Escalator intelligent operation and maintenance early warning method and system

Abstract

The application discloses an escalator intelligent operation and maintenance early warning method and system, collects running state data and braking timing data, positions a driving component friction accumulation section through an environment corrected running deviation trend, and identifies speed aggregation characteristics of elastic failure and friction failure through batch dispersion of shutdown time length; generates component wear marks by comparing state differences of each speed section of the friction section, constructs a fault mode after being associated with a braking failure type, determines a fault urgency rating based on a decline inflection point of a power consumption recovery rate after previous similar maintenance, determines a maintenance window according to a residual intervention time length margin of each fault, generates a maintenance task scheme by matching the lowest level effective intervention type, selects high-quality dispatched personnel and plans a response route according to a historical recovery rate and a rework rate, scores a power consumption recovery amplitude in a first peak period after maintenance and updates a monitoring threshold to output an operation and maintenance early warning instruction, and early perception and closed-loop operation and maintenance management capabilities of escalator mechanical degradation are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for special equipment, and in particular to an intelligent operation and maintenance early warning method and system for escalators. Background Technology

[0002] During long-term service, the two core mechanical components of escalators—drive chains and brakes—deteriorate slowly due to continuous cyclic loads. In the early stages, this deterioration is often invisible, manifesting only in subtle changes in operating energy consumption and shutdown response characteristics. Current maintenance strategies rely heavily on periodic manual inspections and experience-based judgment, lacking quantitative methods for extracting early signs of deterioration. Furthermore, the correlation between energy consumption fluctuations and changes in mechanical condition is affected by external factors such as passenger flow and seasonal temperature variations. Directly comparing operating data from different batches makes it difficult to distinguish between genuine deterioration and normal fluctuations caused by environmental changes.

[0003] The decline in recovery effectiveness after each repair of the same type of fault is an observable indicator that a component has entered the irreversible degradation stage. However, this decay pattern lacks quantitative tracking in the existing operation and maintenance system. The various links of repair timing judgment, dispatch personnel selection and repair effect evaluation are isolated from each other, and the execution results cannot be fed back to subsequent monitoring parameters. Therefore, the improvement of operation and maintenance quality lacks data closed-loop support. Summary of the Invention

[0004] This invention discloses an intelligent operation and maintenance early warning method and system for escalators. It aims to locate the two types of degradation sources, drive and braking, through joint analysis of the deviation characteristics of the operating status and the discrete characteristics of the shutdown response. It combines the correlation modeling of component wear type and braking anomaly source, quantifies the urgency of the fault by the decay inflection point of the power consumption recovery rate of each maintenance, prioritizes maintenance according to the remaining time margin and matches the lowest level intervention type, selects dispatch personnel based on historical quality and efficiency records, and feeds back the maintenance recovery effect to the monitoring threshold to realize the dynamic updating of early warning instructions.

[0005] The first aspect of this invention proposes a smart operation and maintenance early warning method for escalators, comprising the following steps: Collect operating status data and braking timing data, calculate the deviation between each operating status and the benchmark from the operating status data to generate the operating status deviation, and collect the braking cycle dispersion based on the dispersion of the braking cycle duration of each batch of braking timing data. Based on the deviation of the operating state, the step accumulation boundary is located to form a friction zone, and the speed segment with high discrete batch aggregation is identified by using the discrepancy of the braking cycle to determine the braking anomaly source marker. The wear marks of components are generated by comparing the state differences of each speed segment in the friction section. The wear marks of components are used to identify the components related to the brake anomaly source to form an associated fault mode. Based on the associated fault mode, the decline in the power consumption recovery rate of previous similar repairs is identified to determine the fault urgency rating. Based on the fault urgency rating, the critical time period of each fault combination is analyzed to obtain the critical margin set. The critical margin set is used to determine the nearest maintenance time period of the item with the smallest margin to obtain the urgent maintenance window. From the urgent maintenance window, the lowest level maintenance type that can restore friction is identified to establish a maintenance task plan. Based on the maintenance task plan, the personnel with the best performance in the same fault in history are selected to form a response dispatch plan. The power consumption deviation during the first peak period after maintenance is extracted from the response dispatch plan to obtain the maintenance performance score. The component monitoring threshold is updated according to the maintenance performance score and maintenance early warning instructions are output.

[0006] A second aspect of this invention provides an intelligent operation and maintenance early warning system for escalators, comprising: The data acquisition module is used to collect operating status data and braking timing data. It generates operating status deviation by statistically analyzing the deviation between each operating status and the benchmark from the operating status data, and obtains braking cycle dispersion by aggregating the dispersion of braking cycle duration for each batch based on the braking timing data. The trend analysis module is used to locate the step accumulation boundary and form the friction zone based on the deviation of the operating state, and to identify the speed segment with high discrete batch aggregation by using the braking cycle discrete quantity to determine the braking anomaly source marker. The fault diagnosis module is used to compare the state differences of each speed segment in the friction section to generate component wear marks, identify the components related to the brake anomaly source mark based on the component wear marks to form an associated fault mode, and identify the decline in the power consumption recovery rate of previous similar repairs based on the associated fault mode to determine the fault urgency rating. The maintenance planning module is used to analyze the critical time period of each fault combination based on the fault urgency rating to obtain a critical margin set, use the critical margin set to determine the nearest maintenance time period of the item with the smallest margin to obtain an urgent maintenance window, and identify the lowest level maintenance type that can restore friction from the urgent maintenance window to establish a maintenance task plan. The instruction output module is used to select the personnel with the best historical quality and efficiency for the same fault based on the maintenance task plan to form a response dispatch plan, extract the power consumption deviation during the first peak period after maintenance to obtain a maintenance quality and efficiency score based on the response dispatch plan, update the component monitoring threshold based on the maintenance quality and efficiency score, and output maintenance early warning instructions.

[0007] The beneficial effects of this invention are reflected in the following points: First, by extracting the consistency of the deviation direction of consecutive batches from the time sequence of the operating state after environmental temperature and humidity correction, the identification of friction accumulation sections is not affected by fluctuations in environmental temperature and humidity. Elastic failure signals are perceived from the trend of the ratio of the dispersion within the braking stop duration batches between high-load and low-load stop batches. The density peak values ​​of high-dispersion batches in each speed range distinguish between elastic failure and friction failure. The two degradation signals are independently extracted from different data dimensions and jointly point to the degradation location, solving the problem that a single indicator cannot simultaneously perceive both types of degradation sources. Second, the inflection point in the time sequence of the normalized recovery rate of power consumption after each similar maintenance, where it changes from a uniform decrease to an accelerated decrease, is used as the time sequence anchor point for the component entering the irreversible degradation stage. The conservative quantile of the actual failure interval of similar historical faults after the inflection point determines the remaining intervention margin for each fault. This transforms the fault urgency, which originally relied on manual experience, into a quantifiable margin that can be ranked. Furthermore, the causal relationship between component wear type and braking failure source is introduced into historical sample retrieval, making the inflection point identification samples closer to the current actual state of the equipment. Finally, the lowest effective intervention type is matched with the low quantile of the historical maintenance recovery rate to avoid excessive maintenance occupying window time; dispatch personnel are jointly screened by the median of the historical normalized recovery rate and the rework rate, and response routes are planned in combination with real-time location; after maintenance, the quality and efficiency score is calculated based on the first peak power consumption recovery rate, and the score results are mapped to the tightening or loosening of subsequent monitoring thresholds. Each maintenance result directly affects the sensitivity of the next round of detection, and the difference in operational quality and efficiency is incorporated into the closed-loop feedback rather than just recorded in history. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating an intelligent operation and maintenance early warning method for escalators according to the present invention.

[0009] Figure 2 This is a structural block diagram of an intelligent operation and maintenance early warning system for escalators according to the present invention. Detailed Implementation

[0010] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0011] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0012] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0013] The technical solutions of the embodiments of this application will be described below.

[0014] like Figure 1 As shown, this embodiment of the invention provides a smart operation and maintenance early warning method for escalators, including the following steps S110-S150: Step S110: Collect operating status data and braking timing data; statistically analyze the deviation between each operating status and the benchmark from the operating status data to generate the operating status deviation; and collect the braking cycle dispersion based on the dispersion of the braking cycle duration of each batch according to the braking timing data.

[0015] Specifically, operational status data and braking timing data are collected. Operational status data is continuously collected during normal operation by vibration sensors (100Hz, installed on both sides of the main drive bearing housing and the step chain guide rail), temperature sensors (0.1Hz, distributed between the main drive bearing and the reducer housing), three-phase current transformers (10Hz), and ambient temperature and humidity sensors (0.05Hz, installed in the equipment room and escalator operating area). The data is organized using a dual index of batch number and sampling time, divided into batches by operating day, with the timing of each sensor aligned to the same reference clock. Batches with failed sensor readings are marked with a missing component identifier. For batches with missing identifiers, the deviation from the initial value is interpolated using the average of adjacent valid batches, with an additional interpolation identifier attached to the interpolation source. Braking timing data is recorded by the electromagnetic brake action recorder with millisecond-level precision, capturing the start and end times of each braking event. The start and end times, trigger cause codes, operating speed, and ambient temperature of the day are stored using the braking event number as the primary key. Trigger cause codes distinguish between four types of events: normal stop, passenger flow clearing stop, emergency braking, and inspection-mandated stop. Inspection-mandated stops are constrained by control logic and have concentrated durations. During the braking timing sequence generation phase, inspection-mandated stop events are separately categorized into an auxiliary sequence and not mixed with operational stops. The timestamps of the operating status data and braking timing data must be aligned to the same reference clock. If the alignment deviation exceeds 100ms, the corresponding batch is marked with a clock deviation identifier. In subsequent joint analysis phases of operating status and braking, the start time association window for batches marked with clock deviation is widened to cover the timing offset introduced by the deviation.

[0016] In some embodiments, the step of generating an operating state deviation by statistically analyzing the deviations between each operating state and the baseline from the operating state data includes: extracting operating state features from each operating segment of the operating state data to obtain an operating state sequence; calculating the difference between the mean of each operating state and the historical baseline mean for the operating state sequence to obtain an initial value of state deviation; evaluating the batch-to-batch consistency of the deviation amplitude after environmental temperature and humidity correction based on the initial value of state deviation to form an environmental correction deviation; and aggregating the mean of each state deviation after environmental correction based on the environmental correction deviation to determine the operating state deviation.

[0017] Operating status sequences are obtained by extracting state characteristics from each operating segment of the operating status data. The data collected by vibration, temperature, and current sensors during the steady-state operation of the escalator most stably reflects the true mechanical state of the drive components and serves as the benchmark data source for tracking friction degradation trends. For each batch of operating status data sampling sequences, the initial acceleration segment (the first 30 seconds before speed stabilization) and the deceleration segment (the last 15 seconds) are removed. Within the steady-state operating segment, the root mean square value of vibration acceleration, the average temperature at each measuring point, and the average three-phase current are averaged separately using a 5-second sliding window and then concatenated to form the vibration characteristic sequence, temperature characteristic sequence, and current characteristic sequence for that batch. These three sequences are aligned by batch number to form the operating status sequence for that batch. The removal process eliminates the bias caused by transient speed changes in steady-state characteristic estimation. During the initial startup of escalators in cold seasons, the viscosity of the lubricating oil film in the step chain is higher, and the vibration and current characteristic values ​​are significantly higher than the normal operating temperature level. When the difference between the average value of the starting phase and the average value of the steady-state phase of each batch exceeds a certain proportion, a cold start label is added. The starting time of the steady-state phase of the operating state sequence of the batch with the cold start label is delayed until all characteristic values ​​have stabilized before collection. The delay is determined by the average value of historical cold start batches to avoid the data from the warm-up transition phase increasing the estimated steady-state average. For batches with missing sensor labels in the operating state data, the corresponding operating state sequence is constructed using interpolated values. When the batch with the interpolated label participates in the calculation of the state deviation from the initial value, the weight is reduced to 0.7 times to prevent insufficient interpolation accuracy from affecting the deviation trend. When the length of each batch of the operating state sequence is less than the minimum number of sampling points, a short sequence label is added. Short sequences usually appear in batches that are abnormally interrupted during operation. The uncertainty of the mean estimation of the short sequence label batch is higher than that of the standard batch, and its weight is reduced in the stage of calculating the state deviation from the initial value.

[0018] The initial deviation value is obtained by statistically analyzing the difference between the mean of each operating state and the historical baseline mean in the operating state sequence. For each batch in the operating state sequence, the root mean square of vibration, the mean of temperature, and the mean of current in the steady-state segment are subtracted from their corresponding historical baseline mean values. These three deviations are then normalized to their respective dimensions and weighted according to their sensitivity weights to friction degradation to obtain the initial deviation value for that batch. The historical baseline mean is the average of similar operating batches within the first three months of equipment operation. The first three months correspond to the operating state after the escalator's break-in period, when the drive chain friction tends to stabilize. Thereafter, the deviation reflects the actual degradation relative to the initial healthy state, rather than the break-in effect. The start time of the steady-state segment for batches identified by cold starts in the operating state sequence has been delayed and corrected. The corresponding initial deviation value is consistent with the benchmark comparison caliber of non-cold start batches. For batches originating from cold starts, a cold start source identifier is added to the initial deviation value. These batches are not included in the calculation of the average value during the environmental correction deviation assessment stage to prevent a systematic increase in cold start values ​​from raising the environmental baseline and underestimating the actual degradation of other batches. The initial value of the state deviation corresponding to the batch of the operation state sequence interpolation is included in the statistics with a weight of 0.7. The initial value of the state deviation is assigned a low confidence flag to the batch from which the interpolation originates. The low confidence flag is transmitted synchronously without interruption during the subsequent environmental correction and aggregation stages. When the initial value of the state deviation is positive for three consecutive batches and the absolute value increases batch by batch, a unidirectional cumulative flag is marked. As the pitch of the ladder chain gradually lengthens due to wear, the frictional resistance between it and the sprocket increases monotonically. The unidirectional cumulative flag is the macroscopic manifestation of this mechanical degradation in the state deviation sequence. The cumulative slope of the corresponding batch is recorded separately when the operation state deviation is output for use in step boundary identification.

[0019] The environmental correction deviation is calculated based on the consistency of the deviation amplitude after temperature and humidity correction, determined by the initial state deviation value. The initial state deviation value is grouped according to two dimensions: ambient temperature (in 5°C increments) and relative humidity (in 10% increments). The average initial state deviation value of historical batches within the same environmental group is defined as the environmental correction reference value for that group. Subtracting the corresponding group reference value from the current batch's initial state deviation value yields the corrected deviation after environmental effect isolation. The environmental correction deviation is stored using the batch number as the primary key, along with the corresponding environmental group identifier. The viscosity of the drive chain lubricating oil film varies by approximately 5% to 15% with ambient temperature. Under high humidity, a water film adheres to the guide rail, resulting in lower contact friction. Direct comparison of state deviations across batches can misjudge fluctuations caused by temperature and humidity as degradation signals. The two-dimensional grouping quantitatively isolates the environmental effect from the initial state deviation value. Batches with the initial state deviation value originating from a cold start are not included in the calculation of group reference values ​​to prevent high cold start values ​​from raising the environmental group baseline and thus underestimating the true degradation of other batches. When the number of historical batches in the same environmental group is less than the minimum sample threshold, the reference value is replaced by the average of all batches, with an additional grouping deficiency flag. For batches with missing environmental temperature and humidity sensor readings, the grouping is determined by interpolating the environmental averages of the two batches before and after the missing batches, and the source of the missing interpolation is marked with an environmental missing flag. When the absolute value of the environmental correction deviation for each batch exceeds a certain percentage of the absolute value of the deviation from the initial state value, an overcorrection flag is marked. Overcorrection indicates that the estimated deviation of the environmental grouping reference value is too large. The deviation of the operating state of the batch marked with overcorrection is conservatively substituted with the initial state deviation value before correction.

[0020] The operational status deviation is determined by aggregating the average of each state deviation after environmental correction based on the environmental correction deviation. The state deviation time series after environmental effect stripping reflects the true friction degradation trend of the escalator drive system relative to the healthy baseline. The fluctuation of state deviation, which briefly declines after lubrication and maintenance and then rises again due to wear, is superimposed on the long-term degradation trend. A sliding window is needed to smooth short-term fluctuations to extract reliable trend signals. The average of the correction deviations of each batch of environmental correction deviation is taken according to the sliding window to form the main time series of operational status deviation. Within the window, the weight of low confidence batches is reduced to 0.7 times, and the weight of overcorrection batches is reduced to 0.8 times. When both types of reduction are triggered simultaneously, they are multiplied and superimposed. The comprehensive weight is not less than 0.56 times to avoid the correction deviation of a single batch being completely ignored due to excessively low weight. The confidence field of the operational status deviation corresponding to the batches with missing environmental indicators and insufficient grouping indicators of environmental correction deviation is reduced according to the difference between the historical environmental fluctuation amplitude and the average value of each environmental group. The confidence reduction of outdoor open places is higher than that of constant temperature indoor environments. When the standard deviation of the corrected deviation sequence within the sliding window of the operating status deviation exceeds a certain percentage of the absolute value of the window mean, a high fluctuation indicator is displayed. High fluctuations typically occur in transitional batches where the deviation briefly decreases immediately after lubrication and maintenance, then rises again due to wear. During the friction resistance section positioning stage, the confidence level of the corresponding boundary decreases. The ratio of the slope of the operating status deviation within the corresponding batch range to the average slope of the entire batch constitutes the friction resistance acceleration index. When the friction resistance acceleration index exceeds a certain multiple, friction resistance acceleration warnings are added to the operating status deviation data. These warnings are used for subsequent step boundary identification to capture degradation acceleration signals earlier.

[0021] In some embodiments, obtaining the braking cycle dispersion based on the dispersion of braking cycle duration for each batch based on the braking timing data includes: extracting the start and end time intervals of each braking event from the braking timing data to generate a braking duration sequence; calculating the ratio of the duration range to the mean of each batch in the braking duration sequence to form a batch duration dispersion; identifying the trend of the ratio of dispersion between high-load and low-load shutdown batches based on the batch duration dispersion to form a discrete increasing record; and determining the braking cycle dispersion based on the deviation ratio of the dispersion of each batch from the initial mean by the discrete increasing record.

[0022] Braking duration sequences are generated by extracting the start and end time intervals of each braking event from braking timing data. The duration of each action of the escalator electromagnetic brake is determined by both the spring preload and the contact state of the friction pads. Under healthy conditions, the braking durations of each braking event under the same operating condition are highly consistent. However, when the springs fatigue or the friction pads wear unevenly, observable discrete fluctuations begin to appear in the braking durations. Therefore, extracting the braking durations from the braking timing data to form a batch sequence is the basic data for quantifying braking degradation. The braking duration of a single braking event is obtained by subtracting the start time from the stop time of each braking event in the braking timing data. The durations of all operational stop events within the same operating day are arranged chronologically to form the batch braking duration sequence. Events triggered by inspection-mandated shutdowns are separately classified into an auxiliary sequence and are not mixed with operational shutdowns in the statistics. Inspection shutdowns are triggered by external commands and their durations are constrained by control logic, resulting in a much lower degree of dispersion than operational shutdowns. Mixing them into the main sequence would suppress the overall batch duration dispersion estimate, thereby masking the true upward trend of dispersion caused by braking component degradation. The braking timing data clock deviation identifier indicates that the absolute value of the braking duration sequence corresponding to the batch has an estimation error. Clock deviation source identifiers are added to the braking events that are the source of the clock deviation in the braking duration sequence. The weight of the clock deviation source events is reduced during the batch duration dispersion calculation stage to reduce their impact on extreme value estimation. When the number of events in each batch of the braking duration sequence is lower than the minimum sample threshold, a sparse batch identifier is marked. Batches with very few escalator downtimes during an operating day are more likely to trigger the sparse identifier. Sparse batches are merged with adjacent batches of braking duration sequences for statistical analysis, with the merging range not exceeding two batches before and after. The merged source triggers a weight reduction during batch duration dispersion calculation to reflect the decrease in cross-batch representativeness.

[0023] The batch duration dispersion is calculated by comparing the range of each batch's duration with the mean in the braking duration sequence. When the brake spring preload is uniform, the duration difference between each braking event is minimal. However, spring fatigue or localized wear of the friction pads leads to uneven torque output between different braking events, thus widening the duration range. Measuring the dispersion by the ratio of the range to the mean eliminates the interference from absolute duration differences between batches. The batch duration dispersion CV_k is obtained by dividing the difference between the maximum and minimum durations of the main sequence of each batch of the braking duration sequence by the batch mean. The calculation formula is as follows: CV_k = (T_max,k - T_min,k) / T_mean,k, where T_max,k and T_min,k are the maximum and minimum values ​​of the main sequence of the k-th batch of the braking duration sequence, and T_mean,k is the average braking duration of the same batch (in seconds). A stable CV_k across batches indicates uniform preload in the braking system, while a continuous increase in CV_k suggests spring fatigue or uneven wear of the friction pads. The clock skew source events in the braking duration sequence have been downweighted. T_max,k and T_min,k are both determined by the extreme values ​​of the main sequence after removing downweighted events, avoiding abnormal duration values ​​introduced by clock skew from becoming extreme values ​​that skew the CV_k estimate. Sparse batch identifiers in the braking duration sequence have been merged with adjacent sequences. After merging, the T_mean,k sample size is sufficient, improving the robustness of the CV_k estimate. The weight of the merged source batch CV_k is reduced during the discrete increasing record identification stage to reflect the decrease in temporal representativeness introduced by cross-batch merging. Batch duration dispersion is marked as high-dispersion batches when the CV_k of each batch exceeds twice the median of the entire batch. Repeated triggering of high-dispersion batches within consecutive operating days indicates observable performance inhomogeneity in the braking system. When the number of high-dispersion batch identifiers exceeds 5 within 20 consecutive batches, a discrete clustering warning is added. This warning is used for subsequent high-dispersion batch aggregation determination.

[0024] To identify the trend of the ratio of high-load and low-load stoppage batch dispersion, a discretely increasing record is formed. Each batch of braking events is categorized into two types based on the corresponding passenger flow density: high load (passenger flow density exceeding 60% of the daily average) or low load (below 20% of the daily average). The ratio of the mean CV_k of high-load batches to the mean CV_k of low-load batches is defined as the load difference dispersion ratio. The load difference dispersion ratio forms a discretely increasing record time sequence using a monthly sliding window. During periods of low passenger flow, the escalator braking spring operates with a smaller compression. Spring fatigue leading to insufficient preload force manifests earlier in low-load conditions, thus affecting braking duration. In contrast, under high-load conditions, the large compression temporarily compensates for preload attenuation, masking the degradation effect. Therefore, the continuous decrease in the load difference dispersion ratio becomes an early observable signal of elastic component degradation, detecting the degradation trend earlier than the traditional judgment relying solely on absolute duration exceeding the threshold. For batches with high dispersion in duration, a low-load concentration marker is added when the batch identifier is concentrated in the low-load group, and a high-load concentration marker is added when the batch identifier is concentrated in the high-load group. These two types of markers are stored separately in the discrete increasing record to distinguish between elastic failure and friction failure, two sources of degradation. Low-load concentration corresponds to spring elastic failure, and high-load concentration corresponds to accelerated wear of friction plates under high load. For batches with sparse dispersion in duration, the source batches are merged and classified into corresponding groups according to the original load classification of each batch before merging, preventing misclassification introduced by merging cross-load batches. For discrete increasing records, when the load difference dispersion ratio decreases unidirectionally for three consecutive months, a trend increasing marker is added. The trend increasing marker triggers a shortened statistical window in the braking cycle dispersion calculation stage to improve the timeliness of the degradation trend response.

[0025] The braking cycle dispersion is determined by the deviation ratio of the dispersion of each batch from the initial mean, based on the discrete incremental records. The absolute values ​​of the factory dispersion of different escalator brake models vary significantly. Directly comparing the absolute values ​​of CV_k across equipment will lead to models with larger initial dispersion being misjudged as severely degraded. Measuring by the proportion of deviation from the initial mean can eliminate baseline differences between models, making the degree of degradation across equipment comparable. The deviation ratio of each batch's CV_k is obtained by subtracting the average CV_k of the same load type at the initial stage of equipment commissioning from the discrete incremental records, and then dividing by the initial mean. This deviation ratio sequence is arranged chronologically by batch to form the main chronological sequence of the braking cycle dispersion. The deviation ratio of batches with increasing trend indicators in the discrete incremental records has a higher weight in the braking cycle dispersion. The increasing trend in the dispersion of batches indicates a stable upward direction and a strong directional signal. This increased weight enhances the sensitivity of the braking cycle dispersion to trend degradation. Discrete incremental records of low-load and high-load concentrated marker batches independently maintain the deviation ratio timing of the two types of sources in the discrete quantity of the braking cycle. The parallel output of the two timing sequences enables the subsequent identification stage to quantify the risk of elastic failure and friction failure according to the source type, preventing mutual interference between the two types of degradation signals and causing misjudgment of the source. When the deviation ratio of each batch of the discrete quantity of the braking cycle exceeds a certain multiple of the initial average, a high deviation marker is marked. When the batch with high deviation markers accounts for more than 50% of the total number of nearly 20 batches, a high deviation dense warning is added. The high deviation dense warning is used for subsequent high discrete batch aggregation judgment. The proportion of high deviation batches of the two concentrated sources constitutes the prior input for fault type weight allocation.

[0026] Step S120: Based on the deviation of the operating state, locate the step cumulative boundary to form the friction section, and use the brake cycle discrete quantity to identify the speed segment with high discrete batch aggregation to determine the brake anomaly source marker.

[0027] Specifically, the step accumulation boundary is located based on the deviation of the operating state to form a friction zone. When wear and debris accumulation on a link of a ladder chain causes the meshing clearance to exceed the critical limit, the friction increases sharply due to jamming with the sprocket. The deviation of the operating state increases abruptly rather than gradually in a short period. This step corresponds to the physical turning point of the drive system from uniform degradation to accelerated degradation, and is the natural starting boundary of the friction zone. When the difference between adjacent batches in the main time sequence of the operating state deviation exceeds a certain proportion of the moving average, it is identified as a step candidate point. When the difference between three consecutive batches exceeds the threshold and is in the same direction, the first batch exceeding the threshold is identified as the starting point of the step accumulation boundary. The range of batches whose operating state deviation continues to be higher than the average of several batches before the starting point is defined as a friction zone. The termination condition is that the average of the operating state deviation continuously falls back to within a certain proportion of the average before the starting point. When an accelerated friction warning exists, the threshold for identifying step candidate points is tightened. After tightening, an accelerated degradation source identifier is added to the boundary of the friction acceleration warning source for the friction zone. A batch with high fluctuations in operating status contributes a maximum of one valid count in the continuous direction consistency determination, preventing fluctuations in operating status of lubrication and maintenance transition batches from being misjudged as the start of a step change. The starting batch number, the number of consecutive batches, and the average operating status deviation within each friction resistance section constitute a section characteristic triplet. When the average within a section exceeds a certain multiple of the historical benchmark, a high friction resistance section is marked. This high friction resistance section mark serves as the primary basis for wear degree grading during the component wear marking generation stage.

[0028] In some embodiments, the step of identifying speed segments with clustered high discrete batches using the brake cycle discrete quantity to determine brake anomaly source markers includes: extracting the time of the over-threshold discrete batch and operating condition parameters from the brake cycle discrete quantity to form a high discrete batch operating condition set; analyzing the running speed characteristics of each batch's braking time to determine the batch speed distribution; statistically analyzing the peak density of high discrete batches and their time drift direction in each speed interval to establish clustered speed intervals based on the batch speed distribution; and generating brake anomaly source markers by centrally determining low-speed segments as elastic failures and high-speed segments as frictional failures based on the clustered speed intervals.

[0029] The discrepancy between the braking cycle and the time of the discrepancy is extracted to form a highly discrete batch operating condition set. The deviation ratio field of the braking cycle discrepancy is scanned batch by batch. Batch numbers whose deviation ratio exceeds a certain multiple of the initial average are extracted and used as search keys to locate all corresponding braking events from the braking time-series data. The four operating condition parameters—braking time, operating speed, passenger flow density, and daily temperature—are combined with the batch number and written into the highly discrete batch operating condition set. The introduction of the temperature parameter is used to distinguish the normal duration fluctuation of the braking spring caused by thermal expansion and contraction in low-temperature environments. When escalators are installed in outdoor open environments, the elastic modulus of the springs in winter low-temperature batches is about 3% to 8% lower than that in summer. The slight extension of braking duration caused by low temperature and the extension caused by elastic degradation are highly similar in amplitude and cannot be distinguished by duration alone. When the daily temperature is more than two standard deviations below the local historical monthly average temperature, a low-temperature batch identifier is added. The threshold for judging the low-temperature batch identifier is relaxed during the batch speed distribution analysis stage to prevent small changes in elastic modulus caused by low temperature from being mistakenly classified as elastic failure. Batches with increasing discrepancies in braking cycles are assigned trend source identifiers within the high-discrepancy batch set. This increasing trend indicates that the high discrepancy in this batch has a continuous evolutionary characteristic rather than a sporadic disturbance. Batches with trend source identifiers receive increased weight during the aggregation speed range identification stage, allowing the trend-degrading signal to contribute more significantly to the density peak calculation. Batches with low-load and high-load discrepancies in braking cycles are assigned corresponding source identifiers within the high-discrepancy batch set, respectively. These two types of source identifiers trigger priority searches for low-speed and high-speed segments during the batch speed distribution analysis stage, ensuring that differences in operating condition sources are addressed in subsequent aggregation determination.

[0030] For example, the step of analyzing the braking speed characteristics of each batch in the highly discrete batch operating condition set to determine the batch speed distribution includes: extracting the braking start time and operating condition characteristics of each batch in the highly discrete batch operating condition set to form a braking start operating condition set; using the braking start operating condition set to analyze the corresponding time operating speed value to obtain the braking speed value; based on the braking speed value, statistically analyzing the braking frequency time period cluster distribution in each preset speed interval to determine the speed interval frequency quantity; and based on the speed interval frequency quantity, aggregating the frequency ratio of each interval and the abnormal missing speed interval to determine the batch speed distribution.

[0031] For highly discrete batch operating condition sets, the braking start time and operating condition characteristics of each brake initiation within each batch are extracted to form a braking start operating condition set. All braking events are extracted batch by batch from the highly discrete batch operating condition set according to batch number. The start time of each braking event is combined with four operating condition characteristics: operating speed, passenger flow density, ambient temperature, and trigger cause code, and then written into the braking start operating condition set. When extracting low-temperature batch identifiers from the braking start operating condition set, their validity must be verified by comparing them with the historical average temperature of the current month. After verification, the low-temperature identifier is transferred to the corresponding item to ensure complete coverage of subsequent temperature corrections without omitting batches crossing month boundaries. Supplementary identifiers are extracted from adjacent batches. The consistency of the operating time between adjacent batches and the original batch must be verified. The operating time is estimated using the median of the hourly distribution of braking event trigger times within each batch. When the median difference exceeds 2 hours, a time period deviation identifier is added to the supplementary source. The weight of the time period deviation identifier is further reduced during the speed value analysis stage at the braking moment. After verifying the consistency between the number of records in each batch of the braking start condition set and the number of braking events in the corresponding batch of the high discrete batch condition set, missing records trigger an event missing flag. For batches with missing event flags, the frequency denominator in the speed interval frequency statistics stage is determined by the actual number of valid records rather than the rated number of braking events, to prevent falsely low frequency density due to missing records. If the pass rate for the completeness verification of the braking start condition set is lower than 85% of the entire batch, an additional batch completeness warning is issued. This warning triggers a slight relaxation of the candidate density threshold in the subsequent speed interval identification stage to compensate for the density underestimation caused by insufficient samples.

[0032] The braking moment speed value is obtained by parsing the running speed value at the corresponding moment in the braking start condition set. Elastic failure is most exposed in low-speed braking, while friction failure is more prominent in high-speed braking. Accurately extracting the running speed at each braking moment is a prerequisite for distinguishing the two types of failure sources. The variable frequency controller speed feedback timing uses the start time of each event in the braking start condition set as the query key. The timing resolution is usually 20ms. The speed extraction uses linear interpolation instead of nearest neighbor value. The interpolation accuracy must be no less than 1% of the rated speed. The extraction result constitutes the main body of the braking moment speed value. The escalator goes through a low-speed running section and then stops during the passenger flow clearing stage at the end of the operating day. The speed of the deceleration section before stopping is significantly different from that of the constant speed section. If the stopping moment speed is used directly, the low-speed section will be grouped with high frequency triggers. The trigger reason code for the passenger flow clearing and stopping event is replaced by the constant speed section speed at the deceleration start moment in the braking start condition set. The deceleration start moment is determined by the first moment when the speed difference in the speed feedback timing changes from positive to negative. The braking moment speed value is marked with a clearing deceleration source identifier for this type of record. The low-temperature batch label must be temperature-corrected to unify the speed values ​​between low-temperature and normal-temperature batches. The correction factor is determined by statistical analysis of historical low-temperature batch speed deviations for the equipment, and the correction amount is typically between 1% and 3% of the rated speed. After correction, the braking speed values ​​of low-temperature batches can be directly compared with those of normal-temperature batches. The confidence level of the braking speed value corresponding to the time-period deviation label is lowered, and the weight of the confidence level is reduced during the frequency statistics stage of the speed interval. When both time-period deviation and supplementary label exist simultaneously, the two reductions are multiplied and added together, and the lower limit of the weight after the addition is not less than 0.6 times.

[0033] The frequency of each speed interval is determined by statistically analyzing the braking frequency distribution within each preset speed interval based on the braking speed values ​​at the braking moment. All braking speed values ​​are recorded and grouped into three intervals: low speed (0 to 60% of rated speed), medium speed (60% to 80%), and high speed (80% to 100%). Within each interval, braking frequency is statistically analyzed for four time periods: morning peak, daytime, evening peak, and nighttime, forming a frequency matrix for the three intervals and four time periods. The frequency of each speed interval is organized using a triple index of batch number, speed interval, and time period identifier. During morning and evening peak hours, when passenger density is high and there are more low-speed stops, the residual frequency is calculated after deducting the time period baseline frequency. The time period baseline frequency is the average number of braking events in the same historical batches during the same period. The average calculation window is the batches during the first three months of equipment operation during the same period. A positive residual indicates that the braking frequency in that speed interval during that time period is higher than the baseline. The residual value constitutes the effective statistical quantity of the speed interval frequency. The speed value at the braking moment and the speed of the elevator deceleration source identifier have been replaced with the speed of the constant speed segment. A speed replacement identifier is added to the corresponding item, and the weight of the speed replacement identifier item remains normal. When the proportion of elevator replacement records exceeds 30%, a high replacement ratio identifier is marked. The reliability of the residuals of batches with high replacement ratios is relatively reduced. This ratio value is recorded in the confidence field during the batch speed distribution summary stage. A speed interval in which the residual frequency is positive for three consecutive statistical batches is marked with a continuously high identifier. The continuously high identifier triggers the determination of the frequency proportion of this interval with the number of valid records as the denominator during the batch speed distribution generation stage, ensuring that the frequency density estimation is not biased by the differences between cross-batch samples.

[0034] The batch speed distribution is determined by aggregating the frequency proportion of each speed interval and identifying abnormally missing speed intervals. The residual frequencies of the three speed intervals are each normalized to a frequency proportion Q_i, calculated as follows: Q_i = ΔF_i / Σ(ΔF_j), where ΔF_i is the residual frequency of the i-th speed interval, ΔF_i = F_i - F_base,i, F_i is the measured frequency of this interval in the current batch, F_base,i is the historical baseline frequency for the same period, and Σ(ΔF_j) is the sum of the residual frequencies of the three intervals (j is the summation sequence number of the speed intervals, traversing the low, medium, and high intervals). Q_i reflects the proportion of each speed interval exceeding the baseline. The proportion statistics only include intervals where ΔF_i is greater than zero, and the lower limit of the proportion of intervals where ΔF_i is not greater than zero is zero. The Q_i sequence of the three intervals constitutes the main body of the batch speed distribution. When Σ(ΔF_j) is zero, it is marked as a low-frequency indicator for the entire interval. Batches with low frequencies in the entire interval do not participate in the density statistics of the aggregated speed intervals. Historically, batches with a normal frequency (Q_i) of at least 10% are marked with a missing interval where the current batch's ΔF_i is zero or extremely low. When the escalator control logic detects a risk of elastic failure, it may proactively avoid triggering braking in the low-speed range, resulting in a frequency gap in the low-speed range while the frequency in the high-speed range is excessively high. During the speed range clustering identification phase, the missing interval triggers a joint analysis of the density peak values ​​of its adjacent intervals to ensure that the actively avoided gap does not obscure the true source of clustering. For intervals with a persistently high frequency, the weight of the Q_i interval is increased during batch speed distribution output. The increase is determined by the ratio of the number of batches with persistently high frequencies in that interval to the total number of batches in the statistical window. When the standard deviation of Q_i in each speed interval continuously decreases across batches, a distribution rigidity flag is added to the batch speed distribution. Distribution rigidity is a macroscopic signal of degraded braking system speed characteristics. The distribution rigidity flag triggers a shortening of the candidate decision window during the speed range clustering identification phase to accelerate peak recognition response.

[0035] Clustering speed intervals are established by statistically analyzing the peak density of highly discrete batches in each speed interval and their time drift direction. Within a sliding time window of the batch speed distribution, the proportion of highly discrete batches in each speed interval to the total number of batches in that window is used as the interval density D_i. The peak density of the interval is calculated using the following formula: D_i = N_high,i / N_window, where N_high,i is the number of highly discrete batches in the i-th speed interval within the window, and N_window is the total number of batches in the window. When D_i exceeds a certain multiple of the average density of the entire speed interval, it is confirmed as a candidate clustering speed interval. The time drift direction of the candidate interval is determined by the direction in which the batch number corresponding to the density peak moves with the monthly window. Drifting towards the latest batch indicates that the clustering period is expanding and the failure process is not being contained. The interval drift indicator triggers a priority search for the density peak in the drift target interval. When the peak value of D_i is consistent with the drift direction, a drift consistency indicator is added to the clustering speed interval. Drift consistency indicates that the clustering degree in that interval is intensifying, and the braking anomaly source marker increases the confidence in determining the failure type of the drift consistency source interval. When discrete centralized early warnings exist, the threshold for determining density peak values ​​has been tightened. For clustering speed ranges, precise determination labels are added to batches originating from discrete centralized early warnings. These precise determination label ranges do not apply the relaxed low-speed / high-speed classification boundaries during the elastic and frictional failure classification stages. The batch speed distribution supplementary label batches have had their weights reduced during density statistics; D_i uses the reduced effective batch number as the denominator to prevent a systematically high density peak value when the proportion of low-weight batches is too high. When clustering is triggered only in the medium-speed segment, the medium-speed segment clustering label is not included in either of the two failure classifications. Braking anomaly source markers add unclassified labels to the medium-speed segment clustering sources. The unclassified label range triggers supplementary diagnosis during the associated fault mode identification stage.

[0036] Based on the convergence speed range, low-speed segments are concentrated and classified as elastic failures, while high-speed segments are concentrated and classified as friction failures, generating braking anomaly source markers. When the density peak value of the candidate interval in the low-speed segment of the convergence speed range exceeds the judgment threshold, it is confirmed as an elastic failure source. After the elasticity of the brake spring fails, the initial compression in the low-speed segment is insufficient to provide the rated braking torque. In the early stage of failure, it only affects the consistency of low-speed braking event duration, while high-speed braking is not sensitive due to inertia. The braking anomaly source marker adds a failure initial speed range field to the elastic failure type to record the speed range in which the failure first appears. When the density peak value of the candidate interval in the high-speed segment of the convergence speed range exceeds the judgment threshold, it is confirmed as a friction failure source. After the friction pads wear, the contact area decreases, resulting in the largest braking torque gap in high-speed braking. The braking anomaly source marker adds a severity rating field to the friction failure type. The rating is determined by the ratio of the difference between the high-speed segment density peak value and the judgment threshold normalized to the historical maximum excess. The drift consistency marker adds a drift aggravation marker to the braking anomaly source marker. The drift aggravation marker triggers an earlier maintenance intervention time limit during the fault urgency rating stage. The sources of clustering identified through joint analysis within the missing identifier intervals are distinguished by joint judgment identifiers in the braking anomaly source marking. The confidence level of joint judgment sources is lower than that of direct clustering sources, and their weight is reduced during the associated fault mode identification stage. When elastic failure and friction failure are triggered simultaneously, a composite failure identifier is added. In the associated fault mode identification stage, both types of related components are included in the retrieval scope. The ratio of the peak density of each failure type constitutes the relative severity index of the two types of failures, serving as a priori basis for weight allocation in the associated fault mode identification stage.

[0037] Step S130: Compare the state differences of each speed segment in the friction section to generate component wear marks. Based on the component wear marks, identify the brake abnormality source and mark the related components to form an associated fault mode. Based on the associated fault mode, identify the decline inflection point of power consumption recovery rate of previous similar repairs to determine the fault urgency rating.

[0038] Specifically, wear markers are generated by comparing the state differences of different speed segments in the friction resistance section. The friction increase caused by the wear of the escalator drive chain rises approximately synchronously and is generally evenly distributed across all speed segments. The friction increase caused by localized wear of the guide rail is more concentrated in the high-speed segment due to the amplification effect of contact force. The difference between the state of the friction resistance section at high speed and low speed segments and the deviation from the mean is defined as the speed difference state. When the difference state exceeds a certain proportion of the friction resistance section mean, it is judged as high-speed concentrated wear and a guide rail wear source marker is added. When the difference state is close to zero or uniformly high across all speed segments, it is judged as uniformly distributed wear and a chain wear source marker is added. The high friction resistance section is identified as the source of the high friction resistance section with a large amplitude of speed difference state. The component wear marker adds a high wear degree marker to the high friction resistance source. The high wear degree marker prioritizes matching historical maintenance samples of equal or greater severity during the associated fault mode identification stage to avoid the mild maintenance samples lowering the average recovery rate and underestimating the current degradation degree. The wear rating of the component wear marker corresponding to the batch of accelerated degradation source identifier in the friction section is upgraded by one level. The friction slope of the accelerated degradation batch is higher than that of the steady-state batch, and the upgrade makes the component wear marker more sensitive to the current degradation rate. When the component wear marker guide rail wear source identifier and chain wear source identifier are triggered simultaneously in the same section, a composite wear identifier is added. Composite wear indicates that multiple contact pairs in the ladder system are degrading synchronously. In the associated fault mode identification stage, the historical sample search range for composite wear sources is expanded to include records of joint maintenance of multiple components.

[0039] The wear indicators on components identify the sources of brake anomalies, forming a associated fault mode for related components. The chain wear indicator and the friction failure indicator on the brake anomaly are both driven by the same cause: continuous cyclic load. Their simultaneous occurrence in the same batch indicates that the drive chain and brake friction pads are in a state of synchronous degradation. The associated fault mode adds a synchronous degradation indicator to this combination. Under synchronous degradation conditions, the accelerated deterioration of the two types of components is superimposed. Maintenance intervention must cover both types of components simultaneously; otherwise, repairing one component will lead to the continued degradation of the other, causing recurrence. The guide rail wear indicator and the elastic failure indicator on the brake anomaly are physically independent, but the vibration and impact caused by localized wear on the guide rail act periodically on the brake spring at a rate of approximately 0.2 to 0.5g (g is the acceleration due to gravity), accelerating the relaxation of the spring preload. Their simultaneous occurrence in the same batch indicates a causal relationship. The associated fault mode adds a causal association indicator to this combination. The historical sample retrieval scope triggered by the causal association indicator is expanded to include composite maintenance records containing guide rail repairs, ensuring that the identification of the inflection point of recovery rate decline is not inaccurate due to insufficient single-component maintenance records. The component wear marker composite wear identifier batch-related failure mode incorporates historical maintenance data related to both chain and guide rail components. The recovery rate degradation sequence of each component's independent maintenance is used, and the component with the greater degradation among the two types of components is used to determine the final failure urgency rating. This prevents the high recovery rate of mildly degraded components from masking the true risk of deeply degraded components. The brake anomaly source marker drift aggravation identifier batch-related failure mode historical sample retrieval period is tightened to the last 24 months. Increased drift indicates an accelerating failure rate. Early maintenance samples have higher recovery rates than recent batches and are less representative. Tightening the retrieval period makes the recovery rate mean estimate closer to the current equipment condition.

[0040] In some embodiments, determining the fault urgency rating based on the decline inflection point of the power consumption recovery rate of previous similar repairs identified by the associated fault mode includes: using the associated fault mode to retrieve records of the deviation of operating status before and after each repair of the same model and the same fault to obtain the status records before and after repair; calculating the baseline correction deviation recovery ratio based on the status records before and after repair to determine the single status recovery rate; identifying the inflection point of the continuous repair recovery rate from uniform decline to accelerated decline based on the single status recovery rate to determine the amount of recovery rate decline; and forming a fault urgency rating based on the repair intervention time limit after the recovery rate decline amount is collected into the accelerated decline section.

[0041] The system retrieves operational status records before and after maintenance for the same model and fault using associated fault patterns. The fault type and equipment model fields are combined as the search key. All maintenance events of the same model and fault are extracted from the historical maintenance database. Braking component failure is reflected in operational status characteristic values ​​due to residual drag from incomplete brake release. Therefore, the average operational status deviation of the last batch before maintenance and the average operational status deviation of the first batch after maintenance constitute a pre- and post-maintenance status record for each maintenance event. This record can characterize both the degradation and recovery of the braking components. The maintenance event number is used as the primary key to store the average deviation values ​​and the corresponding batch number. For records with a joint maintenance source identifier retrieved from the batch-based causal association identification of associated fault patterns, the average post-maintenance status deviation must be separated from the guide rail maintenance contribution, retaining only the brake component recovery amount. The separated amount is estimated from the historical average status change of individual maintenance events for similar guide rails. The guide rail maintenance contribution typically accounts for 15% to 30% of the total recovery amount from joint maintenance. The separated result is appended with the source identifier. When the sample size of pre- and post-maintenance status records is less than the minimum inflection point identification threshold, it is supplemented with cross-model samples of the same type of fault. The cross-model supplementation source is marked with a cross-model identifier. The weight of cross-model identified samples is reduced to 0.8 times in the single state recovery rate calculation stage. The upper limit of the number of cross-model samples is twice the sample size of the local machine to prevent cross-model differences from skewing the average recovery rate. When the average deviation of the pre-maintenance status of each maintenance in the pre-maintenance status records is more than twice the historical average of the entire batch, an abnormally high pre-maintenance status identifier is marked. An abnormally high pre-maintenance status indicates that the equipment was in a severely degraded state before the maintenance. This identifier triggers a correction in the deviation recovery ratio calculation stage to compensate for the systematic suppression of the recovery ratio by the high pre-maintenance status.

[0042] The single-time state recovery rate is determined based on the deviation recovery ratio calculated from the pre- and post-maintenance state records. As the cumulative travel of the escalator step chain exceeds its design life, the adhesion efficiency of each lubrication replenishment decreases. After replacing the friction plates, the break-in period of the new and old mating surfaces is extended with each replacement. Therefore, the recovery effect of maintenance on friction exhibits a gradually diminishing trend. The deviation recovery ratio R_m is calculated by dividing the difference between the pre-maintenance and post-maintenance state deviation by the difference between the mean pre-maintenance state deviation and the healthy baseline. The trend of R_m value with the number of maintenance directly depicts the evolution trajectory of component repairability. A deviation recovery ratio R_m close to 1.0 for each maintenance in the pre- and post-maintenance state records indicates that the maintenance has completely restored the state deviation to the healthy baseline. An R_m exceeding 1.0 indicates overcompensation, ending at 1.0. A negative R_m value indicates an increased state deviation after maintenance. Negative values ​​are not included in the mean calculation but are retained for maintenance quality traceability. The mean deviation of the post-maintenance state from the combined maintenance source identifier item in the pre- and post-maintenance state records has been stripped of the guide rail contribution. R_m reflects the recovery effect of the brake component alone. Without stripping, R_m is approximately 0.10 to 0.20 higher. After stripping, the estimation accuracy is significantly improved. The single-state recovery rate is represented by the deviation of R_m minus the weighted mean R_mean of historical R_m for the same model and fault (weighting rules are described below). The deviation is arranged chronologically according to the maintenance events. A continuously negative deviation with an increasing absolute value is a direct signal that the recovery rate is entering a decline channel. A negative deviation in the single-state recovery rate indicates that the maintenance effect is lower than the historical average. The negative amplitude of the deviation gradually increases from near zero, which is a direct manifestation of the decline in repairability in the maintenance records.

[0043] For example, determining the single-time state recovery rate based on the baseline correction deviation recovery ratio calculated from the pre- and post-maintenance state records includes: obtaining the pre-maintenance state baseline by statistically analyzing the average deviation of the last batch of operating states before maintenance using the pre- and post-maintenance state records; obtaining the post-maintenance state deviation average by statistically analyzing the average deviation of the first batch of operating states after maintenance based on the pre- and post-maintenance state records; calculating the deviation recovery ratio after removing overcompensation based on the pre-maintenance state baseline and the post-maintenance state deviation average; and determining the single-time state recovery rate by aggregating the deviation recovery ratio against the average historical recovery rate of this model and this fault.

[0044] The pre-maintenance status baseline is obtained by statistically analyzing the deviation of the last batch of operating status before maintenance using the status records before and after maintenance. The average deviation of the operating status of the last 5 batches before maintenance for each maintenance event in the status records before and after maintenance constitutes the pre-maintenance status baseline P_pre,m for that maintenance. The last 5 batches represent the stable friction level of the equipment before maintenance. When the number of batches is too small, the impact of a single high-load disturbance on the average is too large. When the number of batches is too large, the early low-friction batches lower the average and underestimate the degree of degradation before maintenance. The escalator continues to operate after the maintenance notice is issued and before the actual shutdown for maintenance. Friction may further accumulate during the waiting period. When the interval between the last batch and the maintenance time exceeds 3 operating days, a "large time interval" indicator is marked. The pre-maintenance status baseline for the "large time interval" item is changed to the average of the last 7 batches. The 7-batch window includes the friction fluctuation during the waiting period in the baseline to avoid the occasional peak of the last batch dominating the P_pre,m estimation. The pre-maintenance baseline environment correction has been completed in the operational status deviation stage. Here, the corrected environmental value is directly called. When there is a large environmental fluctuation between batches at the time of maintenance, the difference in the environmental correction range within the last 5 batches must be verified. When there is a large batch combination with environmental fluctuation, an environmental fluctuation sampling label is added. The confidence level of the environmental fluctuation sampling label item is lowered in the deviation recovery ratio calculation stage.

[0045] The mean deviation of the post-maintenance state is obtained by statistically analyzing the deviation of the first batch of operating states after maintenance based on the status records before and after maintenance. The mean post-maintenance state P_post,m is calculated by taking the mean deviation of the first three batches of operating states after each maintenance event. The first batch of states reflects the immediate effect of the maintenance. Taking the mean of three batches smooths out the temporary low value introduced before the lubricating oil film has stabilized. If there are too many batches, the state will naturally recover during the break-in period, which will systematically raise the mean and overestimate the maintenance effect. After escalator maintenance, a lubricating oil film reconstruction phase is required. It usually takes about half an operating day for the new oil film to fully adhere. When the lubricating oil film is not fully established, the initial low state is a normal process phenomenon and not a true sign of insufficient recovery. When the initial state is more than 10% lower than the second batch, an oil film establishment delay indicator is added. The mean post-maintenance state for the oil film establishment delay indicator item is changed to the mean of the second to fourth batches to skip the oil film establishment transition period, ensuring that P_post,m reflects the true recovery level of the braking components and not a low value during the process transition period. In the pre- and post-maintenance status records, if the interval between the first batch of samples after maintenance and the maintenance time exceeds two operating days, a recovery delay indicator is marked. This indicates a situation where the installation personnel did not complete the lubrication and sealing process according to procedures, resulting in an inflated initial condition and subsequent decline in subsequent batches. To address the source of the recovery delay, the lowest value among the first three batches is used instead of the average value for the post-maintenance status average, as the lowest value is closer to the true recovery level after the maintenance is fully effective. If the post-maintenance status average is lower than the historical health benchmark, an overcompensation indicator is displayed. Overcompensation typically occurs when the friction coefficient of a new component is lower than that of the original component during installation, or when the initial operating condition is lower than the health benchmark. The overcompensation indicator item is set to 1.0 during the deviation from the recovery ratio calculation phase to eliminate the effect of overcompensation.

[0046] The deviation recovery ratio, excluding overcompensated components, is calculated based on the average deviation between the pre-maintenance baseline and the post-maintenance baseline. The difference between the pre-maintenance baseline and the average post-maintenance baseline deviation is the deviation recovery amount for this maintenance. The recovery amount is divided by the difference between the pre-maintenance baseline and the historical health baseline to obtain the deviation recovery ratio R_m, calculated as follows: R_m=(P_pre,m-P_post,m) / (P_pre,m-P_base), where P_pre,m is the average deviation of the last batch of operating status before the m-th maintenance, P_post,m is the average deviation of the first batch of operating status after maintenance, and P_base is the average deviation of the historical health baseline status of the equipment (P_base=0 according to the definition of operating status deviation). R_m=1.0 corresponds to complete recovery of the status deviation to the health baseline, and R_m=0.5 corresponds to only half of the deviation being recovered in this maintenance. The magnitude of the continuous decrease in R_m with the increase of maintenance frequency reflects the gradual decline in the repairability of the component. If the pre-maintenance baseline is abnormally high, the pre-maintenance indicator P_pre,m will be biased, leading to an increased denominator and a systematically low R_m. A correction factor, estimated by the ratio of the historical normal degradation maintenance P_pre,m to the mean P_base, is introduced during calculation. After correction, the R_m estimation deviation narrows from approximately 0.12 to approximately 0.05. If the overcompensation indicator R_m is cut off at 1.0, the post-maintenance deviation will actually deviate further from the baseline, typically corresponding to installation errors or incorrect component models. Negative values ​​are not included in the calculation of the single-time recovery rate mean but are recorded for maintenance quality traceability. If the pre-maintenance baseline time interval is too large, the indicator P_pre,m has already taken the extended window mean. When the difference between the extended mean and the mean of the last 5 batches exceeds a certain proportion of the baseline value, a baseline fluctuation indicator is added to the deviation recovery ratio. The weight of the baseline fluctuation indicator is reduced during the single-time recovery rate calculation stage. The reduction in weight is proportional to the baseline fluctuation ratio; the larger the fluctuation, the lower the representativeness and the smaller the weight. The sampling identifier item R_m of the pre-maintenance state baseline environmental fluctuation has environmental deviation. The environmental deviation identifier is added to the sampling source of the deviation recovery comparison environmental fluctuation. The environmental deviation identifier item is compared with the mean value by grouping according to environmental conditions in the single state recovery rate summary stage to eliminate the interference of environmental differences on cross-maintenance event comparison.

[0047] The single-time recovery rate is determined by the deviation from the historical average recovery rate of the same model and fault based on the deviation recovery ratio. The absolute value of the deviation recovery ratio is affected by the initial severity of the fault and batch differences in the maintenance process. Direct comparison across maintenance events would include these two types of noise in the trend judgment. Therefore, the deviation relative to the historical weighted average of the same model and fault is used as the aggregation object. After removing the fault severity and process fluctuations, the trend of repairability changing with the number of maintenance is retained separately. The recovery rate deviation of the maintenance is obtained by subtracting the weighted average of the historical R_m of the same model and fault from each effective item R_m of the deviation recovery ratio. When calculating the weighted average, the weight of the cross-model identifier source is reduced to 0.8 times, and the weight of the benchmark fluctuation identifier source is reduced according to the fluctuation ratio. The deviation is arranged in the time sequence of maintenance events to form the main time sequence of the single-time recovery rate. The deviation amount corresponding to the environmental deviation indicator item of the recovery ratio is compared with the historical average of similar environmental conditions. After environmental grouping, the sample size of each group must not be lower than the minimum inflection point identification threshold. For environmental groups with insufficient sample size, the average of the entire batch is substituted and an insufficient group indicator is added. The threshold for the differential sequence judgment item of the insufficient group indicator item is appropriately relaxed in the decline inflection point identification stage to compensate for the differential estimation error introduced by insufficient sample. The source of the oil film establishment delay indicator P_post,m of the single state recovery rate is the average of the second to fourth batches. The R_m of this type of maintenance is usually higher than that of the no-delay item by about 0.05 to 0.10. Before the deviation amount is calculated, this systematically high amount must be deducted from R_m to ensure that the deviation amount of each maintenance is consistent. The deduction amount is determined by the statistical difference of the average R_m of the same type of maintenance with and without delay in history.

[0048] The recovery rate degradation is determined by identifying the inflection point where the continuous maintenance recovery rate transitions from a uniform decline to an accelerated decline based on the single-state recovery rate. The single-state recovery rate deviation time series captures the repairability degradation signal—how much can be recovered with each maintenance—rather than the state-exceeding signal—whether the current friction exceeds the threshold. The repairability degradation signal can be observed when the recovery rate begins to decline, typically triggering an alarm weeks to months earlier than the detected value. The difference between the deviations of adjacent maintenance events in the single-state recovery rate deviation time series forms a difference sequence. When the difference sequence changes from small fluctuations close to zero to a sustained negative value with a monotonically increasing absolute value, it is determined to be an accelerated decline phase. The first maintenance event number that meets this condition is identified as the degradation inflection point. The absolute value of the difference between the average deviations of three maintenance events before and after the inflection point is defined as the recovery rate degradation. The inflection point signifies that the maintenance repair effect can no longer offset the cumulative rate of structural degradation, and the component transitions from a slow, repairable degradation phase to an accelerated deterioration phase. When the single-state recovery rate deviation is negative for three consecutive times and its absolute value increases monotonically, the corresponding item is marked with a decline channel entry indicator. The decline channel entry indicator triggers a tightening of the threshold for judging accelerated decline in the differential sequence, ensuring that the inflection point is captured in the early stage of the decline channel. The cross-model indicator sample differential sequence of recovery rate decline participates in inflection point identification with a weight of 0.8. The inflection point identification results of cross-model and local samples are weighted and merged. When the proportion of cross-model samples exceeds 50%, the recovery rate decline is marked with a cross-model dominant indicator. The conservative intervention time limit of the cross-model dominant batch is mainly referenced with the local sample to prevent the time limit estimation bias introduced by model differences from affecting the intervention priority ranking.

[0049] A fault urgency rating is established based on the maintenance intervention time limit after the recovery rate decline enters the accelerated decline phase. The actual time interval from the inflection point of historical similar faults to the next fault trigger or forced shutdown after the inflection point of the recovery rate decline constitutes the intervention time limit sample set. The 10th percentile is taken as the conservative intervention time limit. When the ratio of the time elapsed since the last maintenance to the conservative intervention time limit exceeds 0.6, the fault urgency rating enters Level 2 warning; when it exceeds 0.85, it enters Level 3 urgency. The 10th percentile corresponds to 10% of equipment having experienced forced shutdown within this time limit. Compared to taking the mean, the conservative estimate advances the intervention window and reserves a buffer for on-site dispatch, which is particularly important for escalators in public transportation facilities with uninterrupted operation. The conservative intervention time limit for cross-model recovery rate decline is determined by weighting the 10th percentile of the current model as the primary indicator and the 10th percentile of the cross-model model as a secondary indicator. When the difference between the two exceeds 30% of the average intervention time limit, an additional model difference warning is issued. The fault item corresponding to the model difference warning is treated with a higher level of conservatism to cover the risk of underestimating the time limit introduced by model differences. When the recovery rate decline exceeds a certain percentage of the historical average for similar faults, the fault urgency rating will be forcibly raised by one level based on the three-level mapping result, but not exceeding three levels. The upward movement reflects that the current batch's decline has exceeded the historical typical level. The deeper the degree to which the repair effect is consistently below the average, the lower the repairability of the components. The associated fault mode synchronous degradation identifier source fault urgency rating will be simultaneously raised by one level. If it is already at level three urgency, it will remain unchanged and will not be superimposed.

[0050] Step S140: Based on the fault urgency rating, analyze the critical time period of each fault combination to obtain the critical margin set. Use the critical margin set to determine the nearest maintenance time period of the item with the smallest margin to obtain the urgent maintenance window. From the urgent maintenance window, identify the lowest level maintenance type that can restore friction and establish a maintenance task plan.

[0051] Specifically, the critical margin set is obtained by analyzing the critical time period of each fault combination based on the fault urgency rating. The remaining available time T_margin for each fault item is calculated by the following formula: T_margin,f = T_conservative,f - T_elapsed,f, where T_conservative,f is the conservative intervention time limit (days) corresponding to the f-th fault item, T_elapsed,f is the time elapsed since the last maintenance (days), and T_margin,f is the remaining available time (days). When T_margin,f is negative, it indicates that the fault has exceeded the conservative intervention time limit. The critical margin set is arranged in ascending order of T_margin,f for each fault item, with negative items listed first and processed first. When escalators exhibit both chain wear and elasticity failure as degradation types, the conservative intervention timeframes for these two types of faults are often asynchronous. The chain wear source in the fault urgency rating may have already reached level three urgency while the elasticity failure source remains at level two warning. Each fault item must be independently sorted according to T_margin,f. Mixed sorting results in higher margins for mild degradation faults, which inflate the ranking of severe degradation faults, leading to a delay in the most urgent faults. The synchronous degradation indicator T_margin,f in the fault urgency rating is determined by the smaller value between the two types of components. Under synchronous degradation conditions, the criticality of any component triggers a joint intervention requirement. The critical margin set adds a joint criticality indicator to the synchronous degradation source. During the urgent maintenance window determination phase, the joint criticality indicator requires that the downtime cover the synchronous processing time for both types of components. The model difference warning source in the fault urgency rating has its T_margin,f further compressed by 10%. After compression, the margin is more conservative, and the intervention ranking shifts forward. The critical margin set adds a conservative compression indicator to the model difference warning source. The conservative compression amount corresponds to the risk of timeframe estimation bias introduced by model differences. When the median of all fault entries T_margin,f in the critical margin set falls below a certain threshold, an overall critical warning is added. An overall critical warning indicates that multiple types of faults are entering the intervention window at the same time. Urgent maintenance windows must be scheduled for centralized processing. Centralized scheduling allows for shared downtime periods to avoid repeated occupation of operational shifts by frequent planned downtime.

[0052] The most recent maintenance period for the item with the smallest margin is determined using the critical margin set to obtain the urgent maintenance window. The fault item with the smallest margin corresponds to the most urgent fault. The latest intervention time is obtained by adding 0.85 times the T_margin,f of the fault item to the current time. Before this time, the earliest downtime in the operation plan with a duration not less than the estimated maintenance time is retrieved as the maintenance window. If the duration is insufficient, it is postponed to the next downtime that meets the conditions. The urgent maintenance window is stored with the start and end times of the final downtime as the primary key. The downtime of the critical margin set combined with the critical source must cover the sum of the maintenance time of the two types of components. If the required tools and operation positions do not conflict, the two types of operations are carried out in parallel; otherwise, they are arranged sequentially, and a sufficiently long downtime is retrieved again. The urgent maintenance window adds a parallel operation identifier to the joint critical source. When the critical margin set has an overall critical warning, a single downtime window that can accommodate all urgent fault handling time is prioritized. If a single downtime window is insufficient, two downtime windows are arranged and a split execution identifier is added for batch arrangement of operations during the maintenance task plan generation stage. The latest intervention time of the source is conservatively compressed based on the compressed value and is not subject to callback. Once the urgent maintenance window period is determined, a conflict detection is performed with the scheduled high-passenger-flow operation plan. If there is a conflict, the search continues to non-conflict periods. The interval between the final downtime period and the latest intervention time for each fault constitutes the maintenance execution margin. If the margin is less than the interval between two downtime periods, an urgent scheduling warning is added, and conservative man-hours are reserved during the maintenance task plan generation stage.

[0053] In some embodiments, the step of identifying the lowest-level repair type capable of restoring friction resistance from the urgent repair window and establishing a repair task plan includes: obtaining the friction resistance deviation by analyzing the abnormal and sudden increase in friction resistance of the corresponding component when triggered by the urgent repair window; using the friction resistance deviation to match the stuck recovery type that reaches the lowest intervention level to form a baseline recovery type; adjusting the reserved time according to the baseline recovery type based on the historical probability of overestimating the same type of repair time to determine the repair time estimate; and constructing a repair task plan by collecting the repair type and resource requirement list from the repair time estimate.

[0054] The friction deviation is obtained by analyzing the magnitude of the sudden increase in friction after an abnormal stabilization at the trigger of the emergency maintenance window. In the time series of deviations in the operating status of several batches prior to the trigger of the emergency maintenance window, if the deviation, which should have gradually increased with wear, did not increase for more than three consecutive batches, and then suddenly increased by a certain percentage of the average value of the stabilization period in a single batch, it is identified as an abnormal stabilization followed by a sudden increase. The difference between the sudden increase and the corresponding component's health benchmark is defined as the current friction deviation of that component. When a worn section of an escalator step chain is temporarily sealed by a lubricating oil film, the friction of the corresponding link remains stable. However, due to temperature changes or stress causing the oil film to peel off, the worn section directly contacts the lubricating part, causing a sudden increase in friction. This type of sudden increase is often larger and develops faster than the deviation caused by gradual wear. Simply relying on gradual deviation to estimate the friction deviation will underestimate the current severity of degradation. When an urgent maintenance window is triggered without a clear abnormal stable mode, the friction deviation is directly taken as the difference between the average deviation of the corresponding operating state at the trigger time and the healthy baseline. The deviation of progressive wear is usually lower than that of the sudden increase mode after abnormal stability. The two types of sources are distinguished by the sudden increase source indicator and the progressive source indicator. When the sudden increase source indicator exceeds twice the average of the stable segment, a deep sudden increase indicator is marked. The component corresponding to the deep sudden increase has entered the rapid failure stage. During the baseline recovery type matching stage, the lubrication and maintenance level is directly excluded for deep sudden increase sources. When the deviation of the progressive source indicator is below a certain threshold, a low deviation indicator is marked. Low deviation indicates that the current friction degradation is still in the reversible stage. During the baseline recovery type matching stage, the lubrication and maintenance level is preferentially matched for low deviation sources to minimize maintenance resource consumption.

[0055] A baseline recovery type is established by matching the friction deviation to the minimum intervention level of the jamming recovery type. The jamming recovery type refers to the lowest level of intervention that can be applied when the moving parts are prone to jamming due to excessive resistance caused by wear or lubrication failure. The maximum recoverable friction deviation range for the three types—lubrication replenishment, clearance adjustment, and local cleaning—is determined by the 25th percentile of the deviation recovery amplitude after similar historical maintenance. The 25th percentile, rather than the mean, is used to cover cases of low recovery effectiveness and prevent insufficient actual recovery after matching based on the mean, which could lead to continued excessive friction. The friction deviation is compared one by one with the recoverable range of each jamming recovery type. If the deviation falls within the lubrication replenishment recoverable range, the lubrication replenishment type is matched; if it exceeds the lubrication replenishment upper limit but falls within the clearance adjustment recoverable range, the clearance adjustment type is matched; if it exceeds the clearance adjustment upper limit, the local cleaning type is matched. When the match is lubrication replenishment or clearance adjustment, a minimum-level constraint indicator is added. The baseline recovery type records the matching result and the recoverable range margin. The smaller the margin, the closer the current deviation is to the upper limit of that type, and the higher the risk of recurrence after maintenance. The "Sudden Increase in Friction Deviation Depth" indicator has excluded lubrication replenishment levels. Matching starts with clearance adjustment. Structural damage caused by a sudden increase in depth may exceed the clearance adjustment recovery limit. The baseline recovery type adds a recovery limit risk indicator to the source of the sudden increase in depth. This recovery limit risk indicator triggers local cleaning as an alternative solution during the maintenance time estimation stage. The "Low Deviation in Friction Deviation" indicator has a large margin of recovery after lubrication replenishment. The baseline recovery type adds a generous margin indicator to the source of low deviation. This generous margin indicator uses historical averages rather than overestimations for the corresponding maintenance time reservation during the maintenance time estimation stage, shortening the time occupied by urgent maintenance windows.

[0056] Based on the baseline recovery type, the estimated maintenance time is determined by adjusting the reserved time according to the historical probability of overestimating similar maintenance time. Overestimation of similar maintenance time in the past is defined as the actual completion time being shorter than the reserved time. Overestimation wastes time during urgent maintenance windows, preventing other faults from being scheduled. The historical probability of overestimation is the proportion of actual maintenance time shorter than the reserved time in the historical records of the same model and maintenance type. If the overestimation probability exceeds 60%, it indicates that the reserved time for this type of maintenance is habitually too high, and the estimated maintenance time must be adjusted downwards to the 75th percentile of historical time, not the mean. The baseline recovery type's margin indicator item has low dispersion and high probability of overestimation in historical lubrication supplementary maintenance time. The reserved time is taken as the 75th percentile of history. Shortening the reserved time ensures that faults with ample margin in urgent maintenance windows do not occupy excessive downtime, reserving more window time for faults with smaller margins. The baseline recovery type's upper limit risk indicator item for historical local cleaning has many uncertainties in man-hours and a low probability of overestimation. The reserved man-hours are taken from the 90th percentile of historical data to cover these uncertainties and avoid forced work interruptions due to insufficient reserves. Local cleaning requires section-by-section inspection of worn parts, and the uncertainty stems from the breadth of the wear distribution. For the fault item with the smallest recoverable margin in the baseline recovery type, the reserved man-hours are increased by an additional 5%. This fault item is the most difficult to recover and often takes longer for operators to handle; the 5% increase corresponds to the average of the historical distribution of extended operation times. The estimated maintenance man-hours are then compared with the available time in the urgent maintenance window. If the total man-hours exceed the available time, the reserved man-hours are compressed item by item according to the recoverable margin, from largest to smallest, to the historical minimum. If the compressed man-hours still exceed the available time, a phased assessment is triggered. The assessment results are then transmitted to the maintenance task plan generation stage to form corresponding multiple downtime arrangements.

[0057] Maintenance task plans are constructed by compiling maintenance types and resource requirements lists based on estimated maintenance man-hours. The required personnel skill levels, tool lists, and spare parts models are retrieved from the maintenance database by maintenance type. These three resource requirements are then compared with current maintenance inventory. If spare parts inventory is insufficient, a spare parts shortage flag is added. Spare parts in shortage must be prepared before the urgent maintenance window, with the lead time determined by the 90th percentile of the supplier's historical delivery dates. For maintenance man-hours, the parallel operation flags must be verified to ensure that the required operation locations of the two types of operations do not conflict. Due to limited space on escalators, the operation locations for step chain maintenance and brake maintenance often overlap and cannot be carried out simultaneously. The maintenance task plan adds sequential operation flags to conflicting sources. After sequential arrangement, the total man-hours increase and must be re-verified to ensure that it does not exceed the available time of the urgent maintenance window. If it does, a phased shutdown arrangement is triggered. The minimum constraint flags for maintenance man-hours primarily require standard lubricants or standard clearance adjustment tools, with delivery times significantly shorter than those for component replacement. In urgent scheduling scenarios, spare parts are prioritized for procurement to ensure that maintenance execution is not delayed due to stockpiling delays. After the maintenance task plan is generated, the total number of personnel with the same skills required during the same downtime period is compared with the current upper limit of dispatchable personnel. If the upper limit is exceeded, the highest priority fault entries are retained according to T_margin,f from smallest to largest. Fault entries with larger margins are postponed to the next downtime period and a postponement execution mark is added. If T_margin,f is negative, postponement is not allowed and temporary emergency repair arrangements must be added. Temporary emergency repairs prioritize the use of the nearest available personnel to fill emergency intervention needs that cannot be covered by planned dispatch.

[0058] Step S150: Based on the maintenance task plan, select the personnel with the best performance in the same fault in history to form a response dispatch plan. Extract the power consumption deviation during the first peak period after maintenance from the response dispatch plan to obtain the maintenance performance score. Update the component monitoring threshold according to the maintenance performance score and output the operation and maintenance early warning instruction.

[0059] In some embodiments, the step of selecting personnel with the best historical performance and efficiency for the same fault based on the maintenance task plan to form a response dispatch plan includes: extracting fault types and required maintenance capability levels from the maintenance task plan to determine maintenance capability requirements; statistically analyzing historical performance and efficiency records of personnel repairing similar faults to construct a candidate personnel performance and efficiency set; identifying personnel with low rework rates and nearby dispatchability based on the candidate personnel performance and efficiency set to determine preferred dispatch personnel; and planning the optimal response route based on the preferred dispatch personnel's comprehensive dispatch time and the geographical coordinates of the ladder group to form a response dispatch plan.

[0060] The maintenance capability requirements are determined by extracting fault types and required maintenance capability levels from the maintenance task plan. The fault type field and maintenance type level field for each item in the maintenance task plan are combined and mapped to the maintenance personnel qualification level system. Lubrication maintenance corresponds to basic maintenance qualification, clearance adjustment to intermediate qualification, and component replacement to advanced qualification. Each qualification level corresponds to specific operational skill certification items. The maintenance capability requirements store the qualification level and certification item list using the fault item number as the primary key. When two types of processes correspond to different qualification levels, the maintenance capability requirements record the two qualification levels and the corresponding process sequence. Higher qualification level personnel have the ability to handle lower qualification level processes. When the number of dispatched personnel is limited, higher qualification level personnel can undertake the entire sequential process. The maintenance capability requirements add a "handling" label to such handling arrangements to distinguish between single-process dispatch and multi-process dispatch. The spare parts shortage flag in the maintenance task plan corresponds to the additional spare parts coordination qualification requirement for maintenance capability requirements. Spare parts coordination and on-site operations are recorded independently. Maintenance capability requirements add a coordination category flag to spare parts coordination requirements. Requirements with coordination flags are not included in the on-site operator quota. These two types of qualification requirements are retrieved separately during the candidate personnel quality and efficiency set construction phase. The completeness of the qualification list for maintenance capability requirements is verified by the qualification item dictionary in the operations and maintenance database. Certification items missing from the dictionary are marked as "to be supplemented." Requirements with "to be supplemented" flags are retrievald using higher-level compatible qualifications during the candidate personnel quality and efficiency set construction phase.

[0061] A candidate personnel performance evaluation set is constructed by statistically analyzing historical records of personnel performing similar faults to assess maintenance capability requirements. For each qualification level of maintenance capability requirements, all personnel meeting the certification requirements are retrieved from the registered personnel database. Records of similar fault repair events from the most recent 24 months are extracted using the personnel ID as the primary key. The normalized recovery rate of the first peak period after each repair event, along with the rework event marker, constitutes the individual performance evaluation record for that personnel. The candidate personnel performance evaluation set is organized using a dual index of personnel ID and repair event ID. Personnel retrieved from the maintenance capability requirements database who have performed fewer similar fault repairs than the minimum sample threshold are marked with a low-sample label. The performance evaluation of personnel with low-sample labels is considered unstable, and the confidence field for personnel from low-sample sources is lowered in the candidate personnel performance evaluation set. Personnel with low-sample labels are given lower priority than personnel with sufficient samples during the personnel selection phase. The median and interquartile range of the historical recovery rate for each employee in the candidate employee performance evaluation set jointly describe the employee's performance stability. Employees with a high median but a large interquartile range have insufficient operational stability. Employees in the candidate employee performance evaluation set whose interquartile range exceeds 1.5 times the average interquartile range of similar employees are marked with a high volatility label, and their weight is reduced during the selection and screening phase. Employees in the candidate employee performance evaluation set who have experienced rework events in their last three maintenance operations are marked with a recent rework label. Recent rework indicates that the employee's recent operational error rate is relatively high. Employees from which recent rework events originate in the candidate employee performance evaluation set are marked with a recent rework label, and their priority is reduced to the lowest tier in the current batch of dispatches.

[0062] Based on the candidate personnel performance set, personnel with low rework rates and those readily available for dispatch are selected as the preferred dispatchers. Operations such as escalator brake clearance adjustment and chain lubrication rely heavily on manual experience; the historical recovery rate distribution of personnel with equivalent qualifications can range from 15% to 25% across the interquartile range. Dispatching personnel with high recovery rates and low rework rates can significantly improve the effectiveness of single maintenance and reduce repeated downtime. The historical rework rate of each personnel in the candidate personnel performance set is calculated based on the proportion of historical maintenance events marked with rework. Personnel with historical rework rates lower than the average of their peers and historical normalized recovery rates higher than the average of their peers are included in the preferred candidates. Preferred candidates are ranked in descending order of the median of historical normalized recovery rates, with priority given to personnel with the highest recovery rates. Personnel with high volatility in the candidate personnel performance set are not included in the preferred candidates, and personnel with recent rework marks are simultaneously excluded. When both types of marks are triggered, the corresponding personnel in the candidate personnel performance set are marked with a double exclusion mark. Personnel with double exclusion marks are only considered as a last resort when the preferred candidate pool is exhausted. The travel time from the current personnel location to the corresponding escalator geographic coordinates in the maintenance task plan is estimated in real time by the map service. If the travel time exceeds 70% of the available arrival time calculated from the minimum T_margin,f of the critical margin set, the personnel are marked with a distance exceeding the limit. Although personnel with the distance exceeding the limit have excellent performance and efficiency, they cannot arrive on time. The preferred dispatch personnel are re-ranked from the remaining candidates according to the recovery rate. Distance constraints and performance and efficiency constraints are equally important, and arrival timeliness is not ignored due to excellent performance and efficiency. Personnel with low sample quality and efficiency in the candidate personnel performance and efficiency set are added to the preferred candidate pool when there is insufficient personnel. Preferred dispatch personnel with low sample source are marked with a low sample source. During the maintenance performance and efficiency scoring stage, the current maintenance record of personnel with low sample source is included in their historical performance and efficiency time series. After the current maintenance is completed, the low sample mark is updated.

[0063] A response dispatch plan is generated based on the optimal response route planned according to the selected personnel's overall dispatch time and the geographical coordinates of the ladder group. The current location coordinates of the selected personnel and the corresponding ladder group geographical coordinates of the maintenance task plan are input into the route planning module. The route planning integrates three travel modes: walking, public transportation, and driving, selecting the route with the shortest arrival time as the recommended route. The arrival time of the recommended route must be at least 30 minutes earlier than the start time of the urgent maintenance window. This 30-minute interval corresponds to the pre-operation time required for maintenance preparation and equipment confirmation. If the arrival time does not meet the requirement, a closer candidate is retrieved from the candidate personnel quality and efficiency set to replace the original candidate. When multiple dispatched personnel are going to the same ladder group, the response dispatch plan checks the route overlap. For sections with an overlap exceeding 70%, it is recommended that personnel travel together to save traffic resources. A collaborative travel suggestion is added, and this suggestion triggers the generation of a meeting point and meeting time during the response execution phase. Prioritizing dispatched personnel with a dual-source identifier, both types of procedures must be completed within the urgent repair window. The dispatch plan generates a procedure time allocation scheme for these personnel. The two types of procedures are arranged according to the estimated repair time and reserved time, with a 5-minute tool changeover time allowed between procedures. The total time must not exceed the available time within the urgent repair window to pass. Each item in the dispatch plan is organized by dispatched personnel number, recommended route, estimated arrival time, and the fault item number they are responsible for. The time difference between the estimated arrival time and the start time of the urgent repair window constitutes the response time margin. If the response time margin is less than 15 minutes, the dispatch plan adds a time-urgent identifier to the corresponding item. This time-urgent identifier triggers simultaneous notifications via SMS and the system to ensure timely delivery of dispatch information.

[0064] The maintenance quality and efficiency score is obtained by extracting the power consumption deviation during the first peak period after maintenance in response to the dispatch plan. The decrease in the average deviation of the operating status during the morning peak period on the first official operating day after maintenance compared to the same period before maintenance is defined as the power consumption deviation during the first peak period. The morning peak period has the highest passenger flow density and the drive motor load is closest to the rated state. The power consumption deviation during this period best represents the true recovery range of the components under actual operating conditions after maintenance. After the maintenance of the fault items assigned to each dispatched personnel in response to the dispatch plan is completed, the power consumption deviation during the first peak period is divided by the friction deviation before maintenance to obtain the normalized recovery rate. The normalized recovery rate eliminates the differences in the initial deviation degree of different faults, allowing for horizontal comparison of the quality and efficiency of different personnel across faults. The maintenance quality and efficiency score is summarized and stored by maintenance personnel number based on the normalized recovery rate. When power consumption deviation is estimated during the first peak period after maintenance, affected by seasonal or abnormal passenger flow, the maintenance quality and efficiency score must verify the ratio of passenger flow density during the first peak period to the historical passenger flow density for the same period. If the ratio deviates from the average by more than 20%, an abnormal passenger flow indicator is added to the corresponding item of the maintenance quality and efficiency score. The normalized recovery rate of the abnormal passenger flow indicator item is replaced by the power consumption deviation of adjacent normal passenger flow batches to ensure that the maintenance quality and efficiency score reflects the maintenance effect rather than passenger flow fluctuations. After a valid record is added to the maintenance quality and efficiency score of personnel with low sample indicators in response to the dispatch plan, the low sample indicator is updated in the personnel's historical quality and efficiency time series. The subsequent priority of this personnel is re-evaluated based on the updated time series, and the low sample indicator gradually falls off as the sample accumulates. When the normalized recovery rate of each personnel's maintenance quality and efficiency score is lower than the historical average of the same type of fault twice consecutively, a low quality and efficiency indicator is marked. During the operation and maintenance early warning instruction generation stage, the monitoring threshold of the elevator group under the personnel's responsibility is tightened. The incomplete recovery of the source of low quality and efficiency must be compensated by detection in a shorter cycle.

[0065] The system updates component monitoring thresholds and outputs maintenance warning commands based on maintenance quality and efficiency scores. When the normalized recovery rate of each fault item in the maintenance quality and efficiency score is higher than the historical average, it indicates sufficient recovery during the maintenance. The corresponding component monitoring threshold can be appropriately relaxed based on the current threshold. The relaxation amount is determined by mapping the proportion of the normalized recovery rate exceeding the average. A higher recovery rate indicates a greater reduction in friction resistance and a larger buffer space before the next alarm trigger. When the normalized recovery rate is lower than the historical average, the maintenance recovery is insufficient, and the residual friction resistance deviation is relatively large. The corresponding component monitoring threshold must be tightened. The tightening amount is determined by mapping the proportion of the recovery rate below the average. The maintenance warning command adds a tightened monitoring identifier to the source of the threshold tightening. The tightened monitoring identifier will be used as the basis for determining deviation exceeding limits in the subsequent deviation statistics phase of the elevator group. The monitoring threshold for maintenance early warning commands for elevator groups with low maintenance efficiency ratings is tightened by an additional 5% on top of the tightened recovery rate threshold. Residual friction deviations introduced by non-standard operations by low-efficiency personnel are usually distributed in areas with uneven operation coverage. This additional tightening ensures that residual deviations are detected early, preventing delays in triggering alarms due to lenient thresholds until the next periodic inspection. The threshold update for the passenger flow anomaly rating in the maintenance efficiency rating is calculated based on the power consumption deviation after replacement. A passenger flow anomaly replacement flag is added to the corresponding maintenance early warning command threshold update. The confidence level of the passenger flow anomaly replacement source threshold update is lower than that of the normal peak period source. This confidence level is recorded in the maintenance early warning command metadata for traceability in the next maintenance cycle. When multiple monitoring threshold updates for the same elevator group in the maintenance early warning command have the same direction, the threshold update amount is accumulated. For elevator groups with multiple sufficient recoveries, the threshold is gradually relaxed, but the upper limit is set to the healthy baseline threshold to prevent excessive relaxation from masking the true degradation signal. The maintenance early warning command adds a threshold upper limit flag to items that reach the upper limit. Elevator groups with the upper limit flag will no longer have their thresholds relaxed, maintaining the healthy baseline threshold for long-term monitoring.

[0066] To implement the above-described method embodiments, a smart operation and maintenance early warning method for escalators is proposed to achieve the corresponding functions and technical effects. See also... Figure 2 , Figure 2 This diagram illustrates a structural block diagram of an intelligent escalator maintenance and early warning system 200 according to an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The intelligent escalator maintenance and early warning system 200 provided in this embodiment includes: The data acquisition module 201 is used to collect operating status data and braking timing data, generate operating status deviation from the benchmark by statistically analyzing the deviation of each operating status from the benchmark based on the operating status data, and obtain braking cycle dispersion by aggregating the dispersion of braking cycle duration for each batch based on the braking timing data. Trend analysis module 202 is used to locate the step accumulation boundary and form a friction zone based on the deviation of the operating state, and to identify the speed segment with high discrete batch aggregation by using the brake cycle discrete amount to determine the brake anomaly source marker. The fault diagnosis module 203 is used to compare the state differences of each speed segment in the friction section to generate component wear marks, determine the related components of the braking anomaly source mark based on the component wear marks to form an associated fault mode, and identify the decline inflection point of power consumption recovery rate of previous similar repairs based on the associated fault mode to determine the fault urgency rating. The maintenance planning module 204 is used to analyze the critical time period of each fault combination based on the fault urgency rating to obtain a critical margin set, use the critical margin set to determine the nearest maintenance time period of the minimum margin item to obtain an urgent maintenance window, and identify the lowest level maintenance type that can restore friction from the urgent maintenance window to establish a maintenance task plan. The instruction output module 205 is used to select the personnel with the best historical quality and efficiency for the same fault based on the maintenance task plan to form a response dispatch plan, extract the power consumption deviation during the first peak period after maintenance to obtain a maintenance quality and efficiency score based on the response dispatch plan, update the component monitoring threshold based on the maintenance quality and efficiency score, and output an operation and maintenance early warning instruction.

[0067] The aforementioned intelligent operation and maintenance early warning system 200 for escalators can implement one of the intelligent operation and maintenance early warning methods for escalators described in the above method embodiments. The options in the above method embodiments are also applicable to this embodiment, and will not be detailed here. The remaining content of this application embodiment can be referred to the content of the above method embodiments, and will not be repeated in this embodiment.

[0068] The above embodiments are not an exhaustive list based on the present invention, and there may be many other embodiments not listed. Any substitutions and improvements made without departing from the concept of the present invention are within the protection scope of the present invention.

Claims

1. A smart operation and maintenance early warning method for escalators, characterized in that, include: Collect operating status data and braking timing data, calculate the deviation between each operating status and the benchmark from the operating status data to generate the operating status deviation, and collect the braking cycle dispersion based on the dispersion of the braking cycle duration of each batch of braking timing data. Based on the deviation of the operating state, the step accumulation boundary is located to form the friction zone, and the speed segment with high discrete batch aggregation is identified by using the discrepancy of the braking cycle to determine the braking anomaly source marker. The wear marks of components are generated by comparing the state differences of each speed segment in the friction section. The wear marks of components are used to identify the components related to the brake anomaly source to form an associated fault mode. Based on the associated fault mode, the decline in the power consumption recovery rate of previous similar repairs is identified to determine the fault urgency rating. Based on the fault urgency rating, the critical time period of each fault combination is analyzed to obtain the critical margin set. The critical margin set is used to determine the nearest maintenance time period of the item with the smallest margin to obtain the urgent maintenance window. From the urgent maintenance window, the lowest level maintenance type that can restore friction is identified to establish a maintenance task plan. Based on the maintenance task plan, the personnel with the best performance in the same fault in history are selected to form a response dispatch plan. The power consumption deviation during the first peak period after maintenance is extracted from the response dispatch plan to obtain the maintenance performance score. The component monitoring threshold is updated according to the maintenance performance score and maintenance early warning instructions are output.

2. The method according to claim 1, characterized in that, The step of generating the operating status deviation by statistically analyzing the deviations between each operating status and the baseline from the operating status data includes: The running state sequence is obtained by extracting the state features of each running segment from the running state data; The initial value of the state deviation is obtained by statistically analyzing the difference between the mean of each running state and the historical benchmark mean of the running state sequence. Based on the deviation from the initial value of the state, the consistency of the deviation amplitude after environmental temperature and humidity correction between batches is formed to generate the environmental correction deviation amount. The operational state deviation is determined by aggregating the mean of each state deviation after environmental correction based on the aforementioned environmental correction deviation.

3. The method according to claim 1, characterized in that, The method of obtaining the braking cycle dispersion based on the dispersion of the braking cycle duration of each batch of braking time sequence data includes: The braking time sequence data is used to extract the start and end time intervals of each braking event to generate a braking duration sequence; The batch duration dispersion is calculated by comparing the range and mean of each batch duration in the braking duration sequence. The ratio trend of the dispersion of high-load and low-load shutdown batches is identified based on the dispersion of the batch duration to form a discretely increasing record; The braking cycle dispersion is determined based on the deviation ratio between the dispersion of each batch and the initial mean in the discrete incremental record collection.

4. The method according to claim 1, characterized in that, The step of identifying speed segments with highly discrete batch clusters and determining braking anomaly source markers using the braking cycle discreteness includes: The time of the over-threshold discrete batch and the operating condition parameters are extracted from the discrete amount of the braking cycle to form a highly discrete batch operating condition set; The speed distribution of each batch is determined by analyzing the running speed characteristics at the braking moment of each batch in the highly discrete batch operating condition set. Based on the statistical analysis of the peak density of highly discrete batches in each speed interval and their time drift direction, a clustering speed interval is established. Based on the aforementioned aggregation speed range, the low-speed segment is identified as elastic failure and the high-speed segment is identified as friction failure, generating a braking anomaly source marker.

5. The method according to claim 1, characterized in that, The method of determining the fault urgency rating based on the decline inflection point of power consumption recovery rate of similar repairs in the context of the associated fault mode identification includes: By using the associated fault mode, records of operational status deviations before and after each maintenance for the same model and fault are retrieved to obtain the status records before and after maintenance. The single-time state recovery rate is determined by calculating the baseline correction deviation recovery ratio based on the pre- and post-maintenance state records. Based on the single-state recovery rate, the inflection point where the continuous maintenance recovery rate changes from a uniform decline to an accelerated decline is identified, and the amount of recovery rate decline is determined. A fault urgency rating is formed based on the maintenance intervention time limit after the recovery rate decline enters the accelerated decline zone.

6. The method according to claim 1, characterized in that, The step of identifying the lowest-level repair type capable of restoring friction from the urgent repair window and establishing a repair task plan includes: The friction deviation is obtained by analyzing the magnitude of the sudden increase in friction after the abnormal stabilization of the corresponding component when the emergency maintenance window is triggered. The baseline recovery type is formed by matching the friction deviation to achieve the lowest intervention level of the stuck recovery type; Based on the aforementioned baseline recovery type, the estimated maintenance time is determined by adjusting the reserved time according to the historical probability of overestimation of similar maintenance time. A maintenance task plan is constructed by compiling a list of maintenance types and resource requirements based on the estimated maintenance man-hours.

7. The method according to claim 1, characterized in that, The step of selecting personnel with the best historical performance in handling similar faults based on the maintenance task plan to form a response dispatch plan includes: The maintenance capability requirements are determined by extracting the fault type and the required maintenance capability level from the maintenance task plan. A candidate personnel performance set is constructed by statistically analyzing the historical performance records of repair personnel for similar faults to meet the aforementioned maintenance capability requirements. Based on the candidate personnel quality and efficiency set, personnel with low maintenance rework recurrence rate and nearby dispatchable personnel are identified to determine the preferred dispatch personnel; Based on the optimal dispatch time and the geographical coordinates of the selected personnel, an optimal response route is planned to form a response dispatch plan.

8. The method according to claim 4, characterized in that, The step of analyzing the braking speed characteristics of each batch in the highly discrete batch operating condition set to determine the batch speed distribution includes: Extract the braking start time and operating condition characteristics of each braking instance within the highly discrete batch operating condition set to form a braking start operating condition set. The braking speed value is obtained by parsing the operating speed value at the corresponding moment using the braking initiation condition set. Based on the speed value at the braking moment, the frequency of braking frequency within each preset speed interval is statistically distributed to determine the frequency of the speed interval; The batch speed distribution is determined by aggregating the frequency proportion of each speed interval and the abnormal missing speed intervals.

9. The method according to claim 5, characterized in that, The determination of the single-state recovery rate based on the baseline correction deviation recovery ratio calculated from the pre- and post-maintenance state records includes: The baseline state before maintenance is obtained by statistically analyzing the average deviation of the last batch of operating states before maintenance using the aforementioned state records before and after maintenance. Based on the state records before and after maintenance, the mean deviation of the first batch of operating states after maintenance is statistically analyzed to obtain the mean deviation of the state after maintenance. The deviation recovery ratio, after removing overcompensation, is calculated based on the average deviation between the pre-maintenance baseline and the post-maintenance baseline. The single-state recovery rate is determined based on the deviation of the average historical recovery rate of this model for this fault from the aforementioned deviation recovery ratio aggregation.

10. An intelligent operation and maintenance early warning system for escalators, characterized in that, include: The data acquisition module is used to collect operating status data and braking timing data. It generates operating status deviation by statistically analyzing the deviation between each operating status and the benchmark from the operating status data, and obtains braking cycle dispersion by aggregating the dispersion of braking cycle duration for each batch based on the braking timing data. The trend analysis module is used to locate the step accumulation boundary and form the friction zone based on the deviation of the operating state, and to identify the speed segment with high discrete batch aggregation by using the braking cycle discrete quantity to determine the braking anomaly source marker. The fault diagnosis module is used to compare the state differences of each speed segment in the friction section to generate component wear marks, identify the components related to the brake anomaly source mark based on the component wear marks to form an associated fault mode, and identify the decline in the power consumption recovery rate of previous similar repairs based on the associated fault mode to determine the fault urgency rating. The maintenance planning module is used to analyze the critical time period of each fault combination based on the fault urgency rating to obtain a critical margin set, use the critical margin set to determine the nearest maintenance time period of the item with the smallest margin to obtain an urgent maintenance window, and identify the lowest level maintenance type that can restore friction from the urgent maintenance window to establish a maintenance task plan. The instruction output module is used to select the personnel with the best historical quality and efficiency for the same fault based on the maintenance task plan to form a response dispatch plan, extract the power consumption deviation during the first peak period after maintenance to obtain a maintenance quality and efficiency score based on the response dispatch plan, update the component monitoring threshold based on the maintenance quality and efficiency score, and output maintenance early warning instructions.

Images