Escalator multi-source data anomaly detection method and system

Abstract

The invention discloses an escalator multi-source data anomaly detection method and system, and relates to the technical field of escalator intelligent detection.The method comprises the steps that an original operation data set is constructed by monitoring the vibration acceleration, the operation temperature, the operation current and the speed multi-source time sequence of an escalator in real time; performing time correction, deletion repair and noise suppression on the original data, and extracting a period parameter, a trend parameter and a residual parameter; calculating a period stability deviation coefficient, a trend evolution deviation coefficient and a multi-source residual consistency coefficient, performing evaluation, and judging whether the period structure, the operation trend and the residual consistency are abnormal or not; generating mark information for the abnormal condition; and finally, uniformly summarizing all the abnormal characteristics, and storing the abnormal characteristics, the original data and the characteristic parameters together to realize closed-loop abnormal detection and recording of the running state of the escalator. According to the method, the cycle, trend and systematic abnormity can be accurately recognized, and the reliability and safety of escalator operation monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection technology for escalators, specifically to a method and system for detecting anomalies in multi-source data of escalators. Background Technology

[0002] With the rapid development of urban public transportation and large commercial complexes, the operational safety and reliability of escalators, as high-frequency, continuously operating electromechanical equipment, have received widespread attention.

[0003] Current escalator operation monitoring relies heavily on single or limited sensor parameters for status assessment, typically employing threshold alarms or post-fault analysis methods. This approach fails to comprehensively reflect changes in the equipment's operational characteristics under complex conditions. Furthermore, escalator operation is influenced by multiple factors, including load fluctuations, operational cycle variations, environmental disturbances, and component aging. This results in operational data exhibiting clear periodicity, trends, and systematic coupling characteristics, making it difficult for traditional monitoring methods to uniformly characterize and systematically manage abnormal information from multiple data sources.

[0004] Furthermore, existing technologies for storing and utilizing anomaly identification results are mostly limited to single-record processes, lacking structured data accumulation that reflects historical operational evolution, which is detrimental to subsequent operational analysis and maintenance decision support. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for detecting anomalies in multi-source data of escalators, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for detecting anomalies in multi-source data of escalators, comprising the following steps:

[0007] Step 1: Monitor the motion status of the pedestrian overpass escalator in real time; synchronously collect and record timestamps, and encapsulate the multi-source time series of vibration acceleration, operating temperature, operating current and speed with timestamps to construct the original multi-source operation dataset of the escalator;

[0008] Step 2: By performing time correction, missing data repair, and noise suppression on the original multi-source operation dataset of the escalator, a stable operation data sequence is formed; based on multi-cycle cumulative analysis, the cycle parameters of normal escalator operation are extracted and a shared cyclic cycle structure is constructed, thereby separating the cycle component parameters, trend parameters, and residual parameters of each channel;

[0009] Step 3: Obtain the average energy value of the current cycle. Compared with historical cycle energy average The periodic stability deviation coefficient ZPX is calculated and compared with the periodic stability threshold Zth to determine whether the periodic structure of the escalator is qualified. If it is not qualified, the first strategy is applied to generate periodic anomaly marker information.

[0010] Step 4: By obtaining the measured trend sequence Xsc(t) and the predicted trend sequence Xyq(t), calculate the trend evolution deviation coefficient QYP and compare it with the trend evolution threshold Qth to determine whether the relatively stable evolution trajectory of the escalator's running trend has deviated. If it deviates, a second strategy is given to generate trend anomaly marker information.

[0011] Step 5: By acquiring the residual sequence r(t) corresponding to each acquisition channel, calculate the multi-source residual consistency coefficient DCY and compare it with the residual consistency threshold Dth to determine whether the multi-source residual consistency is qualified. If it is not qualified, the third strategy is applied to generate systematic anomaly confirmation marker information. Combining the periodic anomaly marker information and the trend anomaly marker information, the abnormal characteristics of the current operating status of the escalator are uniformly summarized and stored together with the corresponding multi-source operating data and feature parameters to complete the closed-loop recording of the escalator anomaly detection results.

[0012] Preferably, step one includes:

[0013] S11. By monitoring the mechanical, electrical, and motion states of escalators operating in pedestrian overpasses under conditions of continuous start-stop, high load, long-term operation, and significant fluctuations in passenger flow, and collecting basic operating data of the escalators, including the following steps:

[0014] S111. Vibration acceleration sensors are installed in the drive motor housing, gearbox housing, main drive bearing housing, and handrail belt bearing installation positions of the escalator to collect the vibration acceleration time series a(t) during the escalator operation in real time.

[0015] S112. By installing a current sensor in the power supply circuit or inside the control cabinet of the escalator drive motor, the operating temperature time series T(t) of the drive motor during the escalator operation is collected in real time.

[0016] S113. By installing a current sensor in the power supply circuit or inside the control cabinet of the escalator drive motor, the operating current time sequence I(t) of the drive motor during the escalator operation is collected in real time.

[0017] S114. By installing speed sensors and displacement sensors at the step chain drive part, main drive shaft or step running track position of the escalator, the speed time sequence v(t) generated by the step motion during the operation of the escalator is collected in real time.

[0018] S12. By configuring a unified clock module or high-precision system clock in the data acquisition and control device of the escalator, the acquisition process of vibration sensors, temperature sensors, current sensors, and speed or displacement sensors installed in different locations is synchronously triggered or synchronously sampled and controlled. While various sensors are acquiring data, the corresponding acquisition time identifier is recorded for each sampled data; the acquisition time identifier is organized into a timestamp sequence t.

[0019] S13. The collected vibration acceleration time series a(t), operating temperature time series T(t), operating current time series I(t), velocity or displacement time series v(t), and corresponding timestamp sequence t are uniformly packaged to form the original multi-source operation dataset of the escalator.

[0020] Preferably, step two includes:

[0021] S21. Based on the original multi-source operation dataset of the escalator, timestamp alignment and interpolation repair technology is used to uniformly correct and complete the sampling offset, time misalignment and local data loss of different acquisition channels during operation; filtering and smoothing technology is used to suppress high-frequency interference in vibration signals, transient fluctuations in current signals and measurement noise in temperature and speed data; and a stable operation data sequence is established.

[0022] S22. During the continuous operation of the escalator, the stable operation data sequence is accumulated and stored over multiple operating cycles to form a historical stable operation dataset. Based on the historical stable operation dataset, the periodic change pattern of multi-source operation data under normal operating conditions is analyzed using the time series periodic feature analysis method, and the operating cycle parameter W of the stable operation rhythm of the escalator is extracted.

[0023] S23. Based on the running cycle parameter W, the period alignment and time index mapping method is used to perform period segmentation and alignment processing on the stable running data sequence in the time dimension, and the continuous running data is divided into multiple period segments of the same length according to the running cycle parameter; the cross-cycle feature alignment and structured modeling method is used to uniformly describe the changing relationship of running data at the same period position of different acquisition channels, and construct a shared cyclic period matrix Q.

[0024] S24. Based on the shared cyclic periodic matrix Q and the stable operating data sequence, and combined with the index position of the stable operating data on the time axis, a periodic position mapping and periodic component instantiation extraction method is used to perform periodic matching processing on the operating data of each acquisition channel in the stable operating data sequence. During the processing, according to the relative position of the operating data in its respective periodic segment, the periodic structure information corresponding to the current position is extracted from the shared cyclic periodic matrix and mapped to each acquisition channel to form periodic component parameters. ;

[0025] S25. Based on the stable operation data sequence and periodic component parameters, the periodic influence stripping and operation state decomposition method is adopted to separate the changes caused by periodic operation behavior in the stable operation data; by removing the corresponding periodic component parameters from the stable operation data of each acquisition channel, the trend parameter Xtrend(t) is constructed; the remaining changes in the stable operation data that are not explained by the periodic components are organized into the residual parameter Xres(t).

[0026] Preferably, step three includes:

[0027] S31, through the obtained periodic component parameters Using the periodic energy statistical analysis method with the operating cycle parameter W, and after dimensionless processing, the average value of the current periodic energy is obtained. Compared with historical cycle energy average .

[0028] Preferably, step three further includes:

[0029] S32, by obtaining the average energy of the current cycle Compared with historical cycle energy average After dimensionless processing, the periodic stability deviation coefficient ZPX is calculated and obtained.

[0030] S33. By setting a preset periodic stability threshold Zth, and comparing the periodic stability deviation coefficient ZPX with the periodic stability threshold Zth, the first evaluation result is obtained, including:

[0031] When the periodic stability deviation coefficient ZPX ≤ periodic stability threshold Zth, it indicates that the periodic structure of the escalator is qualified and continuous monitoring is required.

[0032] When the periodic stability deviation coefficient ZPX > the periodic stability threshold Zth, it indicates that the escalator's periodic structure is substandard, with abnormal deviations in periodic energy distribution. This poses a risk of periodic rhythm disorder, load fluctuations, or mechanical wear evolution, triggering the first warning instruction and generating the first strategy: reducing the periodic component parameters by 15% to 30%. The participation rate in relevant statistical calculations is reduced to minimize the impact of periodic changes caused by the normal periodic start-stop of escalators and load fluctuations on the anomaly judgment results, thereby generating periodic anomaly marker information.

[0033] Preferably, step four includes:

[0034] S41. Based on the trend parameter Xtrend(t) and the residual parameter Xres(t), the trend parameter is used as a time series expression that truly reflects the long-term evolution state of the escalator during continuous operation, and the measured trend sequence Xsc(t) of the corresponding acquisition channel is constructed.

[0035] S42. Based on the trend parameter Xtrend(t) and residual parameter Xres(t), and combined with the historical stable operation dataset, the trend evolution characteristics of the corresponding acquisition channel in the historical stable operation data are used as input. The trend prediction model is used to model the change law of the trend residual sequence in the time dimension. By fitting and extrapolating the time evolution relationship of the trend residual sequence under the historical stable operation state, the trend change reference result corresponding to the current operation time period is obtained, and the predicted trend sequence Xyq(t) is formed.

[0036] Preferably, step four further includes:

[0037] S43. After dimensionless processing of the obtained measured trend sequence Xsc(t) and predicted trend sequence Xyq(t), the trend evolution deviation coefficient QYP is calculated.

[0038] S44. By setting a preset trend evolution threshold Qth, and comparing the trend evolution deviation coefficient QYP with the trend evolution threshold Qth, the second evaluation result is obtained, including:

[0039] When the trend evolution deviation coefficient QYP ≤ the trend evolution threshold Qth, it indicates that the escalator's operating trend is relatively stable and the evolution trajectory has not deviated, and there is no risk of non-periodic degradation. Continuous monitoring is required.

[0040] When the trend evolution deviation coefficient QYP > the trend evolution threshold Qth, it indicates that the relatively stable evolution trajectory of the escalator's operating trend has deviated, and there is a risk of non-periodic abnormal evolution caused by mechanical wear, electrical performance degradation, or long-term load changes. This triggers a second warning instruction and generates a second strategy: marking the current operating state as a potential degradation state; adjusting the discrimination weights of the trend parameter and the residual parameter, increasing the discrimination weight corresponding to the trend parameter by 10% to 25%, and simultaneously increasing the discrimination weight corresponding to the residual parameter by 5% to 15%, thereby enhancing the ability to identify non-periodic evolution anomalies and long-term degradation trends; and generating trend anomaly marking information.

[0041] Preferably, step five includes:

[0042] S51. Based on the measured trend sequence Xsc(t) and the predicted trend sequence Xyq(t) obtained in step four, the time-by-time difference calculation method is used to perform difference calculation on the measured trend data and the predicted trend data of the same acquisition channel at the same running time to obtain the residual change of the corresponding acquisition channel in the time dimension; and the residual change of each acquisition channel in the continuous operation process is sorted in time order to form the residual sequence r(t) of each acquisition channel.

[0043] Preferably, step five further includes:

[0044] S52. Based on the residual sequence r(t) corresponding to each acquisition channel, the correlation analysis method is used to quantitatively calculate the consistency of the changing trends between the residual sequences of any two different acquisition channels; by statistically summarizing the residual correlation results corresponding to all different acquisition channels, and after dimensionless processing, the multi-source residual consistency coefficient DCY is calculated.

[0045] S53. By setting a residual consistency threshold Dth, and comparing the multi-source residual consistency coefficient DCY with the residual consistency threshold Dth, the third evaluation results are obtained, including:

[0046] When the multi-source residual consistency coefficient DCY < residual consistency threshold Dth, it indicates that the multi-source residual consistency is qualified and should be continuously monitored.

[0047] When the multi-source residual consistency coefficient DCY ≥ residual consistency threshold Dth, it indicates that the multi-source residual consistency is unqualified. The direction and trend of residual changes in multiple acquisition channels are consistent, and the abnormal deviation shows cross-channel synchronous evolution characteristics. There is a systemic operational risk caused by structural wear accumulation, system performance degradation, or overall operating condition deterioration. This triggers the third early warning instruction and generates the third strategy: determine the current escalator operating status as a systemic abnormal state; confirm that the abnormal impact has evolved from local disturbance to cross-channel consistency deviation, with the risk of persistence and spread; and generate systemic abnormality confirmation mark information.

[0048] S54. Based on the generated periodic anomaly marker information, trend anomaly marker information, and systemic anomaly confirmation marker information, the abnormal characteristics of the current operating status of the escalator are uniformly summarized; the anomaly marker information, along with the corresponding multi-source operating data, operating cycle parameters, and trend parameters, are stored in the historical operating database; the required periodic structure data and trend evolution reference data are analyzed to complete the closed-loop recording of the multi-source data anomaly detection process of the escalator.

[0049] Preferably, an escalator multi-source data anomaly detection system includes:

[0050] The multi-source operation data acquisition module is used to monitor the motion status of pedestrian overpass escalators in real time; it synchronously collects and records timestamps, and encapsulates the multi-source time series of vibration acceleration, operating temperature, operating current and speed with timestamps to construct the original multi-source operation dataset of the escalator;

[0051] The data preprocessing module is used to perform time correction, missing data repair, and noise suppression on the original multi-source operation dataset of the escalator to form a stable operation data sequence. Based on multi-cycle cumulative analysis, the periodic parameters of the escalator's normal operation are extracted and a shared cyclic periodic structure is constructed, thereby separating the periodic component parameters, trend parameters, and residual parameters of each channel.

[0052] The periodic stability assessment module is used to obtain the average energy of the current period. Compared with historical cycle energy average The periodic stability deviation coefficient ZPX is calculated and compared with the periodic stability threshold Zth to determine whether the periodic structure of the escalator is qualified. If it is not qualified, the first strategy is applied to generate periodic anomaly marker information.

[0053] The trend evolution deviation assessment module is used to obtain the measured trend sequence Xsc(t) and the predicted trend sequence Xyq(t), calculate the trend evolution deviation coefficient QYP, and compare it with the trend evolution threshold Qth to determine whether the escalator's running trend has deviated from the relatively stable evolution trajectory. If it deviates, a second strategy is given to generate trend anomaly marker information.

[0054] The multi-source residual consistency assessment and recording module is used to acquire the residual sequence r(t) corresponding to each acquisition channel, calculate the multi-source residual consistency coefficient DCY, and compare it with the residual consistency threshold Dth to determine whether the multi-source residual consistency is qualified. If it is not qualified, a third strategy is adopted to generate systematic anomaly confirmation marker information. Combining periodic anomaly marker information and trend anomaly marker information, the abnormal characteristics of the current operating status of the escalator are uniformly summarized and stored together with the corresponding multi-source operating data and feature parameters to complete the closed-loop recording of escalator anomaly detection results.

[0055] This invention provides a method and system for detecting anomalies in multi-source data of escalators. It has the following beneficial effects:

[0056] (1) The method and system for detecting anomalies in multi-source data of escalators encapsulates multi-source time series such as vibration, temperature, current and speed through unified timestamps, and performs time correction, missing data repair and noise suppression processing, which effectively eliminates the impact of asynchronous sampling of multiple sensors and unstable data quality on the analysis results, and improves the reliability of escalator operation data modeling and analysis.

[0057] (2) The method and system for detecting anomalies in multi-source data of escalators, by constructing a shared cyclic structure, separates the periodic components, trend components and residual components in multi-source operation data, avoids the interference of normal periodic start-stop and load fluctuations on anomaly judgment, makes anomaly identification more focused on actual operation deviation, and improves the accuracy and pertinence of anomaly detection.

[0058] (3) The method and system for detecting anomalies in multi-source data of escalators can achieve a multi-dimensional comprehensive characterization of escalator operation anomalies by constructing periodic stability deviation coefficient, trend evolution deviation coefficient and multi-source residual consistency coefficient respectively. It can identify short-term periodic disorder, as well as long-term degradation trend and systemic anomaly risk, thus enhancing the comprehensiveness of operation status assessment.

[0059] (4) The method and system for detecting anomalies in multi-source data of escalators uniformly summarizes periodic anomaly markers, trend anomaly markers and systemic anomaly confirmation markers, and stores them together with the corresponding multi-source operating data and characteristic parameters to build a complete historical operating record, providing a reliable data foundation for subsequent operation status tracing, trend analysis and operation and maintenance decision-making. Attached Figure Description

[0060] Figure 1 This is a schematic diagram illustrating the steps of a multi-source data anomaly detection method for escalators according to the present invention;

[0061] Figure 2 This is a flowchart of a multi-source data anomaly detection system for escalators according to the present invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0063] Example 1

[0064] Please see Figure 1 This invention provides a method for detecting anomalies in multi-source data of escalators, comprising the following steps:

[0065] Step 1: Monitor the motion status of the pedestrian overpass escalator in real time; synchronously collect and record timestamps, and encapsulate the multi-source time series of vibration acceleration, operating temperature, operating current and speed with timestamps to construct the original multi-source operation dataset of the escalator;

[0066] Step 2: By performing time correction, missing data repair, and noise suppression on the original multi-source operation dataset of the escalator, a stable operation data sequence is formed; based on multi-cycle cumulative analysis, the cycle parameters of normal escalator operation are extracted and a shared cyclic cycle structure is constructed, thereby separating the cycle component parameters, trend parameters, and residual parameters of each channel;

[0067] Step 3: Obtain the average energy value of the current cycle. Compared with historical cycle energy average The periodic stability deviation coefficient ZPX is calculated and compared with the periodic stability threshold Zth to determine whether the periodic structure of the escalator is qualified. If it is not qualified, the first strategy is applied to generate periodic anomaly marker information.

[0068] Step 4: By obtaining the measured trend sequence Xsc(t) and the predicted trend sequence Xyq(t), calculate the trend evolution deviation coefficient QYP and compare it with the trend evolution threshold Qth to determine whether the relatively stable evolution trajectory of the escalator's running trend has deviated. If it deviates, a second strategy is given to generate trend anomaly marker information.

[0069] Step 5: By acquiring the residual sequence r(t) corresponding to each acquisition channel, calculate the multi-source residual consistency coefficient DCY and compare it with the residual consistency threshold Dth to determine whether the multi-source residual consistency is qualified. If it is not qualified, the third strategy is applied to generate systematic anomaly confirmation marker information. Combining the periodic anomaly marker information and the trend anomaly marker information, the abnormal characteristics of the current operating status of the escalator are uniformly summarized and stored together with the corresponding multi-source operating data and feature parameters to complete the closed-loop recording of the escalator anomaly detection results.

[0070] In this embodiment, by performing decomposition analysis of the periodicity, trend, and residuals of the multi-source operation data of the escalator, and combining the periodic stability, trend evolution deviation, and multi-source residual consistency for comprehensive evaluation, the system achieves step-by-step identification and unified labeling of operation anomalies from the local to the system level, avoiding misjudgment based on a single indicator and improving the accuracy and reliability of escalator operation anomaly detection.

[0071] Example 2

[0072] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically, step one includes:

[0073] S11. By monitoring the mechanical, electrical, and motion states of escalators operating in pedestrian overpasses under conditions of continuous start-stop, high load, long-term operation, and significant fluctuations in passenger flow, and collecting basic operating data of the escalators, including the following steps:

[0074] S111. Vibration acceleration sensors are installed in the drive motor housing, gearbox housing, main drive bearing housing, and handrail belt bearing installation positions of the escalator to collect the vibration acceleration time series a(t) during the escalator operation in real time.

[0075] S112. By installing a current sensor in the power supply circuit or inside the control cabinet of the escalator drive motor, the operating temperature time series T(t) of the drive motor during the escalator operation is collected in real time.

[0076] S113. By installing a current sensor in the power supply circuit or inside the control cabinet of the escalator drive motor, the operating current time sequence I(t) of the drive motor during the escalator operation is collected in real time.

[0077] S114. By installing speed sensors and displacement sensors at the step chain drive part, main drive shaft or step running track position of the escalator, the speed time sequence v(t) generated by the step motion during the operation of the escalator is collected in real time.

[0078] S12. By configuring a unified clock module or high-precision system clock in the data acquisition and control device of the escalator, the acquisition process of vibration sensors, temperature sensors, current sensors, and speed or displacement sensors installed in different locations is synchronously triggered or synchronously sampled and controlled. While various sensors are acquiring data, the corresponding acquisition time identifier is recorded for each sampled data; the acquisition time identifier is organized into a timestamp sequence t.

[0079] S13. The collected vibration acceleration time series a(t), operating temperature time series T(t), operating current time series I(t), velocity or displacement time series v(t), and corresponding timestamp sequence t are uniformly packaged to form the original multi-source operation dataset of the escalator.

[0080] In this embodiment, by deploying multiple types of sensors in the key mechanical and electrical parts of the escalator and using a unified clock for synchronous acquisition and timestamp encapsulation, the present invention can obtain multi-source operating data with consistent time and complete dimensions under complex operating conditions, effectively avoiding the problems of multi-channel data misalignment and information fragmentation, and providing a reliable and alignable data foundation for subsequent operating status analysis and anomaly identification.

[0081] Example 3

[0082] This embodiment is an explanation based on Embodiment 2. Please refer to it. Figure 1 Specifically, step two includes:

[0083] S21. Based on the original multi-source operation dataset of the escalator, timestamp alignment and interpolation repair technology is used to uniformly correct and complete the sampling offset, time misalignment and local data loss of different acquisition channels during operation; filtering and smoothing technology is used to suppress high-frequency interference in vibration signals, transient fluctuations in current signals and measurement noise in temperature and speed data; and a stable operation data sequence is established.

[0084] S22. During the continuous operation of the escalator, the stable operation data sequence is accumulated and stored over multiple operating cycles to form a historical stable operation dataset. Based on the historical stable operation dataset, the periodic change pattern of multi-source operation data under normal operating conditions is analyzed using the time series periodic feature analysis method, and the operating cycle parameter W of the stable operation rhythm of the escalator is extracted. This parameter describes the cyclical pattern of multi-source operation data in the time dimension and serves as the basic parameter for subsequent periodic structure construction and anomaly detection analysis.

[0085] S23. Based on the operating cycle parameter W, a cycle alignment and time index mapping method is used to perform cycle segmentation and alignment processing on the stable operating data sequence in the time dimension, dividing the continuous operating data into multiple cycle segments of the same length according to the operating cycle parameter; a cross-cycle feature alignment and structured modeling method is used to uniformly describe the changing relationship of operating data at the same cycle position of different acquisition channels, and construct a shared cyclic cycle matrix Q; used to characterize the correlation and change characteristics of multi-source operating data at the cycle scale;

[0086] S24. Based on the shared cyclic periodic matrix Q and the stable operating data sequence, and combined with the index position of the stable operating data on the time axis, a periodic position mapping and periodic component instantiation extraction method is used to perform periodic matching processing on the operating data of each acquisition channel in the stable operating data sequence. During the processing, according to the relative position of the operating data in its respective periodic segment, the periodic structure information corresponding to the current position is extracted from the shared cyclic periodic matrix and mapped to each acquisition channel to form periodic component parameters. It is used to characterize the normal periodic operation behavior of escalators at the current point in time; to describe the regular changes in multi-source operation data caused by the normal periodic operation of escalators, and to serve as the basic parameters for subsequent periodic influence stripping and abnormal feature construction.

[0087] S25. Based on the stable operation data sequence and periodic component parameters, the periodic influence stripping and operation state decomposition method is adopted to separate the changes caused by periodic operation behavior in the stable operation data; by removing the corresponding periodic component parameters from the stable operation data of each acquisition channel, a trend parameter Xtrend(t) is constructed to reflect the evolution characteristics of the escalator operation state over time; the remaining changes in the stable operation data that are not explained by the periodic components are organized into residual parameters Xres(t); the trend parameter is used to characterize the long-term state change characteristics of the key components of the escalator during continuous operation, and the residual parameter is used to characterize the instantaneous deviation of the operation state from the normal periodic operation behavior, and serves as the core input parameter for subsequent anomaly detection coefficient calculation, threshold determination and operation state assessment.

[0088] In this embodiment, by performing time alignment, noise suppression, and periodic structure modeling on multi-source operational data, and decomposing the operational status into periodic components, trend parameters, and residual parameters, the present invention can effectively remove the interference of normal escalator periodic operation on the data, highlighting the long-term evolution characteristics and instantaneous abnormal deviation characteristics, thereby significantly improving the accuracy and pertinence of subsequent anomaly detection and operational status assessment.

[0089] Example 4

[0090] This embodiment is an explanation based on Embodiment 3. Please refer to it. Figure 1 Specifically, step three includes:

[0091] S31, through the obtained periodic component parameters Using the periodic energy statistical analysis method with the operating cycle parameter W, and after dimensionless processing, the average value of the current periodic energy is obtained. Compared with historical cycle energy average The formula is as follows:

[0092]

[0093] In the formula, t0 represents the start time of the current running cycle. This represents the periodic component signal separated from the multi-source operating data of the i-th acquisition channel;

[0094]

[0095] In the formula, N represents the number of historical stable operating cycle windows included in the statistics, and tk represents the start time of the k-th historical stable operating cycle window.

[0096] In this embodiment, by performing energy statistics on the periodic component parameters and introducing the historical periodic energy average as a reference, the present invention can quantitatively characterize the degree of energy deviation between the current operating cycle and the normal cycle with a unified dimension, avoiding interference from single moment or instantaneous fluctuations on the judgment results, thereby improving the objectivity and reliability of the escalator periodic stability assessment.

[0097] Example 5

[0098] This embodiment is an explanation based on Embodiment 4. Please refer to it. Figure 1 Specifically, step three also includes:

[0099] S32, by obtaining the average energy of the current cycle Compared with historical cycle energy average After dimensionless processing, the periodic stability deviation coefficient ZPX is calculated using the following formula:

[0100]

[0101] S33. By setting a preset periodic stability threshold Zth, and comparing the periodic stability deviation coefficient ZPX with the periodic stability threshold Zth, the first evaluation result is obtained, including:

[0102] When the periodic stability deviation coefficient ZPX ≤ periodic stability threshold Zth, it indicates that the periodic structure of the escalator is qualified and continuous monitoring is required.

[0103] When the periodic stability deviation coefficient ZPX > the periodic stability threshold Zth, it indicates that the escalator's periodic structure is substandard, with abnormal deviations in periodic energy distribution. This poses a risk of periodic rhythm disorder, load fluctuations, or mechanical wear evolution, triggering the first warning instruction and generating the first strategy: reducing the periodic component parameters by 15% to 30%. The participation rate in relevant statistical calculations is reduced to minimize the impact of periodic changes caused by the normal periodic start-stop of escalators and load fluctuations on the anomaly judgment results, thereby generating periodic anomaly marker information.

[0104] The method for obtaining the periodic stability threshold Zth is as follows: By statistically analyzing a large amount of multi-source periodic operation data of escalators under normal operating conditions, the average periodic energy and its fluctuation range in different operating cycles are extracted to obtain the distribution range of the periodic stability deviation coefficient ZPX under periodic stable and abnormal states. Combining long-term escalator operation monitoring experience and the professional judgment of maintenance personnel, a reasonable critical judgment value for distinguishing between periodic structural stability and periodic structural abnormality is determined. At the same time, referring to relevant industry specifications, operation and maintenance standards and periodic operation characteristic parameters provided by equipment manufacturers, the periodic stability threshold Zth is comprehensively determined to effectively distinguish between normal periodic operation behavior and abnormal periodic rhythm risk states of escalators.

[0105] In this embodiment, by constructing a periodic stability deviation coefficient and combining it with a periodic anomaly marker and weight adjustment strategy, the present invention can effectively reduce the interference of normal start-stop and load periodic factors on the analysis results when the periodic structure experiences abnormal fluctuations. This makes the periodic anomaly identification more focused on the abnormal energy shift itself, thereby improving the accuracy and stability of identifying periodic rhythm disorders and early mechanical wear risks.

[0106] Example 6

[0107] This embodiment is an explanation based on Embodiment 5. Please refer to it. Figure 1 Specifically, step four includes:

[0108] S41. Based on the trend parameter Xtrend(t) and the residual parameter Xres(t), the trend parameter is used as a time series expression that truly reflects the long-term evolution state of the escalator during continuous operation, and the measured trend sequence Xsc(t) of the corresponding acquisition channel is constructed.

[0109] S42. Based on the trend parameter Xtrend(t) and residual parameter Xres(t), and combined with the historical stable operation dataset, the trend evolution characteristics of the corresponding acquisition channel in the historical stable operation data are used as input. The trend prediction model is used to model the change law of the trend residual sequence in the time dimension. By fitting and extrapolating the time evolution relationship of the trend residual sequence under the historical stable operation state, the trend change reference result corresponding to the current operation time period is obtained, and the predicted trend sequence Xyq(t) is formed.

[0110] In this embodiment, by constructing a measured trend sequence and a predicted trend sequence based on historical stable operating data respectively, the present invention can provide a comparable trend evolution reference for the operating status of escalators, enabling a quantitative comparison basis for long-term operating status changes, thereby enhancing the ability to characterize operating trend deviations and potential degradation risks.

[0111] Example 7

[0112] This embodiment is an explanation based on Embodiment 6. Please refer to it. Figure 1 Specifically, step four also includes:

[0113] S43. After dimensionless processing of the obtained measured trend sequence Xsc(t) and predicted trend sequence Xyq(t), the trend evolution deviation coefficient QYP is calculated using the following formula:

[0114]

[0115] In the formula, M represents the number of acquisition channels. This represents the measured trend sequence of the i-th acquisition channel. This represents the predicted trend sequence of the i-th acquisition channel;

[0116] S44. By setting a preset trend evolution threshold Qth, and comparing the trend evolution deviation coefficient QYP with the trend evolution threshold Qth, the second evaluation result is obtained, including:

[0117] When the trend evolution deviation coefficient QYP ≤ the trend evolution threshold Qth, it indicates that the escalator's operating trend is relatively stable and the evolution trajectory has not deviated, and there is no risk of non-periodic degradation. Continuous monitoring is required.

[0118] When the trend evolution deviation coefficient QYP > the trend evolution threshold Qth, it indicates that the relatively stable evolution trajectory of the escalator's operating trend has deviated, and there is a risk of non-periodic abnormal evolution caused by mechanical wear, electrical performance degradation, or long-term load changes. This triggers a second warning instruction and generates a second strategy: marking the current operating state as a potential degradation state; adjusting the discrimination weights of the trend parameter and the residual parameter, increasing the discrimination weight corresponding to the trend parameter by 10% to 25%, and simultaneously increasing the discrimination weight corresponding to the residual parameter by 5% to 15%, thereby enhancing the ability to identify non-periodic evolution anomalies and long-term degradation trends; and generating trend anomaly marking information.

[0119] The trend evolution threshold Qth is obtained by statistically modeling historical trend parameter data of escalators under long-term stable operation, analyzing the distribution characteristics of the deviation between the measured trend sequence and the predicted trend sequence, and extracting the typical value range of the trend evolution deviation coefficient QYP under stable and abnormal evolution states; combining the aging law of equipment, wear characteristics of key components and professional operation and maintenance experience, determining the critical judgment value used to distinguish between normal trend evolution and non-periodic degradation evolution; and referring to the escalator safety operation standards and the life and performance degradation indicators provided by the manufacturer, comprehensively forming the trend evolution threshold Qth.

[0120] In this embodiment, by introducing a trend evolution deviation coefficient and combining it with a preset threshold for unified quantitative comparison, the present invention can transform the stability of the escalator operation trend into a measurable index. When the trend deviates non-periodicly, the discrimination weight of the trend and residual is automatically strengthened, thereby improving the accuracy of identifying long-term degradation and gradual abnormal evolution.

[0121] Example 8

[0122] This embodiment is an explanation based on Embodiment 7. Please refer to it. Figure 1 Specifically, step five includes:

[0123] S51. Based on the measured trend sequence Xsc(t) and the predicted trend sequence Xyq(t) obtained in step four, the time-by-time difference calculation method is used to perform difference calculation on the measured trend data and the predicted trend data of the same acquisition channel at the same running time to obtain the residual change of the corresponding acquisition channel in the time dimension; and the residual change of each acquisition channel in the continuous operation process is sorted in time order to form the residual sequence r(t) of each acquisition channel.

[0124] In this embodiment, by performing time-by-time difference calculations on the measured trend sequence and the predicted trend sequence and constructing a residual sequence, the present invention can accurately characterize the instantaneous deviation features of each acquisition channel in the time dimension, providing stable and comparable basic data for subsequent multi-source residual consistency analysis.

[0125] Example 9

[0126] This embodiment is an explanation based on Embodiment 8. Please refer to it. Figure 1 Specifically, step five also includes:

[0127] S52. Based on the residual sequence r(t) corresponding to each acquisition channel, the correlation analysis method is used to quantitatively calculate the consistency of the changing trends between the residual sequences of any two different acquisition channels; by statistically summarizing the residual correlation results corresponding to all different acquisition channels, and after dimensionless processing, the multi-source residual consistency coefficient DCY is calculated, as follows:

[0128]

[0129] In the formula, M represents the number of acquisition channels. This represents the residual sequence corresponding to the i-th acquisition channel. This represents the residual sequence corresponding to the j-th acquisition channel. This represents the residual correlation function, which calculates the correlation between the residual sequences of the i-th acquisition channel and the j-th acquisition channel in the time dimension, reflecting whether the direction of change of the residuals of different acquisition channels is consistent and whether the trend of change is synchronous.

[0130] S53. By setting a residual consistency threshold Dth, and comparing the multi-source residual consistency coefficient DCY with the residual consistency threshold Dth, the third evaluation results are obtained, including:

[0131] When the multi-source residual consistency coefficient DCY < residual consistency threshold Dth, it indicates that the multi-source residual consistency is qualified and should be continuously monitored.

[0132] When the multi-source residual consistency coefficient DCY ≥ residual consistency threshold Dth, it indicates that the multi-source residual consistency is unqualified. The direction and trend of residual changes in multiple acquisition channels are consistent, and the abnormal deviation shows cross-channel synchronous evolution characteristics. There is a systemic operational risk caused by structural wear accumulation, system performance degradation, or overall operating condition deterioration. This triggers the third early warning instruction and generates the third strategy: determine the current escalator operating status as a systemic abnormal state; confirm that the abnormal impact has evolved from local disturbance to cross-channel consistency deviation, with the risk of persistence and spread; and generate systemic abnormality confirmation mark information.

[0133] The residual consistency threshold Dth is obtained as follows: Correlation statistical analysis is performed on the multi-source residual sequences of the escalator under the condition of no systematic anomaly. The consistency distribution interval of the residual changes of each acquisition channel under normal and systematic anomaly states is extracted to obtain the typical variation range of the multi-source residual consistency coefficient DCY. Combining the structural coupling characteristics of the escalator, the system-level fault evolution mechanism, and the experience judgment of professional technicians, a critical judgment value for distinguishing between local random disturbances and cross-channel systematic anomalies is determined. At the same time, the residual consistency threshold Dth is comprehensively determined with reference to the escalator safety operation specifications and system health assessment standards to identify the systemic operational risk state.

[0134] S54. Based on the generated periodic anomaly marker information, trend anomaly marker information, and systemic anomaly confirmation marker information, the abnormal characteristics of the current operating status of the escalator are uniformly summarized; the anomaly marker information, along with the corresponding multi-source operating data, operating cycle parameters, and trend parameters, are stored in the historical operating database; the required periodic structure data and trend evolution reference data are analyzed to complete the closed-loop recording of the multi-source data anomaly detection process of the escalator.

[0135] In this embodiment, by performing correlation analysis on the residual sequences of multiple acquisition channels and constructing a multi-source residual consistency coefficient, the present invention can identify the synchronous evolution characteristics of abnormal deviations in different channels, realize the effective confirmation of the transition from local anomalies to systemic anomalies, and complete the closed-loop recording of historical data by combining anomaly marking information, providing a reliable basis for the overall operational risk assessment and subsequent analysis of escalators.

[0136] Example 10

[0137] A multi-source data anomaly detection system for escalators, please refer to... Figure 2 Specifically, including:

[0138] The multi-source operation data acquisition module is used to monitor the motion status of pedestrian overpass escalators in real time; it synchronously collects and records timestamps, and encapsulates the multi-source time series of vibration acceleration, operating temperature, operating current and speed with timestamps to construct the original multi-source operation dataset of the escalator;

[0139] The data preprocessing module is used to perform time correction, missing data repair, and noise suppression on the original multi-source operation dataset of the escalator to form a stable operation data sequence. Based on multi-cycle cumulative analysis, the periodic parameters of the escalator's normal operation are extracted and a shared cyclic periodic structure is constructed, thereby separating the periodic component parameters, trend parameters, and residual parameters of each channel.

[0140] The periodic stability assessment module is used to obtain the average energy of the current period. Compared with historical cycle energy average The periodic stability deviation coefficient ZPX is calculated and compared with the periodic stability threshold Zth to determine whether the periodic structure of the escalator is qualified. If it is not qualified, the first strategy is applied to generate periodic anomaly marker information.

[0141] The trend evolution deviation assessment module is used to obtain the measured trend sequence Xsc(t) and the predicted trend sequence Xyq(t), calculate the trend evolution deviation coefficient QYP, and compare it with the trend evolution threshold Qth to determine whether the escalator's running trend has deviated from the relatively stable evolution trajectory. If it deviates, a second strategy is given to generate trend anomaly marker information.

[0142] The multi-source residual consistency assessment and recording module is used to acquire the residual sequence r(t) corresponding to each acquisition channel, calculate the multi-source residual consistency coefficient DCY, and compare it with the residual consistency threshold Dth to determine whether the multi-source residual consistency is qualified. If it is not qualified, a third strategy is adopted to generate systematic anomaly confirmation marker information. Combining periodic anomaly marker information and trend anomaly marker information, the abnormal characteristics of the current operating status of the escalator are uniformly summarized and stored together with the corresponding multi-source operating data and feature parameters to complete the closed-loop recording of escalator anomaly detection results.

[0143] In this embodiment, by modularly and collaboratively designing multi-source operation data acquisition, periodic stability assessment, trend evolution deviation assessment, and multi-source residual consistency analysis, the present invention achieves hierarchical identification and unified summary recording of escalator operation status from periodic structure and long-term trends to cross-channel systemic anomalies, thereby improving the comprehensiveness of anomaly detection, the consistency of judgment results, and the reliability of operation status assessment.

[0144] The threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by those skilled in the art for each set of sample data; as long as it does not affect the ratio between the parameter and the quantized value, it is acceptable.

[0145] The above formulas are all derived from software simulation using a large amount of data and are selected to be close to the actual values. The coefficients in the formulas are set by those skilled in the art according to the actual situation. The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any equivalent substitutions or changes made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the protection scope of the present invention.

Claims

1. A method for detecting anomalies in multi-source data of escalators, characterized in that, Includes the following steps: Step 1: Monitor the movement status of the pedestrian overpass escalator in real time; Synchronously collect and record timestamps, and encapsulate the multi-source time series of vibration acceleration, operating temperature, operating current and speed with timestamps to construct the original multi-source operating dataset of the escalator; Step 2: By performing time correction, missing data repair, and noise suppression on the original multi-source operation dataset of the escalator, a stable operation data sequence is formed; based on multi-cycle cumulative analysis, the cycle parameters of normal escalator operation are extracted and a shared cyclic cycle structure is constructed, thereby separating the cycle component parameters, trend parameters, and residual parameters of each channel; Step 3: Obtain the average energy value of the current cycle. Compared with historical cycle energy average The periodic stability deviation coefficient ZPX is calculated and compared with the periodic stability threshold Zth to determine whether the periodic structure of the escalator is qualified. If it is not qualified, the first strategy is applied to generate periodic anomaly marker information. Step 4: By obtaining the measured trend sequence Xsc(t) and the predicted trend sequence Xyq(t), calculate the trend evolution deviation coefficient QYP and compare it with the trend evolution threshold Qth to determine whether the relatively stable evolution trajectory of the escalator's running trend has deviated. If it deviates, a second strategy is given to generate trend anomaly marker information. Step 5: By acquiring the residual sequence r(t) corresponding to each acquisition channel, calculate the multi-source residual consistency coefficient DCY and compare it with the residual consistency threshold Dth to determine whether the multi-source residual consistency is qualified. If it is not qualified, the third strategy is applied to generate systematic anomaly confirmation marker information. Combining the periodic anomaly marker information and the trend anomaly marker information, the abnormal characteristics of the current operating status of the escalator are uniformly summarized and stored together with the corresponding multi-source operating data and feature parameters to complete the closed-loop recording of the escalator anomaly detection results.

2. The method for detecting multi-source data anomalies in escalators according to claim 1, characterized in that, Step one includes: S11. By monitoring the mechanical, electrical, and motion states of escalators operating in pedestrian overpasses under conditions of continuous start-stop, high load, long-term operation, and significant fluctuations in passenger flow, and collecting basic operating data of the escalators, including the following steps: S111. Vibration acceleration sensors are installed in the drive motor housing, gearbox housing, main drive bearing housing, and handrail belt bearing installation positions of the escalator to collect the vibration acceleration time series a(t) during the escalator operation in real time. S112. By installing a current sensor in the power supply circuit or inside the control cabinet of the escalator drive motor, the operating temperature time series T(t) of the drive motor during the escalator operation is collected in real time. S113. By installing a current sensor in the power supply circuit or inside the control cabinet of the escalator drive motor, the operating current time sequence I(t) of the drive motor during the escalator operation is collected in real time. S114. By installing speed sensors and displacement sensors at the step chain drive part, main drive shaft or step running track position of the escalator, the speed time sequence v(t) generated by the step motion during the operation of the escalator is collected in real time. S12. By configuring a unified clock module or high-precision system clock in the data acquisition and control device of the escalator, the acquisition process of vibration sensors, temperature sensors, current sensors, and speed or displacement sensors installed in different locations is synchronously triggered or synchronously sampled and controlled. While various sensors are acquiring data, the corresponding acquisition time identifier is recorded for each sampled data; the acquisition time identifier is organized into a timestamp sequence t. S13. The collected vibration acceleration time series a(t), operating temperature time series T(t), operating current time series I(t), velocity or displacement time series v(t), and corresponding timestamp sequence t are uniformly packaged to form the original multi-source operation dataset of the escalator.

3. The method for detecting multi-source data anomalies in escalators according to claim 2, characterized in that, Step two includes: S21. Based on the original multi-source operation dataset of the escalator, timestamp alignment and interpolation repair technology is used to uniformly correct and complete the sampling offset, time misalignment and local data loss of different acquisition channels during operation; filtering and smoothing technology is used to suppress high-frequency interference in vibration signals, transient fluctuations in current signals and measurement noise in temperature and speed data; and a stable operation data sequence is established. S22. During the continuous operation of the escalator, the stable operation data sequence is accumulated and stored over multiple operating cycles to form a historical stable operation dataset. Based on the historical stable operation dataset, the periodic change pattern of multi-source operation data under normal operating conditions is analyzed using the time series periodic feature analysis method, and the operating cycle parameter W of the stable operation rhythm of the escalator is extracted. S23. Based on the running cycle parameter W, the period alignment and time index mapping method is used to perform period segmentation and alignment processing on the stable running data sequence in the time dimension, and the continuous running data is divided into multiple period segments of the same length according to the running cycle parameter; the cross-cycle feature alignment and structured modeling method is used to uniformly describe the changing relationship of running data at the same period position of different acquisition channels, and construct a shared cyclic period matrix Q. S24. Based on the shared cyclic periodic matrix Q and the stable operating data sequence, and combined with the index position of the stable operating data on the time axis, a periodic position mapping and periodic component instantiation extraction method is used to perform periodic matching processing on the operating data of each acquisition channel in the stable operating data sequence. During the processing, according to the relative position of the operating data in its respective periodic segment, the periodic structure information corresponding to the current position is extracted from the shared cyclic periodic matrix and mapped to each acquisition channel to form periodic component parameters. ; S25. Based on the stable operation data sequence and periodic component parameters, the periodic influence stripping and operation state decomposition method is adopted to separate the changes caused by periodic operation behavior in the stable operation data; by removing the corresponding periodic component parameters from the stable operation data of each acquisition channel, the trend parameter Xtrend(t) is constructed; the remaining changes in the stable operation data that are not explained by the periodic components are organized into the residual parameter Xres(t).

4. The method for detecting anomalies in multi-source data of escalators according to claim 3, characterized in that, Step three includes: S31, through the obtained periodic component parameters Using the periodic energy statistical analysis method with the operating cycle parameter W, and after dimensionless processing, the average value of the current periodic energy is obtained. Compared with historical cycle energy average .

5. The method for detecting multi-source data anomalies in escalators according to claim 4, characterized in that, Step three also includes: S32, by obtaining the average energy of the current cycle Compared with historical cycle energy average After dimensionless processing, the periodic stability deviation coefficient ZPX is calculated and obtained. S33. By setting a preset periodic stability threshold Zth, and comparing the periodic stability deviation coefficient ZPX with the periodic stability threshold Zth, the first evaluation result is obtained, including: When the periodic stability deviation coefficient ZPX ≤ periodic stability threshold Zth, it indicates that the periodic structure of the escalator is qualified and continuous monitoring is required. When the periodic stability deviation coefficient ZPX > the periodic stability threshold Zth, it indicates that the escalator's periodic structure is substandard, with abnormal deviations in periodic energy distribution. This poses a risk of periodic rhythm disorder, load fluctuations, or mechanical wear evolution, triggering the first warning instruction and generating the first strategy: reducing the periodic component parameters by 15% to 30%. The participation rate in relevant statistical calculations is reduced to minimize the impact of periodic changes caused by the normal periodic start-stop of escalators and load fluctuations on the anomaly judgment results, thereby generating periodic anomaly marker information.

6. The method for detecting multi-source data anomalies in escalators according to claim 5, characterized in that, Step four includes: S41. Based on the trend parameter Xtrend(t) and the residual parameter Xres(t), the trend parameter is used as a time series expression that truly reflects the long-term evolution state of the escalator during continuous operation, and the measured trend sequence Xsc(t) of the corresponding acquisition channel is constructed. S42. Based on the trend parameter Xtrend(t) and residual parameter Xres(t), and combined with the historical stable operation dataset, the trend evolution characteristics of the corresponding acquisition channel in the historical stable operation data are used as input. The trend prediction model is used to model the change law of the trend residual sequence in the time dimension. By fitting and extrapolating the time evolution relationship of the trend residual sequence under the historical stable operation state, the trend change reference result corresponding to the current operation time period is obtained, and the predicted trend sequence Xyq(t) is formed.

7. The method for detecting multi-source data anomalies in escalators according to claim 6, characterized in that, Step four also includes: S43. After dimensionless processing of the obtained measured trend sequence Xsc(t) and predicted trend sequence Xyq(t), the trend evolution deviation coefficient QYP is calculated. S44. By setting a preset trend evolution threshold Qth, and comparing the trend evolution deviation coefficient QYP with the trend evolution threshold Qth, the second evaluation result is obtained, including: When the trend evolution deviation coefficient QYP ≤ the trend evolution threshold Qth, it indicates that the escalator's operating trend is relatively stable and the evolution trajectory has not deviated, and there is no risk of non-periodic degradation. Continuous monitoring is required. When the trend evolution deviation coefficient QYP > the trend evolution threshold Qth, it indicates that the relatively stable evolution trajectory of the escalator's operating trend has deviated, and there is a risk of non-periodic abnormal evolution caused by mechanical wear, electrical performance degradation, or long-term load changes. This triggers a second warning instruction and generates a second strategy: marking the current operating state as a potential degradation state; adjusting the discrimination weights of the trend parameter and the residual parameter, increasing the discrimination weight corresponding to the trend parameter by 10% to 25%, and simultaneously increasing the discrimination weight corresponding to the residual parameter by 5% to 15%, thereby enhancing the ability to identify non-periodic evolution anomalies and long-term degradation trends; and generating trend anomaly marking information.

8. The method for detecting anomalies in multi-source data of escalators according to claim 7, characterized in that, Step five includes: S51. Based on the measured trend sequence Xsc(t) and the predicted trend sequence Xyq(t) obtained in step four, the time-by-time difference calculation method is used to perform difference calculation on the measured trend data and the predicted trend data of the same acquisition channel at the same running time to obtain the residual change of the corresponding acquisition channel in the time dimension; and the residual change of each acquisition channel in the continuous operation process is sorted in time order to form the residual sequence r(t) of each acquisition channel.

9. The method for detecting multi-source data anomalies in escalators according to claim 8, characterized in that, Step five also includes: S52. Based on the residual sequence r(t) corresponding to each acquisition channel, the correlation analysis method is used to quantitatively calculate the consistency of the changing trends between the residual sequences of any two different acquisition channels; by statistically summarizing the residual correlation results corresponding to all different acquisition channels, and after dimensionless processing, the multi-source residual consistency coefficient DCY is calculated. S53. By setting a residual consistency threshold Dth, and comparing the multi-source residual consistency coefficient DCY with the residual consistency threshold Dth, the third evaluation results are obtained, including: When the multi-source residual consistency coefficient DCY < residual consistency threshold Dth, it indicates that the multi-source residual consistency is qualified and should be continuously monitored. When the multi-source residual consistency coefficient DCY ≥ residual consistency threshold Dth, it indicates that the multi-source residual consistency is unqualified. The direction and trend of residual changes in multiple acquisition channels are consistent, and the abnormal deviation shows cross-channel synchronous evolution characteristics. There is a systemic operational risk caused by structural wear accumulation, system performance degradation, or overall operating condition deterioration. This triggers the third early warning instruction and generates the third strategy: determine the current escalator operating status as a systemic abnormal state; confirm that the abnormal impact has evolved from local disturbance to cross-channel consistency deviation, with the risk of persistence and spread; and generate systemic abnormality confirmation mark information. S54. Based on the generated periodic anomaly marker information, trend anomaly marker information, and systemic anomaly confirmation marker information, the abnormal characteristics of the current operating status of the escalator are uniformly summarized; the anomaly marker information, along with the corresponding multi-source operating data, operating cycle parameters, and trend parameters, are stored in the historical operating database; the required periodic structure data and trend evolution reference data are analyzed to complete the closed-loop recording of the multi-source data anomaly detection process of the escalator.

10. A multi-source data anomaly detection system for escalators, applied to the multi-source data anomaly detection method for escalators as described in any one of claims 1 to 9, characterized in that, include: A multi-source operation data acquisition module is used to monitor the movement status of pedestrian overpass escalators in real time; Synchronously collect and record timestamps, and encapsulate the multi-source time series of vibration acceleration, operating temperature, operating current and speed with timestamps to construct the original multi-source operating dataset of the escalator; The data preprocessing module is used to perform time correction, missing data repair, and noise suppression on the original multi-source operation dataset of the escalator to form a stable operation data sequence. Based on multi-cycle cumulative analysis, the periodic parameters of the escalator's normal operation are extracted and a shared cyclic periodic structure is constructed, thereby separating the periodic component parameters, trend parameters, and residual parameters of each channel. The periodic stability assessment module is used to obtain the average energy of the current period. Compared with historical cycle energy average The periodic stability deviation coefficient ZPX is calculated and compared with the periodic stability threshold Zth to determine whether the periodic structure of the escalator is qualified. If it is not qualified, the first strategy is applied to generate periodic anomaly marker information. The trend evolution deviation assessment module is used to obtain the measured trend sequence Xsc(t) and the predicted trend sequence Xyq(t), calculate the trend evolution deviation coefficient QYP, and compare it with the trend evolution threshold Qth to determine whether the escalator's running trend has deviated from the relatively stable evolution trajectory. If it deviates, a second strategy is given to generate trend anomaly marker information. The multi-source residual consistency assessment and recording module is used to acquire the residual sequence r(t) corresponding to each acquisition channel, calculate the multi-source residual consistency coefficient DCY, and compare it with the residual consistency threshold Dth to determine whether the multi-source residual consistency is qualified. If it is not qualified, a third strategy is adopted to generate systematic anomaly confirmation marker information. Combining periodic anomaly marker information and trend anomaly marker information, the abnormal characteristics of the current operating status of the escalator are uniformly summarized and stored together with the corresponding multi-source operating data and feature parameters to complete the closed-loop recording of escalator anomaly detection results.

Images