Portal crane luffing structure state monitoring system

Abstract

The present application relates to the technical field of port machinery state monitoring, and discloses a luffing structure state monitoring system of a port portal crane, comprising: a sensing and collecting unit; a data synchronization and preprocessing unit for data synchronization and preprocessing of vibration data and working condition information; a working condition partition unit for working condition partition of the vibration data segment according to the working condition information; a sample screening and modeling unit for establishing a health baseline model based on historical samples under a health state for each working condition partition label; an abnormality evaluation unit; a health index generation unit for mapping the abnormality score into a health index; and an early warning decision unit. The luffing structure state monitoring system is oriented to the hinge shaft, hinge shaft supporting components and their associated force transmission structure in the luffing structure of the port portal crane, and is used for state monitoring in view of the low-speed, semi-rotation and non-steady-state operation characteristics, and does not rely on the extraction of the rotating mechanical characteristic frequency under the stable rotating speed condition.

Description

Technical Field

[0001] This invention relates to the field of port machinery condition monitoring technology, specifically a port gantry crane luffing structure condition monitoring system. Background Technology

[0002] Port gantry cranes are widely used in bulk cargo handling and general cargo handling operations. Their luffing structure allows for boom angle adjustment, thereby changing the working radius. The luffing structure includes hinge shafts, hinge shaft support components, and their associated force transmission structures, which bear periodic loads and directional switching loads during operation. In existing technologies, the condition monitoring of these components typically employs manual periodic inspections, or monitoring methods based on temperature signals, total vibration values, or fixed thresholds, as well as spectral analysis methods for rotating machinery to identify bearing conditions.

[0003] However, the hinge shaft and its support components perform low-speed, limited-angle reciprocating or semi-rotational motion during amplitude change, exhibiting unsteady operating characteristics. Furthermore, the angular velocity is not constant due to the accompanying start-up, braking, and direction-switching processes, failing to meet the spectral analysis conditions based on stable rotational speed. This makes it difficult to extract stable characteristic frequencies for state identification. The vibration response varies significantly under different amplitude angles, loads, and motion directions, easily leading to misjudgments when using a uniform threshold or model for monitoring. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a monitoring system for the luffing structure of port gantry cranes, which solves the problem that existing technologies often result in significant differences in vibration response under different luffing angles, loads, and directions of motion, and are prone to misjudgment when using a uniform threshold or model for monitoring.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a port gantry crane luffing structure status monitoring system, comprising: The sensing and acquisition unit is used to synchronously acquire vibration data from multiple vibration measuring points arranged along the hinge shaft support and its force transmission path, and to synchronously acquire the working condition information corresponding to the vibration data. The data synchronization and preprocessing unit is used to time-align the vibration data and working condition information from multiple measurement points, and to perform DC removal, abnormal pulse removal, bandpass filtering and normalization on the vibration data. It also divides the data according to a preset sliding time window to obtain vibration data segments corresponding to multiple time windows. The time-frequency transformation unit is used to perform time-frequency transformation on the vibration data segment to obtain the corresponding time-frequency diagram; The working condition partitioning unit is used to discretize and partition the vibration data segment corresponding to the time window based on at least two working condition variables, and generate working condition partitioning labels. The sample screening and modeling unit is used to establish a health baseline model based on historical samples under a healthy state for each working condition zone label, and to establish a corresponding multi-measurement point synchronization relationship baseline. The anomaly assessment unit is used to input real-time vibration data segments and their time-frequency graphs into the health baseline model of the corresponding working condition zone label, and calculate the anomaly score by combining the multi-measuring point synchronization relationship baseline. A health index generation unit is used to normalize the abnormal scores and map them into a health index; The early warning decision unit is used to perform trend assessment and graded early warning based on the health index.

[0006] Preferably, the plurality of vibration measuring points are arranged at key rigid parts on the relevant positions of the hinge support and its force transmission path.

[0007] Preferably, in the sensing and acquisition unit, the sampling frequency of the synchronous vibration acquisition is not less than 25000Hz, and multiple vibration measurement points are synchronously acquired using a unified clock or hardware synchronous triggering.

[0008] Preferably, the operating condition information includes at least two of the following: luffing angle, angular velocity, luffing / lowering direction, load weight, luffing drive current, braking status, and start / stop status.

[0009] Preferably, the data synchronization and preprocessing unit is used to perform DC removal processing, abnormal pulse removal processing, bandpass filtering processing and normalization processing on the vibration data, and to segment the vibration data according to a preset sliding time window to obtain multiple vibration data segments.

[0010] Preferably, the health baseline model established by the sample screening and modeling unit includes: a one-dimensional autoencoder based on the original vibration sequence and a two-dimensional autoencoder based on the time-frequency diagram.

[0011] Preferably, the anomaly assessment unit is used to obtain an anomaly score by weighted fusion of the reconstruction error based on the original vibration sequence, the reconstruction error based on the time-frequency diagram, the deviation of potential features, and the deviation of synchronization consistency of multiple measurement points.

[0012] Preferably, the health index is a value between 1 and 100, and the health index is obtained by normalizing the abnormal score and using a monotonically decreasing mapping function.

[0013] Preferably, a method for monitoring the status of the luffing structure of a port gantry crane includes the following steps: S1: Simultaneously collect vibration data from multiple vibration measurement points and collect the corresponding operating condition information. S2: Synchronize and preprocess the vibration data and working condition information to obtain vibration data segments corresponding to multiple time windows. Perform time-frequency transformation on the vibration data segments corresponding to each time window to generate a time-frequency diagram. S3: Divide the vibration data segment into working condition partitions based on the working condition information to obtain multiple working condition partition labels; S4: For each working condition zone label, establish a health baseline model based on historical samples under healthy conditions, and establish a multi-measurement point synchronization relationship baseline; S5: Input the real-time acquired vibration data segments and their time-frequency graphs into the health baseline model of the corresponding working condition zone label, and calculate the anomaly score by combining the multi-measuring point synchronization relationship baseline. S6: Map the abnormal scores to a health index; S7: Conduct trend assessment and graded early warning based on the aforementioned health index.

[0014] Preferably, an electronic device includes a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the above-described method for monitoring the status of the luffing structure of a port gantry crane.

[0015] This invention provides a status monitoring system for the luffing structure of a port gantry crane. It has the following beneficial effects: 1. This invention addresses the hinge shaft, hinge shaft support components, and associated force transmission structures in the luffing structure of port gantry cranes, and performs state monitoring for low-speed, semi-rotational, and unsteady-state operation characteristics, extracting the characteristic frequencies of rotating machinery under stable speed conditions.

[0016] 2. This invention combines synchronous vibration acquisition with working condition information acquisition, and forms working condition zoning labels based on the working condition information. It establishes a health baseline model for different working condition zoning, thereby reducing the abnormal judgment bias caused by the mixing of different working conditions.

[0017] 3. This invention uses multiple vibration measurement points for synchronous vibration acquisition and combines the synchronous relationship of multiple measurement points for evaluation, which can simultaneously reflect local vibration changes and synchronous changes along the force transmission path.

[0018] 4. This invention simultaneously utilizes the original vibration sequence and time-frequency diagram to establish a healthy baseline model, and integrates the original vibration reconstruction error, time-frequency diagram reconstruction error, potential feature deviation, and multi-measurement point synchronization consistency deviation to form an anomaly score, thereby enhancing the ability to identify abnormal states under complex working conditions. Attached Figure Description

[0019] Figure 1 This is an intentional representation of the health index grading of the present invention; Figure 2This is a schematic diagram of the arrangement and acquisition relationship of vibration measuring points in this invention; Figure 3 This is a schematic diagram of the data synchronization and preprocessing process of the present invention; Figure 4 This is a schematic diagram illustrating the relationship between the original vibration sequence and the time-frequency diagram generated in this invention; Figure 5 This is a schematic diagram of the working condition partitioning process of the present invention; Figure 6 This is a schematic diagram of the anomaly assessment process of the present invention; Figure 7 This is a system architecture diagram of the present invention. Detailed Implementation

[0020] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see the appendix Figure 1 To be continued Figure 7 This invention provides a port gantry crane luffing structure status monitoring system, comprising: The sensing and acquisition unit is used to synchronously acquire vibration data from multiple vibration measuring points arranged along the hinge shaft support and its force transmission path, and to synchronously acquire the working condition information corresponding to the vibration data. The data synchronization and preprocessing unit is used to time-align the vibration data and working condition information from multiple measurement points, and to perform DC removal, abnormal pulse removal, bandpass filtering and normalization on the vibration data. It also divides the data according to a preset sliding time window to obtain vibration data segments corresponding to multiple time windows. The time-frequency transformation unit is used to perform time-frequency transformation on the vibration data segment to obtain the corresponding time-frequency diagram; The working condition partitioning unit is used to discretize and partition the vibration data segment corresponding to the time window based on at least two working condition variables, and generate working condition partitioning labels. The sample screening and modeling unit is used to establish a health baseline model based on historical samples under a healthy state for each working condition zone label, and to establish a corresponding multi-measurement point synchronization relationship baseline. The anomaly assessment unit is used to input real-time vibration data segments and their time-frequency graphs into the health baseline model of the corresponding working condition zone label, and calculate the anomaly score by combining the multi-measuring point synchronization relationship baseline. A health index generation unit is used to normalize the abnormal scores and map them into a health index; The early warning decision unit is used to perform trend assessment and graded early warning based on the health index.

[0022] Specifically, the port gantry crane luffing structure status monitoring system of the present invention is applied to the luffing structure of a port gantry crane, and the monitoring objects are the hinge shaft, hinge shaft support components and their associated force transmission structures. The port gantry crane luffing structure status monitoring system includes a sensor acquisition unit, a data synchronization and preprocessing unit, a working condition zoning unit, a sample screening and modeling unit, an anomaly assessment unit, a health index generation unit, an early warning decision-making unit, and a storage and display unit. Each unit is connected via an industrial communication bus or data interface, and sequentially completes synchronous vibration acquisition, working condition information acquisition, data processing, model calculation, status assessment, result output, and historical storage according to the data flow direction.

[0023] The sensing and acquisition unit is used to synchronously acquire vibration data from multiple vibration measuring points and collect corresponding operating condition information. The unit includes vibration sensors, operating condition sensors, synchronous acquisition hardware, and an acquisition control module. Vibration sensors are installed at multiple vibration measuring points, while operating condition sensors are positioned at locations for amplitude regulation drive, amplitude regulation angle detection, load detection, braking detection, and environmental detection. The synchronous acquisition hardware is electrically connected to both the vibration sensors and the operating condition sensors. The acquisition control module sends sampling control commands to the synchronous acquisition hardware and receives sampling results.

[0024] The data synchronization and preprocessing unit is connected to the sensing and acquisition unit. It receives vibration data and operating condition information, and performs data synchronization and preprocessing on the vibration data and operating condition information to obtain vibration data segments corresponding to multiple time windows. The data synchronization and preprocessing unit includes a time alignment module, a filtering module, a normalization module, a window segmentation module, and an effective window filtering module. The time alignment module performs time alignment on multi-channel vibration data and operating condition information under a unified time reference; the filtering module performs DC removal, abnormal pulse removal, and bandpass filtering on the vibration data; the normalization module normalizes the amplitude of the vibration data; the window segmentation module segments the vibration data according to a preset sliding time window; and the effective window filtering module filters out time windows corresponding to rapid switching time periods based on the operating condition information, or marks such time windows as transition windows.

[0025] The load condition zoning unit, connected to the data synchronization and preprocessing unit, is used to zon the vibration data segments based on load condition information, generating multiple load condition zoning labels. The load condition zoning unit includes a load condition variable extraction module, a load condition interval division module, and a label generation module. The load condition variable extraction module extracts amplitude angle, angular velocity, start / end direction, load weight, amplitude variable drive current, braking status, and start / stop status from the load condition information. The load condition interval division module discretizes these load condition variables. The label generation module combines the discretized load condition variables to generate load condition zoning labels and associates these labels with the corresponding vibration data segments.

[0026] The sample screening and modeling unit is connected to the operating condition zoning unit and is used to build a healthy baseline model for each operating condition zoning label based on historical samples under healthy conditions. The sample screening and modeling unit includes a sample statistics module, a common operating condition screening module, a healthy sample confirmation module, a model training module, and a model management module. The sample statistics module is used to statistically analyze the number of samples, cumulative duration, and frequency of occurrence corresponding to each operating condition zoning label; the common operating condition screening module is used to determine the most common operating condition zoning according to preset screening rules; the healthy sample confirmation module is used to screen historical samples under healthy conditions; the model training module is used to train the healthy baseline model according to the operating condition zoning label; and the model management module is used to save the correspondence between the healthy baseline model and the operating condition zoning label, model parameters, and model version information.

[0027] The anomaly assessment unit is connected to the sample screening and modeling unit. It is used to input real-time acquired vibration data segments into the health baseline model corresponding to the operating condition zoning label and calculate the anomaly score. The anomaly assessment unit includes a model matching module, a raw vibration assessment module, a time-frequency graph assessment module, a latent feature assessment module, a synchronization consistency assessment module, and an anomaly fusion module. The model matching module is used to determine the corresponding health baseline model based on the operating condition zoning label of the current time window; the raw vibration assessment module is used to calculate the raw vibration reconstruction error; the time-frequency graph assessment module is used to calculate the time-frequency graph reconstruction error; the latent feature assessment module is used to calculate the latent feature deviation; the synchronization consistency assessment module is used to calculate the synchronization consistency deviation of multiple measurement points; the anomaly fusion module is used to fuse the above calculation results and output the anomaly score. The multi-measurement point synchronization consistency deviation is calculated based on at least one of the correlation coefficient matrix, cross-spectral consistency, or energy distribution matrix between multiple measurement points. The latent feature deviation is the degree of deviation between the position of the sample in the latent feature space of the health baseline model and the center of the healthy sample distribution.

[0028] The health index generation unit is connected to the anomaly assessment unit and is used to map anomaly scores to a health index. The health index generation unit includes an anomaly normalization module, a mapping calculation module, and a curve generation module. The anomaly normalization module normalizes the anomaly scores; the mapping calculation module calculates the health index based on a preset mapping relationship; and the curve generation module generates a trend curve for the health index based on a continuous time window.

[0029] The early warning decision unit is connected to the health index generation unit and is used for trend assessment and graded early warning based on the health index. The early warning decision unit includes a threshold determination module, a persistence determination module, a trend determination module, a consistency determination module, and an alarm output module. The threshold determination module is used to determine the early warning level based on the interval corresponding to the health index; the persistence determination module is used to determine whether an alarm is valid based on the duration of the interval in which the health index is located within a continuous time window and the continuous window count; the trend determination module is used to determine the degree of trend deterioration based on the rate of decline of the health index; the consistency determination module is used to determine cross-measuring point consistency based on the synchronous anomalies of multiple vibration measuring points; and the alarm output module is used to output the early warning level, abnormal measuring points, suggested actions, and interlocking protection signals.

[0030] The storage and display unit is connected to the sensor acquisition unit, sample screening and modeling unit, anomaly assessment unit, health index generation unit, and early warning decision-making unit, respectively. It stores models, historical samples, vibration data segments, time-frequency graphs, operating condition zoning labels, anomaly scores, health indices, trend curves, alarm records, and model update records. The storage and display unit also outputs health indices, trend curves, operating condition zoning labels, anomaly measurement points, alarm levels, and suggested actions to the display interface for use by operators and maintenance personnel.

[0031] Furthermore, multiple vibration measuring points are arranged at key rigid parts of the hinge support and its force transmission path.

[0032] Specifically, in this embodiment, multiple vibration measuring points are set at key rigid locations along the hinge shaft support and its force transmission path. These key rigid locations include at least measuring points near the supports on both sides of the hinge shaft, and at least a portion of the boom root connection location, connecting plate location, ear plate location, and support rigid location. Eleven vibration measuring points are used, each equipped with a vibration sensor, which is an acceleration sensor. These eleven measuring points are distributed along the hinge shaft support and its force transmission path, ensuring that local vibration changes at the hinge shaft support and the transmission of these local vibration changes along the force transmission path can be recorded under synchronous sampling conditions.

[0033] Furthermore, in the sensing and acquisition unit, the sampling frequency of synchronous vibration acquisition is not less than 25000Hz, and multiple vibration measurement points are synchronously acquired using a unified clock or hardware synchronous triggering. The operating condition information includes at least two of the following: luffing angle, angular velocity, luffing / lowering direction, load weight, luffing drive current, braking status, and start / stop status.

[0034] Specifically, when the sensing and acquisition unit synchronously acquires vibration data from 11 vibration measurement points, the sampling frequency is no less than 25000Hz. Synchronous vibration acquisition uses a unified clock or hardware synchronization triggering method to ensure that the vibration data output from multiple vibration measurement points are on the same time reference. The vibration data is stored in the form of a multi-channel synchronous sequence, with each channel corresponding to one vibration measurement point. Operating condition information is acquired synchronously with the vibration data. This operating condition information includes at least one of the following: luffing angle, angular velocity, luffing / lowering direction, load weight, luffing drive current, braking status, start / stop status, wind speed, and ambient temperature. The luffing angle is output by an angle sensor or encoder; the angular velocity is calculated from the angle sensor or encoder data; the load weight is output by a weighing device; the luffing drive current is acquired by the drive circuit; the braking status and start / stop status are output by the control system; and the wind speed and ambient temperature are output by an environmental monitoring device.

[0035] Furthermore, the data synchronization and preprocessing unit is used to perform DC removal, abnormal pulse removal, bandpass filtering, and normalization on the vibration data, and to segment the vibration data according to a preset sliding time window to obtain multiple vibration data segments.

[0036] Specifically, the data synchronization and preprocessing unit first performs time alignment on the vibration data and operating condition information. For vibration data, alignment is performed according to the synchronization sampling timestamp; for operating condition information, it is mapped to the corresponding vibration data time axis according to its sampling timestamp. After alignment, multi-channel vibration data and operating condition information are obtained, represented by a unified time base. For operating condition information with a sampling frequency lower than the vibration data sampling frequency, interpolation, resampling, or interval assignment are performed according to the time correspondence, so that each time window corresponds to a specific operating condition information.

[0037] After time alignment, the data synchronization and preprocessing unit preprocesses the vibration data. DC removal is used to eliminate DC bias in the vibration data; abnormal pulse removal is used to remove isolated pulses caused by electromagnetic interference, transient impacts, or transmission anomalies; bandpass filtering is used to retain vibration components within a preset frequency band; and normalization is used to standardize the amplitude scale between different vibration measurement points. After these processes, preprocessed vibration data is obtained. The preprocessing sequence is: time alignment, DC removal, abnormal pulse removal, bandpass filtering, and normalization.

[0038] The window segmentation module segments the preprocessed vibration data according to a preset sliding time window, obtaining vibration data segments corresponding to multiple time windows. The time window length is set to 0.2 seconds to 2 seconds, preferably 1 second. Adjacent time windows are set to overlap, with an overlap rate of 25% to 75%, preferably 50%. Each time window contains synchronous vibration data segments corresponding to all vibration measurement points and is associated with the operating condition information within that time window. The effective window filtering module performs a consistency check on the operating condition information of each time window. When the operating condition partition label changes within the same time window, the time window is marked as a transition window; when the operating condition partition label remains consistent within the same time window, the time window is determined as an effective time window. During the healthy baseline model training phase, only vibration data segments corresponding to effective time windows are used; during the real-time status monitoring phase, transition windows are individually marked, and their participation in anomaly assessment is determined according to preset rules.

[0039] A time-frequency transformation is performed on the vibration data segment within each effective time window to generate a corresponding time-frequency map. The time-frequency transformation uses a short-time Fourier transform. For each vibration measuring point and each time window, the input of the short-time Fourier transform is the vibration data segment of that measuring point within that time window, and the output is a time-frequency distribution matrix. The time-frequency distribution matrix is ​​converted into a time-frequency map, where each pixel position corresponds to the energy distribution at different time and frequency positions. Thus, within the same time window, each vibration measuring point corresponds to a set of original vibration sequences and a time-frequency map.

[0040] The window function, window length, and step size of the short-time Fourier transform are set according to the sampling frequency and the target analysis frequency band. The output of the short-time Fourier transform is subjected to amplitude normalization before entering the subsequent model, so that time-frequency plot data at a uniform scale are formed for different time windows and different vibration measurement points. For 11 vibration measurement points, 11 sets of original vibration sequences and 11 time-frequency plots are formed within each effective time window, which serve as input data for the subsequent health baseline model.

[0041] The working condition zoning unit establishes working condition zoning labels based on working condition information. Each working condition zoning label is obtained by combining discrete intervals of multiple working condition variables. These multiple working condition variables include at least two of the following: luffing angle, angular velocity, load weight, lifting / lowering direction, and start / stop status. The working condition zoning label is jointly determined by the luffing angle interval, angular velocity interval, load weight interval, lifting / lowering direction, and start / stop stage. The luffing angle is divided into multiple luffing angle intervals according to a preset angle range; the angular velocity is divided into multiple angular velocity intervals according to a preset velocity range; the load weight is divided into no-load, light-load, medium-load, and heavy-load intervals according to a preset load range; the lifting / lowering direction is divided into lifting working condition and lowering working condition; and the start / stop status is divided into starting stage, steady-state stage, braking stage, and pause stage.

[0042] The working condition zoning unit performs discrete interval matching on the working condition information within each effective time window, and combines the amplitude angle interval, angular velocity interval, load weight interval, start / stop direction, and start / stop phase in a fixed order to generate corresponding working condition zoning labels. The working condition zoning labels are bound and stored with vibration data segments and time-frequency diagrams of all vibration measurement points within the same time window. Once the working condition zoning labels are determined, the same encoding rules are maintained during sample selection, model training, and anomaly assessment to ensure that vibration data segments corresponding to the same working condition zoning label have consistent working condition constraints.

[0043] Furthermore, the health baseline model established by the sample screening and modeling unit includes: a one-dimensional autoencoder based on the original vibration sequence, and a two-dimensional autoencoder based on the time-frequency graph.

[0044] Specifically, the sample screening and modeling unit statistically analyzes the historical samples corresponding to each operating condition partition label. The statistics include the number of samples, cumulative duration, and frequency of occurrence. Based on the statistical results, the most common operating condition partition is determined. The most common operating condition partition is one that simultaneously meets the preset thresholds for the number of samples, cumulative duration, and frequency of occurrence. A health baseline model is established for each of the most common operating condition partitions; no separate health baseline model is established for operating condition partitions that do not meet the thresholds. For operating condition partitions without a separate health baseline model, the health baseline model corresponding to the neighboring operating condition partition label or a fallback health baseline model is used for calculation. When the sample size for a given operating condition partition is insufficient and the anomaly detection criteria cannot be met, only data is recorded. The neighboring operating condition partition label refers to the operating condition partition label that is adjacent to or has the smallest difference from the current operating condition partition label in at least one operating condition variable dimension. The fallback health baseline model is a general baseline model trained based on all healthy samples or cross-operating condition samples, and is invoked when the sample size for the corresponding operating condition partition is insufficient.

[0045] The health sample verification module filters historical samples in a healthy state from the historical samples. Historical samples in a healthy state are those where the equipment has been manually confirmed to be in normal operating condition and has no abnormal alarms within the corresponding observation period. Historical samples with maintenance records, lubrication abnormality records, jamming records, overheating records, or abnormal alarm records are not included in the healthy sample set. After health sample verification, they are archived according to operating condition zoning labels and used as training samples for the corresponding health baseline model.

[0046] For each of the most common operating condition zones, a corresponding health baseline model is established. The health baseline model includes a one-dimensional autoencoder based on the original vibration sequence and a two-dimensional autoencoder based on the time-frequency plot. At the input end, the one-dimensional autoencoder receives the original vibration sequence corresponding to each vibration measurement point under the operating condition zone label, and the two-dimensional autoencoder receives the time-frequency plot corresponding to each vibration measurement point under the operating condition zone label. At the output end, the one-dimensional autoencoder outputs the reconstructed original vibration sequence, and the two-dimensional autoencoder outputs the reconstructed time-frequency plot. The model training objective is to ensure that historical samples under healthy conditions form a small reconstruction error after passing through the corresponding autoencoder, thereby establishing a reconstructed baseline for the healthy state under the operating condition zone label.

[0047] The sample selection and modeling unit also establishes a multi-point synchronization baseline. This baseline is generated based on the relationship matrix between 11 vibration measurement points within the same time window. The relationship matrix includes at least one of the following: correlation coefficient matrix, cross-spectral consistency matrix, and energy distribution matrix. The relationship matrix is ​​calculated from historical samples under healthy conditions and stored separately according to operating condition partition labels. The multi-point synchronization baseline is used to calculate the multi-point synchronization consistency deviation during the real-time anomaly assessment phase.

[0048] Furthermore, the anomaly assessment unit is used to perform weighted fusion of the reconstruction error based on the original vibration sequence, the reconstruction error based on the time-frequency diagram, the deviation of potential features, and the deviation of synchronization consistency of multiple measurement points to obtain an anomaly score; the health index is a value of 1-100, which is obtained by normalizing the anomaly score and using a monotonically decreasing mapping function.

[0049] Specifically, during real-time operation, the anomaly assessment unit performs model matching and anomaly assessment for the current time window. First, it generates a current operating condition partition label based on the operating condition information corresponding to the current time window. Second, it calls the corresponding health baseline model in the model management module based on the current operating condition partition label. Third, it inputs the original vibration sequences of each vibration measuring point within the current time window into a one-dimensional autoencoder, inputs the time-frequency diagrams of each vibration measuring point within the current time window into a two-dimensional autoencoder, and calculates the anomaly score.

[0050] The calculation of the anomaly score includes at least the original vibration reconstruction error, the time-frequency plot reconstruction error, the latent characteristic deviation, and the multi-point synchronization consistency deviation. The original vibration reconstruction error is denoted as... This represents the difference between the original vibration sequence within the current time window and the reconstructed original vibration sequence output by the one-dimensional autoencoder; the time-frequency plot reconstruction error is denoted as... This represents the difference between the time-frequency map within the current time window and the reconstructed time-frequency map output by the two-dimensional autoencoder; the latent feature deviation is denoted as... This represents the degree of deviation between the position of a sample within the current time window in the latent feature space of the healthy baseline model and the baseline position of a historical sample in the same latent feature space under healthy conditions; the multi-point synchronization consistency deviation is denoted as... This indicates the degree of deviation between the relationship matrix of the 11 vibration measurement points within the current time window and the baseline of the multi-measurement point synchronization relationship under the corresponding working condition partition label.

[0051] In this embodiment, the total anomaly score A(t) is calculated according to the following formula: ; Where w1, w2, w3, and w4 are the weighting coefficients for the original vibration reconstruction error, the time-frequency diagram reconstruction error, the potential feature deviation, and the multi-measurement point synchronization consistency deviation, respectively. w1, w2, w3, and w4 are determined before model deployment and remain unchanged during online monitoring corresponding to the same healthy baseline model. When the current time window lacks multi-measurement point synchronization relationship input, Processed according to the preset default rules, and It is still generated according to a unified output format. The default rule is to use preset weights or historical statistical values ​​to replace the calculation when some input data is missing.

[0052] To ensure the comparability of anomaly scores among different operating condition partition labels, the health index generation unit first calculates the total anomaly score. Normalization is performed to obtain normalized outlier scores. The normalization process uses the distribution of abnormal scores of historical health status samples under the corresponding working condition partition label as a benchmark, and calculates the total abnormal score of the current time window. Map to a uniform scale. After normalization, normalize the outlier scores. Mapped to health index Health Index The value ranges from 1 to 100. The higher the value, the closer the hinge, hinge support components and their associated force transmission structures are to a healthy state for the current time window.

[0053] In this embodiment, the health index HI(t) is calculated according to the following formula:

[0054] Where k is the proportionality coefficient. This indicates that the calculation results will be limited to the range of 1 to 100. The proportional coefficient k is determined based on historical sample calibration results before system deployment and remains consistent under the same set of early warning rules. Through the above mapping relationship, when the anomaly score increases, the health index decreases; when the anomaly score decreases, the health index increases. The health index generation unit generates the health index for a continuous time window. Arranged in chronological order, forming a trend curve and stored in the storage and display unit.

[0055] Early warning decision-making unit based on health index Trend assessment and tiered early warning are performed. Trend assessment includes determining the real-time health index value, the moving average health index value, the rate of decline of the health index, the duration of the decline, and cross-measurement point consistency. The real-time health index value is used to identify the state level within the current time window; the moving average health index value is used to eliminate the impact of abnormal fluctuations within a single time window; the rate of decline of the health index is used to identify the trend of the health index over continuous time windows; the duration of the decline is used to determine whether a certain state level has been maintained continuously for a preset duration; and the cross-measurement point consistency is used to identify whether multiple vibration measurement points have experienced synchronous deterioration within the same time period.

[0056] The early warning decision unit outputs early warning levels according to the health index range. A health index of 85 to 100 outputs a Level 0 warning; 70 to 85 outputs a Level 1 warning; 40 to 70 outputs a Level 2 warning; and 1 to 40 outputs a Level 3 warning. A Level 0 warning corresponds to normal operation and routine inspections; a Level 1 warning corresponds to continuing operation, shortening the observation period, and monitoring trend changes; a Level 2 warning corresponds to scheduling a shutdown for inspection; and a Level 3 warning corresponds to immediate shutdown or triggering interlock protection. Each warning level and its corresponding recommended action are output to the storage and display unit.

[0057] To reduce false alarms, the early warning decision unit jointly determines the health index range based on its duration, continuous window count, rate of decline, and consistency across measurement points. If the health index briefly falls into a lower range and then recovers to a higher range, only the event is recorded, and no high-level alarm is generated. If the health index remains in a lower range and the continuous window count increases, the early warning level is maintained or increased. If the rate of decline of the health index exceeds a preset threshold and multiple key vibration measurement points simultaneously exhibit synchronous anomalies, the early warning level is output according to a higher-level rule. If multiple key vibration measurement points simultaneously deteriorate within a continuous time window and the health index enters the level 3 early warning range, the early warning decision unit outputs an interlocking protection signal.

[0058] The results output by the early warning decision unit include the current operating condition zone label and the anomaly score for the current time window. Health index for the current time window The data includes trend curves, abnormal measurement points, early warning levels, recommended actions, and interlock protection status. All these results are stored uniformly in the storage and display unit and linked to a timestamp. The stored data is used for subsequent inspection and verification, model updates, and historical data tracing.

[0059] Furthermore, the present invention includes a method for monitoring the status of the luffing structure of a port gantry crane, comprising the following steps: S1: Simultaneously collect vibration data from multiple vibration measurement points and collect the corresponding operating condition information. S2: Synchronize and preprocess the vibration data and working condition information to obtain vibration data segments corresponding to multiple time windows. Perform time-frequency transformation on the vibration data segments corresponding to each time window to generate a time-frequency diagram. S3: Divide the vibration data segment into working condition partitions based on the working condition information to obtain multiple working condition partition labels; S4: For each working condition zone label, establish a health baseline model based on historical samples under healthy conditions, and establish a multi-measurement point synchronization relationship baseline; S5: Input the real-time acquired vibration data segments and their time-frequency graphs into the health baseline model of the corresponding working condition zone label, and calculate the anomaly score by combining the multi-measuring point synchronization relationship baseline. S6: Map abnormal scores to health indices; S7: Conduct trend assessment and graded early warning based on health index.

[0060] Furthermore, the present invention includes an electronic device comprising a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the above-described method for monitoring the status of the luffing structure of a port gantry crane.

[0061] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A port gantry crane luffing structure status monitoring system, characterized in that, include: The sensing and acquisition unit is used to synchronously acquire vibration data from multiple vibration measuring points arranged along the hinge shaft support and its force transmission path, and to synchronously acquire the working condition information corresponding to the vibration data. The data synchronization and preprocessing unit is used to time-align the vibration data and working condition information from multiple measurement points, and to perform DC removal, abnormal pulse removal, bandpass filtering and normalization on the vibration data. It also divides the data according to a preset sliding time window to obtain vibration data segments corresponding to multiple time windows. The time-frequency transformation unit is used to perform time-frequency transformation on the vibration data segment to obtain the corresponding time-frequency diagram; The working condition partitioning unit is used to discretize and partition the vibration data segment corresponding to the time window based on at least two working condition variables among the luffing angle, angular velocity, load weight, luffing / lowering direction and start / stop status, and generate working condition partitioning labels. The sample screening and modeling unit is used to establish a health baseline model based on historical samples under a healthy state for each working condition zone label, and to establish a corresponding multi-measurement point synchronization relationship baseline. The health baseline model includes a one-dimensional autoencoder based on the original vibration sequence and a two-dimensional autoencoder based on the time-frequency graph; An anomaly assessment unit is used to input real-time vibration data segments into a one-dimensional autoencoder corresponding to the working condition partition label to obtain the original vibration reconstruction error; input the time-frequency graph corresponding to the real-time vibration data segments into a two-dimensional autoencoder corresponding to the working condition partition label to obtain the time-frequency graph reconstruction error; and perform weighted fusion based on the original vibration reconstruction error, the time-frequency graph reconstruction error, the potential feature deviation, and the multi-measurement point synchronization consistency deviation to obtain an anomaly score. A health index generation unit is used to normalize the abnormal scores and map them into a health index; The early warning decision unit is used to perform trend assessment and graded early warning based on the health index.

2. The port gantry crane luffing structure status monitoring system according to claim 1, characterized in that, The multiple vibration measuring points are arranged at key rigid parts of the hinge support and its force transmission path.

3. The port gantry crane luffing structure status monitoring system according to claim 1, characterized in that, In the sensing and acquisition unit, the sampling frequency of the synchronous vibration acquisition is not less than 25000Hz, and multiple vibration measurement points are synchronously acquired by using a unified clock or hardware synchronous triggering.

4. The port gantry crane luffing structure status monitoring system according to claim 1, characterized in that, The operating condition information includes at least two of the following: luffing angle, angular velocity, luffing / lowering direction, load weight, luffing drive current, braking status, and start / stop status.

5. The port gantry crane luffing structure status monitoring system according to claim 1, characterized in that, The data synchronization and preprocessing unit is used to perform DC removal, abnormal pulse removal, bandpass filtering, and normalization on the vibration data, and to segment the vibration data according to a preset sliding time window to obtain multiple vibration data segments.

6. The port gantry crane luffing structure status monitoring system according to claim 1, characterized in that, The sample screening and modeling unit is used to statistically analyze the number of samples, cumulative duration and frequency of occurrence corresponding to each working condition partition label, screen working condition partitions that meet the preset sample number threshold, preset cumulative duration threshold and preset frequency of occurrence threshold, and establish the health baseline model for each screened working condition partition.

7. The port gantry crane luffing structure status monitoring system according to claim 1, characterized in that, The multi-point synchronization consistency deviation is obtained based on the degree of deviation between the relationship matrix of multiple vibration measurement points within the current time window and the multi-point synchronization relationship baseline under the corresponding working condition partition label. The relationship matrix includes at least one of the correlation coefficient matrix, cross-spectral consistency matrix and energy distribution matrix.

8. The port gantry crane luffing structure status monitoring system according to claim 1, characterized in that, The health index is a value from 1 to 100, and it is obtained by normalizing the abnormal score and using a monotonically decreasing mapping function.

9. A method for monitoring the status of the luffing structure of a port gantry crane, characterized in that, The port gantry crane luffing structure status monitoring system according to any one of claims 1-8 includes the following steps: S1: Simultaneously collect vibration data from multiple vibration measurement points and collect the corresponding operating condition information. S2: Synchronize and preprocess the vibration data and working condition information to obtain vibration data segments corresponding to multiple time windows. Perform time-frequency transformation on the vibration data segments corresponding to each time window to generate a time-frequency diagram. S3: Based on at least two working condition variables among the amplitude angle, angular velocity, load weight, start / stop direction and start / stop status, the vibration data segment is discretized and partitioned to obtain multiple working condition partition labels; S4: For each working condition zone label, establish a health baseline model based on historical samples under healthy conditions, and establish a multi-measurement point synchronization relationship baseline; the health baseline model includes a one-dimensional autoencoder based on the original vibration sequence and a two-dimensional autoencoder based on the time-frequency diagram; S5: Input the real-time acquired vibration data segments into a one-dimensional autoencoder corresponding to the working condition zone label to obtain the original vibration reconstruction error; input the time-frequency graph corresponding to the real-time acquired vibration data segments into a two-dimensional autoencoder corresponding to the working condition zone label to obtain the time-frequency graph reconstruction error; and perform weighted fusion based on the original vibration reconstruction error, the time-frequency graph reconstruction error, the potential feature deviation, and the multi-measurement point synchronization consistency deviation to obtain the anomaly score; S6: Map the abnormal scores to a health index; S7: Conduct trend assessment and graded early warning based on the aforementioned health index.

10. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the port gantry crane luffing structure status monitoring method of claim 9.

Images