, or threshold backoff, and the data itself has flaws. In this case, proceed to step 704 to send the result for manual review or subsequent verification, and then proceed to step 705 to mark it as temporarily excluded from baseline update samples to prevent low-confidence samples from contaminating the baseline model. If there is no insufficient sample indicator, it indicates that the data is true and sufficient, and proceed to step 706 for the second-level screening.

[0099] Step 706 is the second-level screening, determining whether it belongs to the current configuration version. The system checks whether the configuration version identifier corresponding to the evaluation result is consistent with the version currently running in the system. If it does not belong to the current configuration version, it indicates that the data comes from an older version before the replacement of key components or sensors, and its data distribution characteristics do not match the current physical state. In this case, proceed to step 709 and mark it as not to be used as a priority baseline update sample. If it belongs to the current configuration version, proceed to step 707 for the third-level screening.

[0100] Step 707 is the third-level screening, determining whether the status level after conflict resolution is normal or low-risk. The system checks whether the final status level falls within the healthy state category. If the status level is medium-risk or high-risk, it indicates that the data contains fault or abnormal information. If used to update the baseline, it would cause the baseline to deviate from the normal state distribution. Therefore, the process proceeds to step 709, marking it as not a priority baseline update sample. If the status level is normal or low-risk, it indicates that the data is a steady-state output collected under healthy equipment conditions. Therefore, the process proceeds to step 708, marking it as a priority sample for baseline update, and further proceeds to step 710 to perform a historical statistical baseline update operation. This high-quality data is used to update and calculate the historical statistical baseline center value, robust scale, and quantile boundaries for the current configuration version.

[0101] Step 711 involves associating the threshold version, baseline version, and evaluation records. Regardless of the final path the evaluation results take, the system will associate and accumulate all key metadata for this evaluation cycle, including the threshold version, baseline version, original evaluation records, conflict resolution process, evidence item values, sample completion methods, and manual review conclusions, forming a complete data chain.

[0102] Step 712 provides support for retrospective analysis, version comparison, and maintenance decisions by linking and accumulating a complete data chain to support subsequent operation and maintenance activities. This includes retrospective analysis of historical faults to trace the evaluation basis at that time, version comparison of the performance of new and old models under different operating conditions, and accurate maintenance planning based on a long-term stable updated baseline model.

[0103] In this embodiment, the present invention takes the hoisting mechanism of a quayside container crane as an example and describes in detail the complete condition evaluation process. First, it receives time-aligned condition monitoring feature values ​​such as vibration, kurtosis, and envelope spectrum, as well as operating condition tags such as hoisting speed uniformity, heavy load, high speed, and high position, and constructs them into an evaluation request data object. and through With the hierarchical evaluation unit index library The system performs precise matching; then it retrieves the standard reference threshold set and historical statistical baseline model bound to the unit, and initiates dual-channel parallel judgment. When the judgment results of the threshold channel and the baseline channel are inconsistent, or when the initial judgment level difference of multiple features within the same unit reaches the preset level difference threshold, the system first identifies the conflict type, then configures the evidence weight vector according to the conflict type and calculates the conflict resolution score and anomaly confidence, and finally outputs the status level and maintenance recommendations. Based on this, for equipment configuration changes and insufficient sample conditions, the system further explains the closed-loop guarantee process, such as configuration version management and recalibration mechanism, adjacent bin interpolation and similar condition migration and completion. The technical effects of this embodiment are: through the structured definition of the evaluation object, precise matching of hierarchical units, and dual-channel parallel judgment, it achieves refined evaluation by working condition; through the correspondence between "conflict type - evidence weight - scoring formula", it stably provides a reliable status level in scenarios such as threshold overrun, baseline drift, and feature level contradiction; and through configuration version management and progressive completion of insufficient samples, it improves the long-term effectiveness and traceability of the evaluation system.

[0104] Example 2 To verify the technical effect of Embodiment 1 of the present invention, the hoisting mechanism of a quay container crane was selected as the verification object. Operating condition signals and multi-source real-time physical signals were synchronously collected during a continuous operating cycle. The operating condition signals included at least the operation stage, lifting load range, speed range, position range, braking state, and linkage state; the real-time physical signals included at least vibration signals, temperature signals, current signals, power signals, and oil status signals. A reference tag set was formed by combining maintenance work orders, inspection records, oil test results, and expert review conclusions.

[0105] Since existing traditional port machinery monitoring systems generally focus only on collecting output signals such as vibration, temperature, and oil levels on the machine side, without simultaneously collecting operating condition signals, this implementation method uses an ablation-based comparative example to verify its effectiveness. The specific settings are as follows: Comparative Example A uses only standard reference thresholds for evaluation, without using operating condition labels, historical statistical baselines, conflict resolution, versioning recalibration, or sample deficiency migration completion; Comparative Example B uses operating condition labels for sub-operating condition threshold evaluation, but does not establish historical statistical baselines; Comparative Example C uses operating condition labels and historical statistical baselines for dual-channel determination, but does not perform conflict type-based evidence weight configuration and conflict resolution, configuration versioning recalibration, or sample deficiency migration completion; Comparative Example D uses operating condition labels, dual-channel determination, and conflict resolution, but does not perform configuration versioning recalibration or sample deficiency migration completion; Example E uses the complete solution of this invention.

[0106] To avoid relying solely on a single accuracy metric, this implementation method selects the following evaluation indicators: consistency rate of state level determination, false alarm rate, false negative rate, stability of the same operating condition level, conflict resolution accuracy, comprehensive score of model switching consistency, and evaluation continuity under insufficient sample scenarios. Specifically, stability of the same operating condition level measures the proportion of evaluation levels remaining consistent or only allowing one level of fluctuation within the same operating condition sub-interval when the equipment state has not undergone substantial changes; evaluation continuity measures the proportion of operating condition sub-intervals that can still continuously output evaluation results under insufficient sample scenarios. The consistency rate of state level determination, false alarm rate, false negative rate, and conflict resolution accuracy are preferably statistically analyzed based on reference labels formed from maintenance work orders, inspection records, oil test results, post-replacement re-inspection conclusions, and expert review results.

[0107] Test results show that, in Comparative Example A, the consistency rate of state level determination is approximately 72%–82%, the false alarm rate is approximately 12%–22%, and the stability of the same operating condition level is approximately 70%–82%; in Comparative Example B, the consistency rate of state level determination is approximately 80%–88%, and the false alarm rate is approximately 8%–15%; in Comparative Example C, the consistency rate of state level determination is approximately 84%–90%, and the false alarm rate is approximately 6%–12%; in Comparative Example D, the consistency rate of state level determination is approximately 87%–93%, and the conflict resolution accuracy is approximately 78%–88%; in Example E, which adopts the complete solution of the present invention, in the preferred embodiment after excluding the results of insufficient sample identification, the consistency rate of state level determination can reach 88%–94%, the false alarm rate can be controlled at 4%–9%, ​​the missed alarm rate can be controlled at 5%–10%, the stability of the same operating condition level can reach 88%–95%, and the conflict resolution accuracy can reach 80%–90%. The percentage results above are preferred implementation results after sufficient reference labels, sampling period and sensor configuration meet the conditions of the embodiment, and obvious invalid samples are excluded. They are used to illustrate the trend of technical effect and do not constitute an absolute limitation on all port crane models and all working conditions.

[0108] Since Comparative Example D already includes different operating conditions, dual channels, and conflict resolution, its consistency rate with Example E in determining the state level is limited under normal scenarios with sufficient samples. The main advantages of Example E are reflected in configuration changes, under insufficient sample conditions, and in long-term evaluation stability.

[0109] Furthermore, in recalibration scenarios following reducer, bearing, or sensor replacement, this invention sets a recalibration observation period and performs consistency verification between the old and new models. For typical operating condition sub-intervals with high frequency of occurrence, preliminary verification can be completed within 30-80 work cycles. For target operating condition sets that need to cover multiple combinations of loads, speeds, positions, and operating stages, recalibration observation and stable switching can typically be completed within 3-7 days, with a comprehensive consistency score of 0.85-0.93 before and after the switch. Compared to the control example that does not perform versioned recalibration, this invention significantly reduces the risk of misjudgment due to threshold failure and baseline mismatch after configuration changes.

[0110] In scenarios with insufficient samples, the evaluation continuity of the target working condition sub-interval is typically 60%–80% without migration completion; when only threshold backoff is used, the evaluation continuity is approximately 75%–88%; after adopting the adjacent working condition binning interpolation, similar working condition migration, and source identification mechanism of this invention, the evaluation continuity can reach 90%–97%, and the output results retain the completion method and migration source information, which is beneficial for subsequent sample accumulation, manual review, and traceability analysis.

[0111] Therefore, this invention is not a simple parallel application of conventional techniques such as monitoring, thresholding, baselines, and recalibration. Rather, it addresses the characteristics of port cranes, such as strong coupling between input conditions and output responses, discrete distribution of operating conditions, and frequent configuration changes. It constructs a closed-loop state evaluation mechanism consisting of control system input condition constraints, multi-source real-time physical signal acquisition, dual-channel parallel judgment, evidence weight configuration and conflict resolution based on conflict type, versioned recalibration, and sample insufficiency migration and completion. This results in superior comprehensive technical effects in terms of evaluation accuracy, level stability, adaptability to configuration changes, and engineering traceability.

[0112] From the perspective of feasibility and technical effect attribution, the above comparison relationship illustrates the synergistic effect among various technical features: the hierarchical evaluation unit limits the range of comparable samples; the dual-channel judgment detects conflicts between the standard threshold and individual baselines; the evidence weight vector configured based on the conflict type achieves differential resolution; configuration version management avoids baseline mismatch after component replacement; and progressive supplementation for insufficient samples ensures continuous output under sparse operating conditions. The above steps are interconnected and work together on the same evaluation request data object. and its corresponding hierarchical evaluation units It is not a simple superposition of independent conventional modules. The data transfer relationship between each step is as follows: the evaluation request data object determines the hierarchical evaluation unit, the hierarchical evaluation unit determines the calling range of threshold, baseline and weight vector, and the conflict resolution results are used as the basis for baseline update and configuration version switch after manual review, sample screening or model switching verification. Therefore, there is a clear input-output dependency relationship between each step.

[0113] In another embodiment, the present invention also provides a system, a computer-readable storage medium, and a computer program product corresponding to the above-described method. The system includes at least a hierarchical evaluation unit construction module, a dual-channel parallel decision module, a conflict type identification module, an evidence weight configuration and conflict resolution module, and a configuration version management and self-calibration module. The system can be deployed on an edge computing gateway or industrial computer for real-time evaluation and collaborates with a cloud platform to complete version management, sample accumulation, model updates, and long-term statistical baseline maintenance. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the method steps of the present invention. The computer program product includes program instructions, which, when executed by a processor, are used to implement the method steps of the present invention.

[0114] Specifically, the system includes: The hierarchical evaluation unit construction module is used to acquire the operating condition signals and status monitoring characteristic values ​​of the port crane. It constructs an evaluation request data object from the current evaluation object, which consists of equipment type identifier, target mechanism identifier, measuring point group identifier, operating condition sub-interval label, and status monitoring characteristic value vector. It generates a search key using the equipment type identifier, target mechanism identifier, measuring point group identifier, and operating condition sub-interval label, and matches the corresponding hierarchical evaluation unit in the preset hierarchical evaluation unit index library. The operating condition sub-interval includes at least load interval, speed interval, position interval, and operation stage. The dual-channel parallel determination module is used to retrieve the standard reference threshold set and historical statistical baseline model bound to the hierarchical evaluation unit in parallel, and compare the state monitoring feature value with the standard reference threshold and the historical statistical baseline respectively to obtain the threshold channel determination result and the baseline channel determination result. The conflict type identification module is used to determine the judgment results of the status monitoring feature values ​​under the standard reference threshold channel and the historical statistical baseline channel, and to identify threshold over-limit conflict, baseline drift conflict and multi-feature level contradiction conflict according to three rules: threshold channel abnormal and baseline channel normal, threshold channel normal and baseline channel abnormal, and the initial judgment level difference of multiple features in the same hierarchical evaluation unit is not less than the preset level difference threshold. The evidence weight configuration and conflict resolution module is used to select evidence items to participate in the calculation according to the conflict type and configure the corresponding weight vector. The evidence items include at least one or more of the following: additional diagnostic features, cross-modal consistency, abnormal duration, recurrence frequency and persistent drift evidence. It is also used to calculate the conflict resolution score based on the selected evidence items and their weights, map the conflict resolution score to the abnormal confidence level, and output the final state level accordingly. The configuration version management and self-calibration module is used to establish a new configuration version and freeze the original threshold and baseline model when a configuration change event or a continuous drift event is detected; to collect steady-state samples of the new configuration in each operating condition sub-interval to establish a candidate new baseline model and start the recalibration observation period; to output the evaluation results of the new and old models in parallel during the observation period and to perform backtracking verification; and to perform model switching when the state level consistency rate, target operating condition sub-interval coverage, evaluation result stability and consistency comprehensive score all meet the corresponding preset thresholds.

[0115] This invention proposes a hierarchical state evaluation and self-calibration method and system for port cranes, addressing the characteristics of port cranes such as strong coupling between input conditions and output responses, discrete distribution of conditions, and frequent configuration changes. The evaluation object is decomposed layer by layer into equipment, mechanisms, measuring points, and sub-intervals of operating conditions. A dual-channel parallel judgment mechanism based on standard reference thresholds and historical statistical baselines is established, and an evidence weight configuration and conflict resolution process based on conflict type are introduced to resolve judgment contradictions. Simultaneously, a configuration version management and self-calibration mechanism are configured. Parallel observation of old and new models, backtracking verification, and step-by-step switching are used to address replacements of key equipment components or long-term drift. In cases of insufficient samples, adjacent sub-bin interpolation, similar condition migration, and threshold rollback supplementation strategies are employed to ensure evaluation continuity. Test results show that, under sufficient verification with reference labels, the consistency rate of state level determination of this scheme can reach 88%–94%, the false alarm rate can be controlled at 4%–9%, ​​the false alarm rate can be controlled at 5%–10%, the stability of the state level under the same working condition can reach 88%–95%, the correctness of conflict resolution can reach 80%–90%, the comprehensive score of recalibration consistency can reach 0.85–0.93, and the evaluation continuity can reach 90%–97% in scenarios with insufficient samples. It has achieved better comprehensive technical results in terms of evaluation accuracy, state level stability, adaptability to configuration changes, and engineering traceability.

[0116] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the invention by those skilled in the art without departing from the spirit and essence of the invention. Such modifications or substitutions should all fall within the scope of the invention, or any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the invention should be covered within the protection scope of the invention. Therefore, the protection scope of the invention should be determined by the scope of the claims.

Claims

1. A method for evaluating and self-calibrating the working conditions of port cranes, characterized in that, The method includes: Step S1: Obtain the operating condition signals and condition monitoring characteristic values ​​of the port crane, and construct an evaluation request data object from the current evaluation object, which consists of equipment type identifier, target mechanism identifier, measuring point group identifier, operating condition sub-interval label, and condition monitoring characteristic value vector. ;by The equipment type identifier, target organization identifier, measurement point group identifier, and working condition sub-interval label are used to generate the search key. And match the corresponding hierarchical evaluation unit in the preset hierarchical evaluation unit index library. ; Step S2: For the hierarchical evaluation unit, retrieve its bound standard reference threshold set and historical statistical baseline model in parallel, compare the state monitoring feature value with the standard reference threshold and the historical statistical baseline respectively, obtain the threshold channel judgment result and the baseline channel judgment result, and form the initial state level of each state monitoring feature; Step S3: Determine the judgment results of the status monitoring feature values ​​in the threshold channel and baseline channel respectively, and identify the conflict type; Step S4: For the identified conflict type, select evidence items to participate in the calculation from the preset evidence item set, and configure corresponding weight vectors for the selected evidence items; calculate the conflict resolution score according to the selected evidence items and their weights, map the conflict resolution score to the anomaly confidence level, and output the final state level according to the anomaly confidence level and the initial state level. Step S5: When a configuration change event or a persistent drift event is detected, a new configuration version is established and the original threshold and baseline model are frozen; steady-state samples of the new configuration in each operating condition sub-interval are collected to establish a candidate new baseline model, and a recalibration observation period is started; during the observation period, the evaluation results of the new and old models are output in parallel and backtracking verification is performed; when the candidate new baseline model meets the model switching conditions, the model switching is performed.

2. The method according to claim 1, characterized in that, The current evaluation object in step S1 is the evaluation request data object. , ,in For equipment type identification, Identify the target organization. For grouping and identifying the measurement points, This is the label vector for the sub-interval of the operating condition. The status monitoring feature value vector; the hierarchical evaluation unit By unique identifier key Determined; when At that time, the current evaluation object is mapped to the hierarchical evaluation unit. The measurement point grouping identifier and the working condition sub-interval label vector are determined according to the preset measurement point grouping rules and the working condition binning rules, respectively. The continuous working condition label in the working condition sub-interval label vector is determined by the preset binning boundary, and the discrete state label is determined by the control system state word. When there is no precisely matched hierarchical evaluation unit, an unmatched identifier is output, or the sample insufficient processing flow is entered according to the preset similar working condition rules. The evaluation request data object, the lookup key, and the hierarchical evaluation unit index are all data structures that are read and called by a computer, and the matching is based on the matching condition that the field values ​​in the lookup key are consistent.

3. The method according to claim 1, characterized in that, The historical statistical baseline model in step S2 is established based on the output segments corresponding to steady-state samples in multiple work cycles under the same or similar working condition labels, and is characterized by quantile boundaries and / or robust center-scale models. The historical statistical baseline model outputs at least two of the following: baseline center value, quantile boundary, robust scale, and baseline deviation. The historical statistical baseline deviation is determined based on the normalized deviation of the current state monitoring feature value from the baseline center value, robust scale, and / or quantile boundary of the historical statistical baseline model, and is used to characterize the degree of deviation of the current state monitoring feature value from the historical statistical baseline model. The similar operating condition label refers to an operating condition label that has the same equipment type, target mechanism and measurement point grouping, and whose bin distance or similarity in the continuous operating condition dimension meets the preset conditions.

4. The method as described in claim 3, characterized in that, Step S3 specifically includes: The threshold channel determination result is determined based on the threshold type of the status monitoring feature; when the standard reference threshold is the upper limit threshold, the current feature value being greater than the upper limit threshold is considered an abnormal threshold channel; when the standard reference threshold is the lower limit threshold, the current feature value being less than the lower limit threshold is considered an abnormal threshold channel; when the standard reference threshold is the normal range threshold, the current feature value falling outside the normal range threshold is considered an abnormal threshold channel; the standard reference threshold is configured to correspond to the status monitoring feature. set up The preset baseline anomaly detection threshold is used to determine the baseline deviation based on historical statistics. and The relationship determines the baseline channel determination result; when Not less than When, determine that the baseline channel is abnormal; when Less than At that time, it was confirmed that the baseline channel was normal; The standard reference threshold is used to compare the current state monitoring feature value. The comparison is used to determine the threshold channel determination result; the baseline anomaly determination threshold. Used for deviation from historical statistical baseline The baseline channel determination result is determined by comparison; the standard reference threshold is compared with... They correspond to different judgment channels, and the two are not compared with each other; When the threshold channel of the same state monitoring feature is abnormal while the baseline channel is normal, it is identified as a threshold out-of-bounds conflict; when the threshold channel of the same state monitoring feature is normal while the baseline channel is abnormal, it is identified as a baseline drift conflict. When the same stratified evaluation unit Inside The state level is obtained by initially mapping each state monitoring feature through the threshold channel and / or baseline channel. to satisfy Not less than the preset grade difference threshold When the conflict is identified as a multi-feature level contradiction, the frequency of recurrence and the duration of abnormality are not considered as conditions for the establishment of this conflict type, but as evidence items for subsequent conflict resolution scores. Before calculating the difference between the maximum and minimum values, the state level is uniformly encoded as an integer level according to the direction of increasing risk.

5. The method as described in claim 4, characterized in that, Step S4 specifically includes: The evidence items include additional diagnostic features, cross-modal consistency, duration of abnormality, frequency of recurrence, and evidence of persistent drift; In the event of a conflict, the target institution is given additional diagnostic evidence. Evidence of cross-modal consistency is Evidence for abnormal duration is Evidence of recurrence frequency is Evidence of continuous drift is The penalty for insufficient samples is The conflict type is The system is based on Call the preset evidence weight vector Then calculate the conflict resolution score. and abnormal confidence They are defined as follows: ； ； in, For bias terms, , , , , These are the positive weighting coefficients for additional diagnostic feature evidence, cross-modal consistency evidence, abnormal duration evidence, recurrence frequency evidence, and persistent drift evidence, respectively. The deduction weight corresponding to the insufficient sample penalty item; The sources of evidence for additional diagnostic features and cross-modal consistency are determined based on the target institution and actual sensor configuration, and are bound to the current configuration version, hierarchical evaluation unit, and operating condition sub-interval; the evidence items are normalized quantitative evidence values. The evidence items are determined based on data from the same configuration version, the same hierarchical evaluation unit, and the same or similar operating condition sub-intervals; different conflict types correspond to different evidence item priorities or weight configurations, and for each conflict type, the weight of the positive evidence item is... , , , and Normalization was applied to make the sum equal to 1; insufficient sample penalty term. by Form is included as a deduction item in the conflict resolution score calculation. It does not participate in the weight normalization of positive evidence items; When a certain evidence item is not selected, is not configured with the corresponding sensor, or is unusable in the current configuration version, hierarchical evaluation unit, and working condition sub-interval, the weight of that evidence item is set to 0; the remaining positive evidence items participating in the calculation are re-normalized so that the sum of the weights of the positive evidence items participating in the calculation is 1.

6. The method as described in claim 5, characterized in that, The model switching conditions in step S5 include at least the state level consistency rate, the target working condition sub-interval coverage rate, the stability of the evaluation results, and the comprehensive consistency score. The conditions for model switching are met when the consistency rate of state level, the coverage rate of target working condition sub-interval, the stability of evaluation results, and the comprehensive score of consistency all meet the corresponding preset thresholds; the stability of evaluation results is determined based on the state level jumps between adjacent evaluation periods during the recalibration observation period.

7. The method as described in claim 6, characterized in that, The method further includes: When the number of samples in a certain sub-interval of a working condition is insufficient, the system first determines whether the evaluation baseline can be supplemented by binning interpolation of adjacent working conditions. If binning of adjacent working conditions still cannot meet the requirements, it determines whether there are similar working condition samples for migration. If similar working condition samples exist, the evaluation baseline is supplemented by migrating similar working conditions, and the migration source information is retained. If the sample requirements are still not met, the system reverts to the standard reference threshold channel to perform initial screening, and marks the result as a sample insufficiency flag. The sample insufficiency flag includes at least one or more of the following: sample insufficiency type, the supplementation method used, and migration source information. When the number of samples meets the requirements and the current evaluation baseline is used to output the results, the system records information such as sufficient samples, no completion method used, and the source of the current evaluation baseline; the adjacent working condition bin interpolation is performed when there are adjacent bins with valid samples in the continuous working condition dimension of the target working condition sub-interval; the similar working condition migration is performed when the similarity between the candidate working condition sub-interval and the target working condition sub-interval is not less than a preset similarity threshold; when the evaluation results are obtained by using adjacent working condition bin interpolation or similar working condition migration, the system records the completion method, the source of the candidate working condition, and the sample status identifier, and the migration completion result retains the insufficient sample identifier.

8. A port crane condition-based hierarchical status evaluation and self-calibration system, characterized in that, The system is configured to perform the method as described in any one of claims 1 to 7, the system comprising: The hierarchical evaluation unit construction module is used to acquire the operating condition signals and status monitoring characteristic values ​​of port cranes. It constructs an evaluation request data object from the current evaluation object, which consists of equipment type identifier, target mechanism identifier, measuring point group identifier, operating condition sub-interval label, and status monitoring characteristic value vector. ;by The equipment type identifier, target organization identifier, measurement point group identifier, and working condition sub-interval label are used to generate the search key. And match the corresponding hierarchical evaluation unit in the preset hierarchical evaluation unit index library. ; The dual-channel parallel judgment module is used to retrieve the standard reference threshold set and historical statistical baseline model bound to the hierarchical evaluation unit in parallel, compare the state monitoring feature value with the standard reference threshold and the historical statistical baseline respectively, obtain the threshold channel judgment result and the baseline channel judgment result, and form the initial judgment state level of each state monitoring feature. The conflict type identification module is used to determine the judgment results of the status monitoring feature values ​​under the threshold channel and the baseline channel, and to identify the conflict type. The evidence weight configuration and conflict resolution module is used to select evidence items for calculation from a preset set of evidence items for the identified conflict type, and configure corresponding weight vectors for the selected evidence items; calculate the conflict resolution score based on the selected evidence items and their weights, map the conflict resolution score to anomaly confidence, and output the final state level based on the anomaly confidence and the initial state level. The configuration version management and self-calibration module is used to establish a new configuration version and freeze the original threshold and baseline model when a configuration change event or a continuous drift event is detected; to collect steady-state samples of the new configuration in each operating condition sub-interval to establish a candidate new baseline model and start the recalibration observation period; to output the evaluation results of the new and old models in parallel during the observation period and to perform backtracking verification; and to perform model switching when the candidate new baseline model meets the model switching conditions.

9. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method of any one of claims 1 to 7.

10. A computer program product comprising program instructions which, when executed by a processor, are used to implement the method of any one of claims 1 to 7.

Images