Port machinery steady state extraction hierarchical storage method and system based on plc event anchor point

Abstract

The application provides a port machine steady state extraction hierarchical storage method and system based on a PLC event anchor point, and relates to the technical field of port crane state monitoring and predictive maintenance. The method collects PLC input and monitoring signals such as vibration and temperature, establishes a unified time base and extracts event anchor points, completes time deviation correction and input / output mapping; constructs a port machine steady state index based on a work stage state machine and a sliding time window, and screens steady state candidate windows; through sub-window feature calculation and non-stationary quantitative index fine screening, a steady state window is obtained, and an event trigger window and a comparison and contrast window are constructed; hierarchical storage is performed according to the retention value score and the dynamic threshold of storage occupancy rate, a working condition index is generated, and similar working condition retrieval and calling are performed based on working condition similarity. The application can improve the purity and alignment accuracy of steady state data, optimize storage efficiency, and provide reliable data support for port machine fault diagnosis, state evaluation, trend analysis and life evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of port crane condition monitoring and predictive maintenance, and particularly relates to a method and system for extracting and hierarchically storing the steady state of port cranes based on PLC event anchors. Background Art

[0002] Port cranes include quay container cranes, portal cranes, rail-mounted gantry cranes, rubber-tired gantry cranes, etc. During the actual operation of such equipment, they usually experience frequent starts and stops, acceleration and deceleration, brake switching, and multi-mechanism linkage operation, and the monitored signals such as vibration and temperature output show obvious non-steady state characteristics. Taking the ship unloading operation of a quay container crane as an example, within a complete operation cycle, the hoisting mechanism and the trolley running mechanism often go through multiple stages such as heavy-load hoisting, uniform transportation, lowering and alignment, and empty spreader return. Even when operated by a skilled driver, the linkage actions of hoisting and the trolley often occur. The traditional analysis method of continuous slicing at equal time intervals is likely to mix the mixed signals of stop, acceleration, deceleration, uniform speed or linkage states in each analysis segment, resulting in a lack of comparability between different segments.

[0003] In the prior art, although there are monitoring schemes for collecting PLC signals or vibration signals, most of the schemes only use PLC data for control or alarm, or only perform continuous storage and offline analysis on the output signals, and fail to form a collaborative technical scheme of "unified time base - event anchor alignment - operation stage identification - steady state candidate window positioning - output fine screening - hierarchical storage of three types of segments - input-output mapping index" around the characteristics of strong non-steady state, multi-mechanism linkage and large amount of high-sampling data of port cranes. This brings at least the following problems: First, it is difficult to stably extract a steady state waveform suitable for spectrum, envelope spectrum, cepstrum or vibration intensity calculation from a large amount of non-steady state data; Second, there is no one-to-one correspondence between the output signal and the input working conditions such as load, speed, position, braking, and linkage, making it difficult to support backtracking and comparison under the same working conditions; Third, there is no high-sampling rate original waveform bound to the input state at that time for sudden shocks or abnormal events; Fourth, the continuous storage of the full amount of original waveforms will bring a large storage and calculation burden.

[0004] In summary, the prior art cannot adapt to the operation scenarios of port cranes with strong non-steady state and multi-mechanism linkage, and it is difficult to provide an integrated data support of high-purity steady state data, accurate input-output mapping, efficient hierarchical storage, and intelligent working condition retrieval, which restricts the development of predictive maintenance applications such as condition evaluation, fault diagnosis, and life assessment. Summary of the Invention

[0005] To address the challenges of extracting high-purity steady-state segments from the operational data of port cranes with strong non-steady states and multi-mechanism linkages, the difficulty in establishing a one-to-one correspondence between input conditions and output responses, and the difficulty in balancing abnormal event backtracking with long-term low-cost storage in existing technologies, this invention proposes a hierarchical storage method and system for steady-state extraction of port cranes based on PLC event anchor points. This solution does not merely synchronously acquire PLC signals along with output signals such as vibration and temperature. Instead, it uses PLC input signals as the primary driver, constructing a complete technical chain encompassing "unified time base and event anchor point alignment—operation stage identification—steady-state candidate identification—output fine screening—construction of three types of segments—hierarchical storage—operation condition similarity retrieval—input-output mapping and retrieval," thereby forming a data foundation that possesses clear operational condition attribution while also considering storage efficiency and subsequent retrieval capabilities.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for extracting and hierarchically storing steady-state data of port machinery based on PLC event anchors, the method comprising the following steps: The system collects the equipment input signals and equipment output signals of the port crane and generates timestamps for each; wherein the equipment input signals include at least two of the following: load information, target mechanism speed information, target mechanism position information, running status word, braking status information, and linkage flag information; and the equipment output signals include at least one of the following: vibration acceleration signal, vibration velocity signal, and temperature signal. The device input signal and the device output signal are processed using a unified time base. Based on the time synchronization, input-side event anchor point extraction, output-side response feature moment extraction, time deviation robust correction, and segmented drift compensation results, a one-to-one correlation relationship between the input signal and the output signal is established. A work phase state machine is constructed based on the device input signals to identify the phase boundaries of the target mechanism in a work cycle and determine the range of the target phase to be captured. Within the target stage range, steady-state candidate constraints are applied based on the equipment input signals to construct a port machinery steady-state index and identify steady-state candidate windows; Within the steady-state candidate window, the output signal of the device is finely screened, and the steady-state candidate window is subdivided into multiple continuous sub-windows. A non-stationary quantification index is constructed based on at least one time-domain feature, frequency-domain feature, envelope feature and / or statistical feature. Impact, transient and strongly non-stationary segments are eliminated to obtain the steady-state window for subsequent analysis. Based on the steady-state window, event trigger window, and comparison window, three types of data segments are constructed, and a retention value score is calculated for each data segment. The hierarchical storage threshold is dynamically adjusted according to the storage occupancy rate so that the three types of data segments are saved as high-sampling-rate original waveforms, downsampled data, and / or feature substitution data, respectively. Generate operating condition indexes for each data segment and establish input-output mapping records. Perform data retrieval and calling based on the similarity of operating conditions to support condition evaluation, fault diagnosis, trend analysis and / or life assessment.

[0007] On the other hand, the present invention also provides a port machinery steady-state extraction and hierarchical storage system based on PLC event anchor points, the system being used to execute the above-described method, the system comprising: The signal acquisition and time base alignment module is used to acquire the equipment input signals and equipment output signals of the port crane and generate timestamps for each; it performs unified time base processing on the equipment input signals and equipment output signals, and establishes a one-to-one correlation between the input signals and output signals based on time synchronization, input-side event anchor point extraction, output-side response feature moment extraction, time deviation robust correction, and segmented drift compensation results; wherein, the equipment input signals include at least two of the following: load information, target mechanism speed information, target mechanism position information, running status word, braking status information, and linkage flag information, and the equipment output signals include at least one of the following: vibration acceleration signal, vibration velocity signal, and temperature signal; The work phase identification module is used to construct a work phase state machine based on the equipment input signals, identify the phase boundaries of the target mechanism in a work cycle, and determine the range of the target phase to be captured. A steady-state candidate identification module is used to apply steady-state candidate constraints based on the equipment input signal within the target stage range, construct a port machinery steady-state index and identify steady-state candidate windows; The output screening module is used to perform output screening on the output signal of the device within the steady-state candidate window, subdivide the steady-state candidate window into multiple continuous sub-windows, construct non-stationary quantification index based on at least one time-domain feature, frequency-domain feature, envelope feature and / or statistical feature, eliminate shock, transient and strongly non-stationary segments, and obtain a steady-state window for subsequent analysis. The data fragment construction and hierarchical storage module is used to construct three types of data fragments based on the steady-state window, event trigger window and comparison window, calculate the retention value score for each data fragment, and dynamically adjust the hierarchical storage threshold according to the storage occupancy rate, so as to save the three types of data fragments as high sampling rate original waveforms, downsampled data and / or feature substitution data respectively. The operating condition indexing and retrieval module is used to generate operating condition indexes for each data segment and establish input-output mapping records. It also performs data retrieval and retrieval based on the similarity of operating conditions to support condition evaluation, fault diagnosis, trend analysis and / or life assessment.

[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 hierarchical storage of port machinery steady-state extraction based on PLC event anchor points.

[0009] Compared with the prior art, the beneficial effects of the present invention are: First, by establishing a one-to-one correspondence between input and output through event anchor point alignment and a unified time base, the errors caused by clock deviations at different acquisition ends, event detection lag, and interference from multi-mechanism linkages in data mapping are reduced, improving the accuracy of the correspondence between input-side operating condition information and output-side monitoring response. Second, by prioritizing the determination of candidate windows from the input side through the operation phase state machine and steady-state candidate constraints, and then finely screening and eliminating non-stationary segments from the output side, the purity and comparability of steady-state data segments can be significantly improved. Third, by collaboratively constructing three types of segments—steady-state windows, event trigger windows, and comparison windows—the needs of steady-state analysis can be met, while retaining the high-sampling-rate original waveforms required for abnormal event backtracking, and forming a conditional input-output sample library. Fourth, through a hierarchical storage mechanism for original waveforms, downsampled data, and feature substitution data, combined with retention value scoring and dynamic thresholds, the storage and computational burden caused by continuous full high-sampling-rate storage can be reduced while ensuring the traceability of key segments. Fifth, through operating condition indexes, operating condition similarity, and input-output mapping index tables, a standardized, searchable, and traceable data foundation is provided for subsequent status evaluation, fault diagnosis, trend analysis, and life assessment.

[0010] 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

[0011] Figure 1 This is a schematic diagram of the overall system structure and the relationship between the functional modules provided by the present invention; Figure 2 This is a schematic diagram of the process for unifying the time base of input and output signals, extracting event anchor points, correcting time deviations, and generating associated records provided by the present invention. Figure 3 This is a schematic diagram of the process for identifying the job stage, calculating the sliding window constraint, and outputting the steady-state candidate window provided by the present invention. Figure 4 This is a schematic diagram illustrating the relationship between the output fine screening within the candidate window, the construction of three types of data segments, and the hierarchical storage provided by the present invention; Figure 5 This is a schematic diagram illustrating the relationship between the generation of the working condition index, the construction of the input-output mapping index table, and the retrieval and calling functions provided by this invention. Detailed Implementation

[0012] 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.

[0013] 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”.

[0014] 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.

[0015] 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.

[0016] 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.

[0017] The following is combined Figures 1 to 5 The implementation details of the embodiments of the present invention will be described in detail below. The following content is only for the convenience of understanding the implementation details and is not necessary for implementing this solution.

[0018] Example 1 like Figure 1As shown, the system of the present invention includes a PLC input acquisition and standardization module 101, an output acquisition module 102, a unified time base and event anchor point alignment module 103, a work phase state machine module 104, a steady-state candidate identification module 105, an output fine screening module 106, a three-segment construction module 107, a hierarchical storage and indexing module 108, and an input-output mapping management module 109. Module 101 is used to acquire and standardize equipment input signals such as load, speed, position, running status word, braking status, and linkage flag; module 102 is used to acquire equipment output signals such as vibration and temperature; module 103 is used to establish a unified time base and complete event anchor point alignment; modules 104 and 105 are used to identify the work phase from the input side and determine the steady-state candidate window; module 106 is used to fine screen the output signals in the candidate window; module 107 is used to construct the steady-state window, event trigger window, and comparison window; module 108 is used to hierarchically store the three types of segments and generate a working condition index; module 109 is used to establish input-output mapping records based on the unified time base, working condition index, and storage location.

[0019] like Figure 2 As shown, the PLC timing stream 201 and the sensor timing stream 202 first enter the unified timing / timestamp standardization module 203 to form a comparable unified time reference under the meaning of coarse synchronization. On this basis, the input-side event anchor point extraction module 204 identifies events such as brake release, brake engagement, target mechanism speed transition from zero to non-zero, target mechanism speed transition from non-zero to zero, linkage start, and linkage end from the PLC input variables, and determines the occurrence time of the corresponding event as the input-side event anchor point time. The timing of the input-side event anchor point can be determined by the timing of changes in the status bit of the PLC input variables, the timing when the speed variable crosses a preset threshold, the timing of changes in the running status word, and / or the timing of changes in the linkage flag.

[0020] The output-side response feature recognition module 205 uses the time of each input-side event anchor point. Centered on the preset search window Within, extract the output-side response feature moments related to the corresponding event from the device's output signal. Preferably, the output-side response characteristic moment is the moment when the short-time energy, short-time root mean square, envelope amplitude, bandwidth energy, or a combination thereof of the device output signal first meets the preset response significance condition; the response significance condition includes: exceeding a preset threshold relative to the mean of the pre-event baseline, exceeding a preset multiple relative to the standard deviation of the pre-event baseline, and / or maintaining a significant deviation from the baseline for a preset duration. For event pairs that do not meet the response significance condition, have multiple competing candidate response moments, or cannot form a unique pairing relationship, they can be determined as pairing failures and will not participate in subsequent time deviation estimation.

[0021] For the input-side event anchor moment of successful pairing With output-side response characteristic time The time deviation correction module 206 first calculates the original time deviation of the current event pair. : To mitigate the impact of misidentification of individual events, localized shocks, or transient noise on bias estimation, the current time bias is... Robust correction is performed on the time deviations of several neighboring events to obtain the corrected time deviation. : in, For event type e i and organization type m i The corresponding weighting coefficients are preferably in the range of 0 to 1; This represents the median of the time deviations of a preset number of neighboring events before and after the current event; r is the parameter for the number of neighboring events.

[0022] After obtaining the corrected time deviation, the time deviation correction module 206 applies the corrected time deviation to any two adjacent output side response characteristic moments. and Segmented drift compensation and additional compensation are applied to the output timestamps t between them to obtain the corrected timestamps under a unified time base. : Among them, the first item The first term represents the basic deviation corresponding to the current anchor point interval; the second term represents the segmented drift compensation term between adjacent anchor points. This represents the braking hysteresis factor caused by the switching of braking states. This indicates the linkage interference coefficient caused by the linkage of the mechanisms; and These represent the braking hysteresis compensation weight and the linkage interference compensation weight, respectively. Preferably, and The selection criteria can be determined based on historical calibration data, institution type, event type, or empirical parameters, with the principle of minimizing the overall time alignment error between the calibrated input-output event pairs.

[0023] Preferably, the braking hysteresis factor It can be obtained by normalizing the time difference between the braking command time and the time when the equipment output signal shows a significant response, according to the length of the anchor point interval; for example, it can be expressed as: in, The timing of the PLC-side braking state switching command. This refers to the significant response moment in the device's output signal corresponding to the switching of the braking state. To prevent extremely small positive numbers with a denominator of zero.

[0024] Preferably, the linkage interference coefficient It can be determined based on the relative motion intensity between the target mechanism and the associated mechanism within the anchor point interval, for example, as follows: in, Let the velocity of the target mechanism be at the k-th sampling time. The velocity of the associated mechanism at the k-th sampling time. Let i be the number of sampling points participating in the calculation within the i-th anchor point interval. To prevent extremely small positive numbers with a denominator of zero, the speeds of the target mechanism and the associated mechanism can be directly obtained from the PLC speed variables or calculated differentially from the corresponding position variables. Sampling points where the target mechanism is in an invalid low-speed range may be excluded from the calculation of the linkage interference coefficient.

[0025] The unified time base correlation engine 207 is based on the corrected timestamp The device output signal is mapped to a time axis under a unified time base, and a one-to-one association relationship is formed between input segments and output segments under this unified time base, generating input-output association record 208. Therefore, subsequent steps S4 to S8 are all based on the association results under the unified time base for stage identification, steady-state candidate window positioning, output fine screening, hierarchical storage, and operating condition index construction.

[0026] like Figure 3 As shown, after the input signal is acquired and standardized (301), the operation stage state machine is first constructed based on the running status word, control logic combination, and mechanism enable signal, and the target stage (302) is identified. The types of stages that the operation stage state machine can identify include, but are not limited to, those mentioned above. Figure 3 The typical stages listed are the hoisting acceleration stage, hoisting constant speed stage, hoisting deceleration stage, trolley acceleration stage, trolley constant speed stage, trolley deceleration stage, lowering and alignment stage, and stopping stage. Based on this, the corresponding target stage range 303 is determined. Within the target stage range, the steady-state candidate identification module 105 uses a sliding time window W to perform window-by-window calculations on the equipment input signals 304. Specifically, based on the target mechanism's speed, load, position or angle, braking state, and the motion state of related mechanisms, it calculates speed stability, load stability, and position or angle constraints, and judges the braking state and linkage suppression constraints 305 to form the port machinery steady-state index SSI(W).

[0027] Specifically, a sliding time window is set. The velocity sequence of the internal target mechanism is v, and the load sequence is q, where This represents the velocity of the target mechanism at the k-th sampling time. This represents the load value at the k-th sampling time. and These represent the mean values ​​of the velocity sequence and the load sequence, respectively. To prevent extremely small positive numbers with a denominator of zero, the velocity stability constraint can be expressed as: The load stability constraint can be expressed as: Therefore, the speed stability constraint is used to characterize the normalized amplitude of the target mechanism's speed fluctuation relative to the average operating level, and the load stability constraint is used to characterize the normalized amplitude of the load fluctuation relative to the average load level. Preferably, the speed change smoothness constraint... It can also be determined based on the average absolute value of the rate of change of the target mechanism's speed, to reflect whether the speed change within the window is gradual.

[0028] For operational scenarios with risks of multi-agency collaboration, the steady-state candidate identification module 105 further calculates the collaboration suppression constraint. A sliding time window is set. The velocity sequence of the internally related mechanism is Their absolute values ​​averaged as Then the linkage interference coefficient can be expressed as: in, For window The average speed of the target mechanism. The speeds of the target mechanism and associated mechanisms are preferably obtained directly from the PLC speed variables, or calculated by differential calculation of the corresponding position variables. When the target mechanism is a hoisting mechanism, the associated mechanism is preferably a trolley mechanism; when the target mechanism is a trolley traveling mechanism, the associated mechanism is preferably a hoisting mechanism; in other port crane scenarios, the associated mechanism may also include a luffing mechanism, a slewing mechanism, or other mechanisms that have a linkage interference relationship with the target mechanism. The larger the linkage interference coefficient, the stronger the motion of the associated mechanism relative to the motion of the target mechanism, and the less suitable the current sliding time window is as a steady-state candidate window.

[0029] In addition to speed, load, and linkage factors, this invention also considers the influence of spatial position on steady-state determination. Let... For sliding time windows The cumulative time the target mechanism is within the preset spatial sensitive zone, where |W| is the total duration of the sliding time window, can be expressed as: The spatially sensitive area includes one or more of the following: end-limiting neighborhood, placement alignment area, track joint influence area, zero-speed switching neighborhood, and other preset control sensitive areas. The larger the proportion of the spatially sensitive area, the more significantly the current sliding time window is affected by position-sensitive factors, and the less suitable it is as a steady-state candidate window.

[0030] For braking state constraints, the steady-state candidate identification module 105 determines the braking state consistency factor B(W) based on whether the braking state remains consistent within the sliding time window W and whether it meets the expectations of the current target stage. Preferably, when the braking state remains stable throughout the sliding time window and matches the target stage, B(W) = 1; when the braking state switches within the sliding time window or does not match the target stage, B(W) = 0. In other embodiments, the braking state consistency factor can also take an intermediate value between 0 and 1 to reflect the degree of consistency of partial satisfaction.

[0031] Based on the above, the steady-state candidate identification module 105 integrates the speed stability constraint. Load stability constraint Constraint on the smoothness of speed change Braking state consistency factor Interference coefficient and the proportion of space-sensitive areas Constructing the port machinery steady-state index: Where w1 to w6 are the contribution coefficients corresponding to each constraint term, preferably satisfying The contribution coefficient can be preset based on the target mechanism type, operation stage type, historical calibration samples, and / or empirical rules. For example, for a hoisting mechanism, the weights of load stability and braking consistency can be appropriately increased; for a trolley running mechanism, the weights of speed stability and space-sensitive area avoidance can be appropriately increased.

[0032] Port Machinery Steady State Index Used for comprehensive characterization of sliding time windows The overall steady-state level, the larger the value, the more suitable the sliding time window is as a steady-state candidate window. Preferably, the steady-state index of the entire window satisfies only the following: Furthermore, the lower limit of the steady-state index of the preset sub-window within the sliding time window satisfies: The sliding time window is determined as a steady-state candidate window 307; wherein... The threshold for determining the entire window. The threshold for determining the lower limit of the sub-window is preferably satisfied. The threshold can be determined based on historical calibration samples, target organization type, operation phase type, and / or empirical rules.

[0033] The system proceeds to the judgment stage based on the calculation results to determine whether the preset constraint 306 is satisfied. If the determination is satisfied, the steady-state candidate window 307 is output; if the determination is not satisfied, the system moves through the sliding window and returns to the constraint calculation step 308 to continue searching and judging the next sliding time window. Figure 3 The typical stage examples listed in the upper right corner are used to illustrate the stage types that the job stage state machine can recognize. They are supplementary explanations of the stage types and do not change the execution order of the main process.

[0034] like Figure 4 As shown, after the steady-state candidate window 401 enters the output fine screening module 402, it is not directly retained as a whole as a steady-state window, but is first subdivided into multiple continuous sub-windows. The output screening module 402 calculates the time domain, frequency domain, and envelope statistical correlation features for each continuous sub-window to quantitatively determine whether impact, transient, and strongly non-stationary segments are mixed in within the candidate window. If necessary, it performs candidate sub-window reorganization and finally outputs the steady-state window 403.

[0035] Specifically, a steady-state candidate window is set. The window is divided into K consecutive sub-windows, and the output signal of the k-th sub-window is... Its root mean square is denoted as The centroid of the spectrum is denoted as Envelope entropy is denoted as The impact factor is denoted as Wherein, the root mean square The spectral centroid is used to characterize the overall energy level of the k-th sub-window. The envelope entropy is used to characterize the spectral energy distribution location of the k-th sub-window. The impact factor is used to characterize the complexity of the statistical distribution of the envelope of the k-th sub-window. Used to characterize whether there is a significant spike impact in the k-th sub-window.

[0036] To further characterize the spectral spread of the k-th sub-window, the output fine screening module 402 calculates the spectral broadening ratio of its spectral energy distribution: in, This represents the minimum effective bandwidth required to cover 50% of the total spectral energy in the k-th sub-window. This represents the minimum effective bandwidth required to cover 95% of the total spectral energy in the k-th sub-window. To prevent extremely small positive numbers with a denominator of zero, a larger spectral broadening ratio indicates a more significant diffusion of spectral energy from the core band to a wider band in the k-th sub-window, and a greater likelihood of including impulses, transients, linked disturbances, or other non-stationary components.

[0037] After obtaining the features of each sub-window mentioned above, the output screening module 402 constructs candidate windows. Non-stationary quantitative indicators : in, The coefficient of variation represents the root mean square of each sub-window, used to characterize the consistency of temporal energy within the candidate window; The coefficient of variation represents the spectral centroid of each sub-window, used to characterize the consistency of the frequency domain structure within the candidate window; Used to characterize the degree of change in the statistical structure of the envelope of adjacent sub-windows; Used to characterize the maximum impact level in each sub-window; This is the average of the spectral broadening ratios of each sub-window, used to characterize the overall spectral diffusion of the candidate window; , , , , The contribution coefficients for each item. The non-stationary quantitative indicators... The larger the value, the stronger the non-stationarity of the candidate window; the smaller the value, the closer the candidate window is to the steady-state window used for subsequent analysis.

[0038] Preferably, when the non-stationary quantization index of the candidate window satisfies: This indicates that the candidate window as a whole has the potential to remain a steady-state window; however, to prevent the candidate window from only partially satisfying the conditions for a short time while the overall situation is not continuous and stable, the output fine screening module 402 further requires that at least M consecutive sub-windows simultaneously satisfy the following: in, For non-stationary quantization threshold, The threshold value for the impact factor. M is the threshold for the spectral broadening ratio, and M is the threshold for the number of consecutive sub-windows. The threshold M for the number of consecutive sub-windows is preset based on the minimum continuous duration required for subsequent analysis, the output signal sampling rate, and the sub-window length. Only when at least... Only when all consecutive sub-windows simultaneously meet the conditions will the consecutive sub-windows be reorganized into a steady-state window 403; otherwise, the candidate window can be completely removed, or the sub-windows that do not meet the conditions can be removed and only the consecutive parts that meet the conditions can be retained.

[0039] Through the above processing, the output screening module 402 realizes the whole-window-level non-stationary quantification judgment, sub-window-level impact and spectrum broadening constraints, and continuous sub-window recombination of steady-state candidate windows, thereby avoiding the mistaken retention of segments that appear stable only in a few short periods of time but actually still contain impact, transient, and strong non-stationary components as steady-state windows. The obtained steady-state window 403 can be further entered into the hierarchical storage mechanism 406; for events such as collisions, abnormal impacts, braking stops, gear skipping, or track irregularity impacts, event trigger windows 404 can be constructed; for conditional samples under different loads, different speeds, different positions, and / or different operation types, comparison windows 405 can be constructed, and they all enter the subsequent hierarchical storage and indexing processing flow.

[0040] like Figure 4 As shown, after the steady-state window 403, event triggering window 404, and comparison window 405 are constructed, the hierarchical storage mechanism 406 does not simply apply fixed preset storage rules to different categories of fragments. Instead, it calculates a retention value score V(G) for each data fragment G and automatically determines the storage level of the data fragment based on the retention value score. The retention value score can be expressed as: Preferably, the steady-state purity Q(G) is determined based on the steady-state candidate identification result, output screening result, and continuous effective steady-state duration corresponding to the data segment. That is, the higher the steady-state candidate confidence, the lower the non-stationarity, and the longer the continuous effective steady-state duration of the segment in the preceding steps, the higher its steady-state purity. The event significance E(G) is determined based on the alarm bit, fault word, velocity mutation degree, and / or output signal impact intensity. That is, the higher the alarm level, the more significant the velocity mutation, and the stronger the output impact contained in the segment, the higher its event significance. The operating condition scarcity R(G) is determined based on the number of historical samples under the operating condition index corresponding to the data segment, and is used to characterize whether the segment belongs to a relatively rare operating condition in the sample library. The diagnostic deviation D(G) is determined based on the degree of deviation of the feature vector of the data segment from the center of historical normal samples of the same operating condition, and is used to characterize whether the segment, although not triggering an explicit alarm, has shown an abnormal deviation trend relative to the center of normal operating conditions. The storage cost C(G) is determined based on the segment sampling rate, number of channels, duration, and data volume corresponding to the storage mode, and is used to characterize the storage resources required to store the segment.

[0041] Preferably, the scarcity of the working condition R(G) is expressed as: in, For fragments The number of historical samples in the corresponding operating condition index bucket. This is the threshold for the number of saturated samples. When there are few historical samples of the same operating condition, the scarcity of that operating condition is high; when there are relatively abundant historical samples of the same operating condition, the scarcity of that operating condition decreases. The storage cost... Represented as: in, Sampling rate, For the number of channels, For the duration of the segment, The number of bytes of data per sampling point. This is a normalized reference constant. Therefore, the higher the sampling rate, the more channels, and the longer the segment, the higher the storage cost.

[0042] After obtaining the retention value score V(G), the hierarchical storage mechanism 406 automatically determines the storage level of the segment based on the retention value score. Preferably, when the retention value score of a segment is not lower than a first threshold, the high-sampling-rate original waveform and its corresponding input segment are stored; when the retention value score of a segment is between the first threshold and the second threshold, the downsampled waveform, key sub-windows, and / or feature substitution data are stored; when the retention value score of a segment is lower than the second threshold, only feature substitution data and index records are stored. Thus, segments with high steady-state purity, significant events, scarce operating conditions, or obvious diagnostic deviations are given priority for higher-level original data retention, while segments with abundant samples and lower value are stored in a lower-cost manner.

[0043] Preferably, the first threshold and the second threshold are dynamically adjusted based on the current storage occupancy rate to adapt to changes in edge-side or local storage resources. For example, the first threshold and the second threshold can be expressed as follows: in, To maintain a high-level baseline threshold, To adjust the base threshold for high-level preservation, To preserve the basic threshold for medium-level, To preserve the base threshold for the adjusted intermediate level, This represents the current storage utilization rate. and These are the dynamic adjustment coefficients for high-level and medium-level thresholds, respectively, and are preferred. When storage occupancy is high, the threshold is increased accordingly to prioritize high-level storage resources; when storage occupancy is low, the threshold can be appropriately lowered to retain more original data.

[0044] exist Figure 4In the hierarchical storage mechanism 406 shown, the original waveform layer preferably stores high-sampling-rate original waveforms of segments with high retention value; the feature layer preferably stores the time-domain features, frequency-domain features, and metadata of various segments; and the index layer preferably stores the operating condition index, equipment identifier, measurement point identifier, time range, job cycle number, and storage location mapping. Therefore, the hierarchical storage mechanism 406 is not a simple fixed-class storage, but rather an automatic decision-making allocation among the original waveform layer, feature layer, and index layer based on the segment retention value score.

[0045] like Figure 5 As shown, the working condition index generation module 501 combines the mechanism / stage / load / speed / position fields 502 and the braking / linkage / environment / event / measuring point fields 503, and writes them together with equipment identifier, measuring point identifier, work cycle number, and time range information into the input-output mapping index table 504. In this embodiment, the working condition index is not only used as a static label, but also serves as the calculation basis for retrieving the same or similar working conditions. Therefore, in addition to storing the original field values, the index table 504 preferably stores a set of discrete fields, a set of continuous parameters, event anchor point sequence features, and linkage mode features used for working condition similarity calculation.

[0046] For any two data segments and The system can be based on the consistency of discrete fields. Continuous parameter proximity Topological similarity of event anchor sequence and similarity of linkage modes Jointly calculate the similarity of working conditions : in, to For each contribution coefficient, the optimal one is selected based on the following conditions. Preferably, the discrete field consistency To characterize whether two segments are consistent in discrete fields such as equipment type, mechanism type, stage identifier, braking status, linkage status, and event type, it can be represented as: Where D is the set of discrete fields involved in the comparison, and 1(·) is an indicator function. It takes the value 1 when the corresponding fields are consistent, and 0 otherwise. Therefore, the more consistent the two segments are in the discrete working condition fields, the higher the consistency of the discrete fields.

[0047] The proximity of continuous parameters Used to characterize the proximity of two segments in terms of load, velocity, position or angle, duration, and / or environmental parameters. Preferably, it can be expressed as: in, , , , Each represents a segment The corresponding statistics include load, velocity, position or angle, and duration. , , , This is the normalized scaling parameter for the corresponding parameters. Therefore, the closer two segments are in terms of continuous operating parameters, the higher the similarity of the continuous parameters.

[0048] The topological similarity of the event anchor sequence This is used to characterize the similarity between two segments in terms of the order of their event anchor sequences, the composition of their event types, and the intervals between adjacent events. Preferably, it can be determined based on the length of the longest common subsequence of the event anchor sequences of the two segments, and can be expressed as: in, and Each represents a segment and fragments The corresponding event anchor sequence, This represents the length of the longest common subsequence between the two sequences. Therefore, the more consistent the order of the event anchor sequences and the more similar the event composition, the higher the topological similarity of the event anchor sequences.

[0049] The similarity of the linkage modes This is used to characterize the similarity between two segments in terms of the intensity, duration, and proportion of linkage between the target mechanism and related mechanisms. Preferably, it can be expressed as: in, and Each represents a segment and fragments The linkage mode characterization quantity can be determined based on one or more of the following: the relative magnitude of the speed of the target mechanism and the speed of the associated mechanism, the proportion of linkage duration, and the number of linkage events.

[0050] In obtaining working condition similarity Then, when the similarity of the working conditions is not lower than a preset similarity threshold At that time, determine the fragment With fragments This belongs to a similar operating condition segment. At this point, index table 504 can be further parsed to obtain the input segment or input variable snapshot position 505, and the output original waveform, downsampled data, or feature data position 506, and respond to the retrieval request 507 to retrieve the corresponding data segment 508. Thus, Figure 5 The process shown is no longer limited to precise retrieval based on "identical fields". Instead, it is more suitable for the highly unsteady, multi-stage, and multi-mechanism linkage operation scenarios of port cranes through the coherent execution of working condition index generation, working condition similarity calculation, location mapping and retrieval call. It can effectively support the comparison of the same working condition and the backtracking of near working condition data.

[0051] In another embodiment, the above method can be deployed in a collaborative architecture of edge computing gateway, industrial computer and cloud platform. The edge side is responsible for real-time acquisition, event anchor identification, steady-state candidate and output screening, while the cloud side is responsible for long-term archiving, index retrieval and cross-device unified management. The present invention can also be embodied as a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method steps of the present invention.

[0052] Example 2 This embodiment focuses on the hoisting mechanism of a quayside container crane. During a single unloading cycle, it collects input signals such as hoisting motor speed, hoisting height, spreader load, brake status, trolley speed, and linkage indicators via a PLC interface. Vibration acceleration signals are collected using vibration sensors installed on the hoisting gearbox, bearing housing, or motor housing. Combined with… Figure 2 The process shown first standardizes and analyzes the PLC input signals, using events such as brake release, lifting speed changing from zero to non-zero, and brake closure as event anchors. Then, the vibration acquisition terminal clock and the PLC-side clock are synchronized, and the time deviation between the two types of data is finely corrected using the event anchors. Subsequently, combined with... Figure 3 The process shown constructs a state machine for each work phase based on the PLC's operating status word and speed variable, identifying the lifting acceleration phase, lifting constant speed phase, lifting deceleration phase, and stopping phase.

[0053] When identifying the uniform lifting speed segment, the system reads the lifting speed sequence, spreader load sequence, lifting height sequence, brake status, and trolley speed sequence within a sliding time window, and calculates the speed stability constraint, load stability constraint, speed change smoothness constraint, braking state consistency factor, linkage interference coefficient, and spatial sensitive area proportion, respectively. For the lifting mechanism, the spatial sensitive area can preferably include the lifting zero-speed switching neighborhood, the container lowering / landing alignment area, and other preset lifting control sensitive areas; for linkage interference, the influence degree is preferably determined based on the ratio of trolley speed to lifting speed. Subsequently, the system constructs a port machinery steady-state index based on the constraints, and only retains sliding time windows where both the overall window steady-state index and the sub-window steady-state index simultaneously meet preset threshold conditions as steady-state candidate windows. If the linkage flag indicates linkage between the trolley and the lifting mechanism, and the linkage interference coefficient exceeds a preset range, the corresponding window is marked as a linkage interference window and its priority is reduced or excluded. This avoids misclassifying windows that, while in the constant-speed lifting phase, are simultaneously affected by trolley linkage, landing alignment correction, or zero-speed switching as steady-state candidate windows. If the conditions are not met, the corresponding window is marked as a linkage interference window and its priority is reduced or it is excluded. For the retained steady-state candidate windows, the system further divides them into multiple continuous sub-windows and calculates the root mean square, spectral centroid, envelope entropy, impact factor, and spectral broadening ratio for each sub-window. The impact factor is used to identify potential collisions, braking switching, or track impact spikes in the candidate window, and the spectral broadening ratio is used to identify potential spectral diffusion anomalies in the candidate window. Subsequently, the system constructs non-stationary quantification indices based on the root mean square coefficient of variation, spectral centroid coefficient of variation, change in envelope entropy of adjacent sub-windows, maximum impact factor, and average spectral broadening ratio for each sub-window. Only when the non-stationary quantification index does not exceed a preset threshold, and at least a preset number of continuous sub-windows simultaneously satisfy the impact factor and spectral broadening ratio constraints, are the corresponding continuous sub-windows reorganized into steady-state windows. Candidate windows that do not meet the conditions can be completely eliminated, or only consecutive sub-windows that meet the conditions can be retained as the final steady-state window. This method avoids mistakenly retaining collision impacts, braking transients, or linkage disturbances interspersed within the hoisting uniform speed segment as steady-state samples. The high-sampling-rate vibration original waveform of the steady-state window is saved together with the corresponding PLC input segment, and based on... Figure 5The system generates a condition index including equipment type, mechanism, stage, load range, speed range, height range, braking status, and linkage status, as shown. In subsequent retrieval calls, the system can not only perform a completely consistent condition search based on the above fields, but also further combine the calculation results of the current segment and historical segments in terms of the proximity of load, speed, and height ranges, the consistency of event anchor sequence, and the similarity of linkage modes to perform a search for similar condition segments. Therefore, in the hoisting mechanism scenario, it can more accurately find historical samples similar to the current steady-state window in terms of box weight, hoisting speed, trolley linkage status, and operation stage, for subsequent spectrum comparison, condition evaluation, and trend analysis.

[0054] Example 3 When track joint impact, collision, gear skipping, or brake interruption occurs during port machinery operation, alarm bits, braking status bits, or speed change information in the PLC input signals can trigger the construction of an event trigger window. The system saves the high-sampling-rate original output waveforms for a preset duration before and after the event, centered on the trigger moment, while simultaneously recording the load, speed, position, linkage flags, and relevant alarm words at that time. In the hierarchical storage stage, the system further calculates a retention value score based on the event salience, diagnostic deviation, and storage cost of the event segment. Since the event trigger window typically has high event salience, it is preferable to retain the high-sampling-rate original waveform and corresponding input segment, and write the event segment and corresponding input segment together into a mapping index table to ensure that abnormal events not only retain simplified features but also retain complete retrospective evidence bound to the input operating conditions.

[0055] Example 4 The system constructs a comparative comparison sample scenario. To establish a conditional input-output mapping sample library under different box weights, speeds, and trolley positions, the system actively filters data segments from multiple work cycles that meet the requirements of the same window length, the same signal type, and clear differences in operating conditions. For each data segment, in addition to generating operating condition fields such as mechanism, stage, load range, speed range, position range, braking state, and linkage state, the system further saves event anchor point sequence features and linkage mode features, and establishes an operating condition similarity retrieval mechanism in the sample library accordingly. For scenarios that require the construction of conditional comparison samples, the system can prioritize returning historical segments with operating condition similarity higher than a preset threshold, thereby improving the accuracy of comparisons with the same operating conditions while ensuring sample representativeness, and providing a more comparable near-operating condition sample basis for subsequent status evaluation, fault diagnosis, trend analysis, and life assessment.

[0056] Example 5 To verify the technical effectiveness of this invention in input / output alignment and steady-state window extraction, the hoisting mechanism and trolley traveling mechanism of a quayside container crane were selected as the verification objects, and actual operation data were continuously collected over multiple workdays. The collected PLC input signals included hoisting speed, trolley speed, spreader load, hoisting position, trolley position, brake status, running status word, linkage indicator, and alarm word; the collected equipment output signals included vibration acceleration signals located at the bearing housing of the hoisting gearbox and the motor housing. The PLC input sampling period was preferably 10ms to 20ms, and the output vibration signal sampling frequency was preferably 2kHz to 10kHz.

[0057] In one implementation, more than 1,000 complete operating cycles are selected from continuously collected data as verification samples. These include normal operating cycles, operating cycles with multi-mechanism linkage interference, and operating cycles containing abnormal events such as braking stop, track joint impact, and collision. At least two technicians with experience in port crane monitoring manually review and annotate a portion of the samples to obtain input-side event reference times, output-side response reference times, and manually confirmed effective steady-state intervals, which serve as verification benchmarks.

[0058] To avoid relying on existing traditional port machinery monitoring systems as a comparison object, this embodiment sets up the following comparative implementation method: Compared with Implementation Method 1, only the PLC input signal and the equipment output signal are uniformly synchronized and timestamp standardized to form a coarse synchronization time axis, but the pairing correction between the input side event anchor point and the output side response characteristic time is not performed, the robust correction of the time deviation of adjacent events is not performed, and the segmented drift compensation, braking hysteresis additional correction and linkage interference additional correction are not performed. Compared with Implementation Method 2, continuous slicing is performed only on the output signal with a fixed window length, and candidate windows are selected based on a single stationarity condition, without constructing steady-state candidate constraints based on speed, load, position, braking state and linkage state, and without performing output fine screening based on multiple continuous sub-windows.

[0059] When verifying the event alignment effect, the correspondence between the input event reference time and the output response reference time obtained through manual verification is used as the truth value. The absolute value of the time deviation of each paired event in the one-to-one association established by the system is calculated, and the average absolute error is statistically analyzed. In a preferred embodiment, the average absolute value of the event alignment error of the method of the present invention is 6ms to 15ms, preferably 8ms to 12ms; the average absolute value of the event alignment error of Comparative Embodiment 1 is 30ms to 80ms, preferably 40ms to 60ms. This shows that the present invention can significantly reduce the input-output mapping error and improve the event association accuracy through a unified time base correction mechanism based on the input-side event anchor point and the output-side response characteristic time.

[0060] When verifying the steady-state extraction effect, the steady-state window output by the system is manually reviewed. The proportion of the duration of the segment confirmed by the manual as a valid steady-state region is calculated out of the total duration of the final window, and this proportion is defined as the steady-state window purity. In a preferred embodiment, the steady-state window purity obtained by the method of the present invention is 85%–95%, preferably 88%–92%; the steady-state window purity of Comparative Embodiment 2 is 60%–78%, preferably 68%–75%. This shows that the present invention, through the combined mechanism of input-side steady-state candidate identification and output-side fine screening, can more effectively eliminate impact, transient, and strongly non-stationary segments, thereby improving the purity and comparability of steady-state data segments.

[0061] In a more specific example, the average absolute value of the event alignment error of the method of the present invention is 9.8 ms, and the steady-state window purity is 90.6%; correspondingly, the average absolute value of the event alignment error of Comparative Embodiment 1 is 47.6 ms, and the steady-state window purity of Comparative Embodiment 2 is 71.8%. This result shows that the present invention does not merely achieve synchronous acquisition of input and output signals, but improves the accuracy of input-output mapping and the quality of steady-state segment extraction through the synergistic effect of event anchor point alignment, operation stage identification, steady-state candidate constraints, and output fine screening. This verification method is consistent with the unified time base and event anchor point alignment, steady-state candidate identification, and output fine screening logic in Embodiment 1.

[0062] Example 6 To verify the technical effectiveness of this invention in preserving key segments, the same data source as in Example 5 was selected, and a key segment verification set was constructed from it. The key segment verification set includes at least one of the following: high-purity steady-state segments used for spectrum, envelope spectrum, or vibration intensity analysis; event-triggered segments containing abnormal events such as collisions, braking stops, track joint impacts, and gear skipping; rare operating condition segments with a small sample size but representative characteristics; and abnormal deviation segments that deviate from the center of historical normal samples under the same operating conditions and have potential diagnostic value. The above key segments were annotated by technicians in conjunction with manual review results.

[0063] To avoid the coarse value judgment caused by using a fixed rule retention method, this embodiment sets up a comparative implementation method 3: all data segments are saved using a fixed rule, that is, the original waveform with a high sampling rate is uniformly saved for data that is determined to be a steady state segment, and only the downsampled waveform or feature substitution data is uniformly saved for non-steady state segments, without calculating the retention value score for each data segment, and without comprehensively considering steady state purity, event significance, working condition scarcity, diagnostic deviation and storage cost.

[0064] The critical segment retention rate is defined as the proportion of data segments successfully retained by the system in either the original waveform form at a high sampling rate or a form that can be fully traced back, out of the total number of critical segments in the manually annotated critical segment verification set. The form that can be fully traced back here includes at least the format where the original waveform, downsampled waveform, critical sub-windows, and corresponding input variable snapshots can be restored to a complete event context through mapping indexes.

[0065] In a preferred embodiment, the critical fragment retention rate of the method of the present invention is 90%–98%, preferably 92%–96%; the critical fragment retention rate of Comparative Embodiment 3 is 68%–85%, preferably 75%–82%. In a more specific example, the critical fragment retention rate of the method of the present invention is 94.1%, and the critical fragment retention rate of Comparative Embodiment 3 is 79.4%.

[0066] Furthermore, under the same total storage budget, the method of this invention, by calculating retention value scores for various data segments and adaptively determining the retention level based on the retention value scores and current storage occupancy, can prioritize the retention of key segments with high purity steady-state analysis value, high anomaly backtracking value, scarce operating conditions, or diagnostic deviations; while reducing the original data retention level for segments with abundant samples, low diagnostic value, and high storage costs. Thus, without relying on continuous high sampling rate original storage of the entire dataset, this invention can still achieve a high retention rate of key segments, while balancing storage efficiency and subsequent backtracking capabilities. This verification method is consistent with the technical effects of retention value scoring, tiered storage, and reducing the burden of full storage in Example 1.

[0067] Example 7 To verify the technical effectiveness of this invention in retrieving fragments of the same or similar operating conditions, an operating condition sample library was constructed based on the data fragments retained in Embodiments 5 and 6. For each data fragment, the equipment type, mechanism field, stage field, load field, speed field, position field, braking field, linkage field, environment field, event anchor point sequence features, and linkage mode features were recorded, and a searchable sample set was established by combining the input-output mapping index table.

[0068] During verification, multiple query segments are randomly selected from the working condition sample library. Technical personnel then create a set of artificial reference segments for each query segment, based on equipment type, mechanism type, stage, load level, speed level, location range, linkage status, and event sequence characteristics. These artificial reference segments do not require complete consistency across all fields, but they must be comparable in subsequent condition evaluation, fault diagnosis, or trend comparison.

[0069] Setting up the comparison implementation method 4: only using the static multi-field precise matching method for fragment retrieval, that is, only when the equipment type, mechanism field, stage field, load field, speed field, position field and braking / linkage status field simultaneously meet the preset consistency conditions, the corresponding historical fragment is returned, without further calculating the proximity of continuous parameters, the topological similarity of event anchor sequence and the similarity of linkage mode.

[0070] The retrieval hit rate is defined as follows: for any query segment, if at least one of the top N candidate results returned by the system belongs to the set of manually referenced segments, it is counted as a hit; the ratio of the number of hits to the total number of queries in all query tasks is defined as the retrieval hit rate. N is preferably 5.

[0071] In a preferred embodiment, the retrieval hit rate of the method of the present invention under the Top-5 condition is 82% to 93%, preferably 85% to 90%; the retrieval hit rate of Comparative Embodiment 4 under the Top-5 condition is 50% to 72%, preferably 58% to 68%. In a more specific example, the Top-5 retrieval hit rate of the method of the present invention is 87.3%, and the Top-5 retrieval hit rate of Comparative Embodiment 4 is 61.5%.

[0072] The above results demonstrate that this invention, by incorporating discrete operating condition fields, continuous operating condition parameters, event anchor sequence features, and linkage mode features into the operating condition index and performing retrieval based on operating condition similarity, avoids the problem of a narrow retrieval range caused by relying solely on identical static fields. This improves the recall capability of historical segments under the same or similar operating conditions, providing more accurate comparable samples for subsequent condition evaluation, fault diagnosis, trend analysis, and life assessment. This verification method is consistent with the operating condition similarity retrieval and input / output mapping call logic in Example 1.

[0073] Example 8 To further verify that this invention is not a simple parallel combination of several independent technical measures, but rather a collaborative technical solution based on PLC input-driven data interception, organization, filtering, storage, and retrieval, the following ablation control system was constructed: System A: Based on the complete system, the pairing and correction of input-side event anchor points and output-side response characteristic moments are removed, and only unified time synchronization and coarse synchronization are retained; System B: Based on the complete system, remove the steady-state candidate constraints based on the input signal and the continuous sub-window output fine screening, and retain only the fixed window slice; System C: Based on the complete system, the retention value scoring and adaptive hierarchical storage are removed, and only fixed rules are used for storage; System D: Based on the complete system, the calculation of working condition similarity is removed, and only static field exact matching retrieval is used.

[0074] The complete system and systems A through D were verified separately, and the event alignment error, steady-state window purity, key fragment retention rate, and Top-5 retrieval hit rate were statistically analyzed.

[0075] In a preferred embodiment, the complete system outperforms any ablation control system in all four of the aforementioned indicators. Specifically, removing event anchor pairing correction significantly increases event alignment error; removing input-driven steady-state candidate identification and output screening significantly reduces steady-state window purity; removing retention value scoring significantly decreases key fragment retention rate; and removing operational condition similarity retrieval significantly decreases Top-5 retrieval hit rate. This demonstrates that the unified time base and event anchor alignment, steady-state candidate identification and output screening, retention value scoring and hierarchical storage, and operational condition similarity retrieval and mapping retrieval in this invention are not isolated but rather mutually supportive, progressively transmitted, and jointly affect the final data quality, fragment retention quality, and retrieval retrieval quality.

[0076] In a more specific example, the average absolute value of the event alignment error for the complete system was 9.8 ms, the steady-state window purity was 90.6%, the key fragment retention rate was 94.1%, and the Top-5 retrieval hit rate was 87.3%. Systems A, B, C, and D all showed significant declines in their respective core metrics, and some metrics had cascading effects. For example, increased event alignment error would further affect the accuracy of input-output mapping and weaken the comparability of subsequent operating condition retrievals; decreased steady-state window purity would affect retention value scoring and the quality of conditionalized samples; insufficient key fragment retention would weaken the effects of anomaly backtracking and historical sample accumulation; and decreased retrieval hit rate would directly weaken the effectiveness of comparisons with the same operating condition.

[0077] Therefore, this invention elevates the PLC input signal from a traditional control variable to a dominant variable for data acquisition and organization, forming a complete link technology solution: "unified time base and event anchor point alignment—operation stage identification—steady-state candidate identification—output fine screening—three-category segment construction—hierarchical storage—operational condition similarity retrieval—input-output mapping and calling." This complete link solution achieves stable improvements in indicators such as event alignment error, steady-state window purity, key segment retention rate, and retrieval hit rate, thus demonstrating its innovation and superiority over the traditional method of relying solely on output-side signal acquisition and continuous storage. This conclusion is consistent with the five beneficial effects described in the invention description section of this specification.

[0078] In another embodiment, the above method can be deployed in a collaborative architecture of edge computing gateways, industrial computers, and cloud platforms. The edge side is responsible for real-time data acquisition, standardization, event anchor point identification, steady-state candidate window identification, output fine screening, construction of three types of fragments, and index generation; the cloud side is responsible for long-term archiving, index retrieval, data access, and unified management across devices. The method steps of this invention can be implemented by the processor executing program instructions stored in memory.

[0079] This invention elevates PLC input signals from simple control variables to dominant variables for data extraction and organization. It constructs a steady-state window extraction, three-segment construction, hierarchical storage, and input-output mapping scheme adapted to the scenarios of strong non-steady-state conditions, multi-mechanism linkage, and high-sampling data in port cranes. This provides a structured, traceable, and efficiently stored data foundation for various subsequent PHM applications.

[0080] 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 extracting and hierarchically storing steady-state data of port machinery based on PLC event anchor points, characterized in that, The method includes the following steps: The system collects the equipment input signals and equipment output signals of the port crane and generates timestamps for each; wherein the equipment input signals include at least two of the following: load information, target mechanism speed information, target mechanism position information, running status word, braking status information, and linkage flag information; and the equipment output signals include at least one of the following: vibration acceleration signal, vibration velocity signal, and temperature signal. The device input signal and the device output signal are processed using a unified time base. Based on the time synchronization, input-side event anchor point extraction, output-side response feature moment extraction, time deviation robust correction, and segmented drift compensation results, a one-to-one correlation relationship between the input signal and the output signal is established. A work phase state machine is constructed based on the device input signals to identify the phase boundaries of the target mechanism in a work cycle and determine the range of the target phase to be captured. Within the target stage range, steady-state candidate constraints are applied based on the equipment input signals to construct a port machinery steady-state index and identify steady-state candidate windows; Within the steady-state candidate window, the output signal of the device is finely screened, and the steady-state candidate window is subdivided into multiple continuous sub-windows. A non-stationary quantification index is constructed based on at least one time-domain feature, frequency-domain feature, envelope feature and / or statistical feature. Impact, transient and strongly non-stationary segments are eliminated to obtain the steady-state window for subsequent analysis. Based on the steady-state window, event trigger window, and comparison window, three types of data segments are constructed, and a retention value score is calculated for each data segment. The hierarchical storage threshold is dynamically adjusted according to the storage occupancy rate so that the three types of data segments are saved as high-sampling-rate original waveforms, downsampled data, and / or feature substitution data, respectively. Generate operating condition indexes for each data segment and establish input-output mapping records. Perform data retrieval and calling based on the similarity of operating conditions to support condition evaluation, fault diagnosis, trend analysis and / or life assessment.

2. The method according to claim 1, characterized in that, The unified time base processing includes: synchronizing and standardizing the timestamps of the device input and output signals to obtain a coarse synchronization timeline; extracting input-side event anchor point timestamps from the device input signals; extracting output-side response feature timestamps related to the corresponding events from the device output signals within a preset search window, centered on each input-side event anchor point timetamp; calculating the original time deviation between the successfully paired input-side event anchor point timestamps and the output-side response feature timestamps; robustly correcting the original time deviation based on the current time deviation and the time deviations of neighboring events; performing segmented drift compensation between adjacent event anchor points; and adding timestamps to the device output signals in conjunction with the effects of braking state switching and mechanism linkage interference. The correction is performed to obtain a correction timestamp under a unified time base; wherein, the input-side event anchor point time includes at least one of the following: brake release event, brake holding event, start event where the target mechanism speed jumps from zero to non-zero, stop event where the target mechanism speed jumps from non-zero to zero, linkage start event, linkage end event; the output-side response characteristic time is: within a preset search window centered on the input-side event anchor point time, the moment when the short-time energy, short-time root mean square, envelope amplitude, frequency band energy, or combination thereof of the device output signal first satisfies the preset response significance condition; the response significance condition includes: the statistic relative to the pre-event baseline exceeds a preset threshold, and / or remains significantly deviated from the baseline for a preset duration.

3. The method according to claim 2, characterized in that, The original time deviation expression is: ； Correcting time deviation for: ； After obtaining the corrected time deviation, any two adjacent output-side response characteristic moments... and Segmented drift compensation and additional compensation are applied to the output timestamps t between them to obtain the corrected timestamps under a unified time base. : ； in, For the input-side event anchor point time, The characteristic moment of the output-side response; The weighted coefficients for the corresponding events and institutions. This represents the median of the time deviations of a preset number of neighboring events before and after the current event, where r is the number of neighboring events parameter; , These are the compensation weights for braking hysteresis and linkage interference, respectively. As the braking hysteresis factor, This refers to the linkage interference coefficient; The braking hysteresis factor Based on the time difference between the braking command time and the time when the equipment output signal shows a significant response, and normalized according to the length of the anchor point interval, the expression is as follows: ； in, The timing of the PLC-side braking state switching command. This refers to the significant response moment in the device's output signal corresponding to the switching of the braking state. To prevent extremely small positive numbers with a denominator of zero; The linkage interference coefficient The expression is determined based on the relative motion intensity between the target mechanism and the associated mechanism within the anchor point interval: ； in, Let the velocity of the target mechanism be at the k-th sampling time. The velocity of the associated mechanism at the k-th sampling time. This represents the number of sampling points involved in the calculation within the i-th anchor point interval.

4. The method according to claim 1, characterized in that, Within the sliding time window, the comprehensive velocity stability constraint... Load stability constraint Constraint on the smoothness of speed change Braking state consistency factor Interference coefficient and the proportion of space-sensitive areas Constructing a port machinery steady-state index The expression is: ； Where w1 to w6 are the contribution coefficients corresponding to each constraint term, satisfying: ； The contribution coefficient is preset based on the target organization type, the operation stage type, historical calibration samples, and / or empirical rules. Only if the steady-state exponent satisfies: ； Furthermore, the lower limit of the steady-state index of the preset sub-window within the sliding time window satisfies: ； The sliding time window is then determined as a steady-state candidate window; wherein, The threshold for determining the entire window. The threshold for determining the lower limit of the sub-window is satisfied. .

5. The method according to claim 1, characterized in that, The steady-state candidate window is subdivided into multiple continuous sub-windows. For the k-th sub-window, the spectral broadening ratio is calculated, and a non-stationary quantization index is constructed based on the time-domain energy stability, frequency-domain structure stability, envelope statistical stability, impact degree, and spectral broadening degree of each continuous sub-window. When the non-stationary quantization index is not greater than a preset non-stationary threshold, and at least a preset number of continuous sub-windows simultaneously satisfy the impact factor threshold and the spectral broadening ratio threshold constraints, the at least preset number of continuous sub-windows are reorganized into a steady-state window. The spectral broadening ratio is the minimum effective bandwidth required to cover 95% of the total spectral energy. The minimum effective bandwidth required to cover 50% of the total spectral energy. The ratio is calculated using the following formula: ； in, To prevent extremely small positive numbers with a denominator of zero; The expression for the non-stationary quantitative indicator is as follows: ； in, It is a non-stationary quantitative indicator. It is the root mean square. The coefficient of variation represents the root mean square of each sub-window, used to characterize the consistency of temporal energy within the candidate window; For the centroid of the spectrum, The coefficient of variation represents the spectral centroid of each sub-window, used to characterize the consistency of the frequency domain structure within the candidate window; For the envelope entropy, Used to characterize the degree of change in the statistical structure of the envelope of adjacent sub-windows; As the impact factor, Used to characterize the maximum impact level in each sub-window; This is the average of the spectral broadening ratios of each sub-window, used to characterize the overall spectral diffusion of the candidate window; , , , , The contribution coefficients for each item.

6. The method according to claim 1, characterized in that, The event trigger window is used to capture one or more events such as collision, abnormal impact, braking stop, gear skipping, and track irregularity impact, and saves the original waveform of a preset length before and after the event, as well as a snapshot of the PLC input variables at the corresponding time. The comparison window extracts segments with the same window length and the same output signal type under different load ranges, different speed ranges, different position ranges, and / or different operation types to construct a conditional input-output mapping sample library. For data segments that do not retain the original waveform, one or more of the following are extracted as feature replacement data: root mean square vibration, peak-to-peak value, kurtosis, envelope energy, dominant frequency amplitude, and temperature statistics.

7. The method according to claim 1, characterized in that, The retention value score is calculated for each data segment, wherein the retention value score The expression is: ； In the formula, the steady-state purity Q(G) is determined based on the steady-state candidate identification result, the output fine screening result, and the continuous effective steady-state duration corresponding to the data segment; the event salience E(G) is determined based on the alarm bit, fault word, speed change degree, and / or output signal impact intensity; the working condition scarcity R(G) is determined based on the number of historical samples under the working condition index corresponding to the data segment, and is used to characterize whether the segment belongs to a relatively rare working condition in the sample library; The diagnostic deviation degree D(G) is determined based on the degree of deviation of the feature vector of the data segment from the center of the historical normal samples under the same working condition. It is used to characterize whether the segment, although not triggering an explicit alarm, has shown an abnormal deviation trend relative to the center of the normal working condition. The storage cost C(G) is determined based on the data volume corresponding to the segment sampling rate, number of channels, duration, and storage mode. It is used to characterize the storage resources required to store the segment. The working condition index includes at least three of the following: mechanism type, stage identifier, load range, speed range, position or angle range, and braking state or linkage state. The expression for the scarcity of the operating condition R(G) is: ； The storage cost C(G) is expressed as follows: ； In the formula, For fragments The number of historical samples in the corresponding operating condition index bucket. This is the threshold for the number of saturated samples; Sampling rate, For the number of channels, For the duration of the segment, The number of bytes of data per sampling point. This is the normalized reference constant.

8. The method according to claim 7, characterized in that, The tiered storage threshold is dynamically adjusted based on the retention value score and current storage occupancy rate. When the retention value score is not lower than the first threshold, the high-sampling-rate original waveform and corresponding input segment are saved. When the retention value score is between the first and second thresholds, downsampled data, key sub-windows, and / or feature substitution data are saved. When the retention value score is lower than the second threshold, only feature substitution data and index records are saved. Simultaneously, the operating condition index uses a multi-field combination identification method and performs data retrieval for the same or similar operating conditions based on operating condition similarity. When the operating condition similarity is not lower than a preset similarity threshold, the corresponding data segment is determined to belong to a similar operating condition segment, and the corresponding input segment, input variable snapshot, output original waveform, downsampled data, and / or feature data are retrieved through the input-output mapping index table. The operating condition similarity is calculated according to the following formula: ； in, For operating condition similarity, For consistency of discrete fields, For continuous parameter proximity, For topological similarity of event anchor sequence, Similarity in linkage patterns; The consistency of discrete fields The expression used to characterize whether two segments are consistent in discrete fields is: ； Where D is the set of discrete fields being compared. (·) This is an indicator function; it takes the value 1 when the corresponding fields match, and 0 otherwise. The topological similarity of the event anchor sequence The expression is: ； in, and Each represents a segment and fragments The corresponding event anchor sequence, This represents the length of the longest common subsequence of the two sequences; In addition, the continuous parameter proximity Similarity to the aforementioned linkage mode It is calculated using the exponential negative distance method.

9. A port machinery steady-state extraction and hierarchical storage system based on PLC event anchor points, characterized in that, The system is configured to perform the method as described in any one of claims 1 to 8, the system comprising: The signal acquisition and time base alignment module is used to acquire the equipment input signals and equipment output signals of the port crane and generate timestamps for each; it performs unified time base processing on the equipment input signals and equipment output signals, and establishes a one-to-one correlation between the input signals and output signals based on time synchronization, input-side event anchor point extraction, output-side response feature moment extraction, time deviation robust correction, and segmented drift compensation results; wherein, the equipment input signals include at least two of the following: load information, target mechanism speed information, target mechanism position information, running status word, braking status information, and linkage flag information, and the equipment output signals include at least one of the following: vibration acceleration signal, vibration velocity signal, and temperature signal; The work phase identification module is used to construct a work phase state machine based on the equipment input signals, identify the phase boundaries of the target mechanism in a work cycle, and determine the range of the target phase to be captured. A steady-state candidate identification module is used to apply steady-state candidate constraints based on the equipment input signal within the target stage range, construct a port machinery steady-state index and identify steady-state candidate windows; The output screening module is used to perform output screening on the output signal of the device within the steady-state candidate window, subdivide the steady-state candidate window into multiple continuous sub-windows, construct non-stationary quantification index based on at least one time-domain feature, frequency-domain feature, envelope feature and / or statistical feature, eliminate shock, transient and strongly non-stationary segments, and obtain a steady-state window for subsequent analysis. The data fragment construction and hierarchical storage module is used to construct three types of data fragments based on the steady-state window, event trigger window and comparison window, calculate the retention value score for each data fragment, and dynamically adjust the hierarchical storage threshold according to the storage occupancy rate, so as to save the three types of data fragments as high sampling rate original waveforms, downsampled data and / or feature substitution data respectively. The operating condition indexing and retrieval module is used to generate operating condition indexes for each data segment and establish input-output mapping records. It also performs data retrieval and retrieval based on the similarity of operating conditions to support condition evaluation, fault diagnosis, trend analysis and / or life assessment.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the port machinery steady-state extraction and hierarchical storage method based on PLC event anchor points as described in any one of claims 1 to 8.

Images