Elevator operation state real-time monitoring method and system based on internet of things

Abstract

The application discloses an elevator operation state real-time monitoring method and system based on Internet of Things, and the content comprises monitoring node registration, operation parameter collection, residual error propagation modeling, working condition decoding and alarm output. The application relates to the technical field of elevator Internet of Things monitoring, and particularly relates to an elevator operation state real-time monitoring method and system based on Internet of Things. In the scheme, a monitoring node identity mapping and trusted evaluation mechanism is constructed, an elevator monitoring topology graph is established in combination with spatial position, time delay and link quality, stable anomaly detection is realized through multi-baseline residual error analysis, topology propagation and hysteresis determination, and further, accurate fault positioning and hierarchical alarm are realized in combination with working condition identification, fault probability correction and risk fusion, so that the accuracy, reliability and intelligent level of elevator state monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of elevator Internet of Things (IoT) monitoring technology, specifically to a method and system for real-time monitoring of elevator operating status based on the Internet of Things. Background Technology

[0002] With the rapid development of IoT technology and smart building construction, elevators, as an important vertical transportation tool in high-rise buildings, are receiving increasing attention for their operational safety and stability. Deploying sensor nodes in the car, landings, machine room, and control cabinet to collect real-time operational parameters such as door operator status, door lock status, car speed, traction current, braking temperature, vibration amplitude, and leveling error, and uploading this data to a monitoring platform using IoT communication technology, has become the main technical approach for elevator operational status monitoring. With the further integration of edge computing, big data analytics, and artificial intelligence technologies, elevator operational status monitoring is expected to develop towards multi-source data fusion, early fault warning, and intelligent operation and maintenance, thereby improving elevator operational safety and maintenance efficiency.

[0003] However, in existing technologies, the data collected by multiple sensor nodes lacks a unified identification and credibility assessment mechanism. The data sources of nodes are chaotic, and reliable nodes are mixed with low-quality nodes, resulting in low reliability of subsequent analysis results. At the same time, most existing methods use a single threshold or isolated features for anomaly judgment, failing to fully consider the correlation between various functional components of the elevator and the fault propagation law, which easily leads to inaccurate anomaly location and frequent false alarms and missed alarms. In addition, operating condition identification and fault diagnosis are often carried out independently, lacking operating condition perception capabilities, which can easily misjudge normal operating condition switching as a fault. It is also difficult to comprehensively quantify and assess the fault type, fault location, and risk level, resulting in alarm information that lacks specificity and is not conducive to timely and accurate fault handling by operation and maintenance personnel. Summary of the Invention

[0004] To address the above issues and overcome the shortcomings of existing technologies, this invention provides a method and system for real-time monitoring of elevator operation status based on the Internet of Things (IoT). Addressing the problems in existing elevator IoT monitoring technologies, such as the lack of unified identification and credibility assessment mechanisms for data sources from multiple sensor nodes, and the difficulty in tracing specific nodes and locations of abnormal data, this solution generates a unique identifier for each monitoring node and establishes a mapping relationship between it and the elevator and its installation location. Simultaneously, it calculates node credibility and filters valid monitoring targets by combining signal strength, reporting continuity, and location consistency. Furthermore, it constructs an elevator monitoring topology map based on spatial distance, latency consistency, and link quality, thereby achieving traceability of monitoring data sources and accurate construction of the monitoring topology. This solution also addresses the shortcomings of existing elevator operation status monitoring methods, which mostly rely on single thresholds or isolated local features to determine anomalies, failing to reflect the propagation and diffusion patterns of faults among various functional components of the elevator, leading to anomalies. To address the issues of inaccurate positioning and numerous false alarms and missed alarms, this solution jointly calculates the comprehensive residual by establishing short-term window baselines, long-term operating baselines, and mode baselines corresponding to different operating conditions. It then uses a monitoring topology map to perform graph propagation of the residuals to form a residual heatmap, enabling local anomalies to spread along the elevator's functional topology. This is further addressed by combining time recursion and a dual-threshold hysteresis mechanism to output the abnormal status. Furthermore, to address the problems in existing elevator fault diagnosis technologies, such as the independence of operating condition identification and fault diagnosis, the tendency to misjudge normal operating condition transitions as faults, and the difficulty in comprehensively quantifying and assessing fault types, fault locations, and risk levels, this solution decodes the current operating condition using a state fingerprint vector. During fault source decomposition, it introduces an operating condition compatibility coefficient to suppress and correct fault categories that do not conform to the current operating condition. Simultaneously, it integrates abnormal evidence values, fault probabilities, and operating condition inconsistencies to generate a comprehensive risk score and risk level.

[0005] The technical solution adopted in this invention is as follows: a method for real-time monitoring of elevator operating status based on the Internet of Things, the method comprising the following steps:

[0006] Step S1: Register monitoring nodes. Generate a unique identifier for each node participating in the monitoring and establish its mapping relationship with the elevator and installation location to form a node object table. At the same time, jointly evaluate the signal strength, reporting continuity and location consistency of each node to calculate the credibility and screen out the effective monitoring nodes. Then, based on the spatial location relationship of the nodes, communication delay characteristics and link quality indicators, construct the elevator monitoring topology map of the screened nodes.

[0007] Step S2: Acquire operating parameters. Collect multi-source operating parameters such as door operator opening and closing status, door lock status, car speed, traction current, and braking temperature to form the original observation sequence. Combine time offset, link quality, and node reliability to generate fusion weights. Based on the state change trend in historical continuous sampling data, recursively compensate for missing or abnormally deviated observations within the sampling window, and then calculate the synchronous sampling value according to the fusion weights.

[0008] Step S3: Residual propagation modeling, robustly standardize the synchronous sampled values ​​and compress extreme fluctuations to generate basic state features; combine short-time difference, long-time difference and cross-parameter coupling terms to form a state fingerprint vector; establish short-time window baseline, long-term operating baseline and mode baseline corresponding to different operating conditions, calculate the deviation between the state fingerprint and the baseline to obtain the comprehensive residual; use the monitoring topology map to propagate the residual to form a residual heat map; then perform time recursion on the residual to obtain anomaly evidence values, and output the abnormal state through a double threshold hysteresis method;

[0009] Step S4: Operating condition decoding. Input the current and previous time-state fingerprint vectors into the operating condition decoding model and output the operating condition category label; combine the topology propagation residual and the residual change rate to decompose the fault source of the abnormal residual, and introduce the operating condition compatibility coefficient to correct the fault category probability; fuse the abnormal evidence value, fault probability and operating condition inconsistency term to generate a risk score and map it to a risk level; select the fault category with the highest corrected probability as the main fault category, and generate an alarm direction based on the topology cluster location;

[0010] Step S5: Alarm output. When the risk level reaches the preset conditions, the elevator number, node identifier, monitoring time, working condition label, main fault category, risk level, fault location and status fingerprint are encapsulated into an alarm event package. Based on the results of manual inspection or subsequent confirmation, the short window baseline, long window baseline, abnormal threshold, working condition fault compatibility relationship and node credibility are incrementally updated.

[0011] Further, in step S1, the registration of the monitoring node specifically includes the following steps:

[0012] Step S11: Generate a unique identifier for the monitoring node and establish a location mapping. For each node participating in the monitoring, read its physical address, elevator number, installation location code, node type and power-on timestamp, and concatenate this information and generate a unique identifier for the node through hash operation. Then, establish a mapping relationship between the unique identifier and the corresponding elevator equipment and installation location to form a node object table, so that all data collected later can be traced back to the specific node and specific location.

[0013] Step S12: Calculate node credibility and screen valid monitoring objects. Jointly evaluate the reception strength, reporting continuity and location consistency of each node during the registration stage to form node credibility. Include nodes with credibility higher than the preset threshold in the formal monitoring object table, and set nodes with insufficient credibility to a pending review status, thereby limiting the data sources that subsequently enter the algorithm link.

[0014] Step S13: Construct an elevator monitoring topology map. The nodes that have passed the trust screening are constructed into an elevator monitoring topology map according to their installation location, functional association and communication link relationship. The adjacency weight between nodes takes into account spatial distance, time delay consistency and link quality, so that subsequent residual propagation and fault attribution can be carried out under topological constraints.

[0015] Further, in step S2, the acquisition of operating parameters specifically includes the following steps:

[0016] Step S21: Collect multi-source operating parameters. The car sensor node, landing sensor node, machine room sensor node and control cabinet sensor node respectively collect the door operator opening and closing status, door lock status, car speed, traction current, braking temperature, vibration amplitude, leveling error, communication delay, load status and floor stopping status, and form the original observation sequence according to a unified sampling period.

[0017] Step S22: Generate fusion weights. For each original observation, adjust the weights based on its time offset, link quality, and observation reliability to obtain the final fusion weights for participating in the synchronous calculation. The closer the time is to the center of the window, the more stable the link, and the more reliable the observation, the greater the weight.

[0018] Step S23: Output synchronous sampled values. For observations that are missing or distorted within the current sampling window, use the recursive compensation values ​​from nearby historical moments to replace them, and then calculate the synchronous sampled values ​​according to the fusion weights.

[0019] Furthermore, in step S3, the residual propagation modeling specifically includes the following steps:

[0020] Step S31: Generate basic state features. Perform robust standardization on the synchronous sampled values ​​according to the historical normal distribution, and use the hyperbolic tangent function to compress extreme fluctuations, so that the operating parameters of different dimensions are mapped to a unified feature space to form basic state features.

[0021] Step S32: Generate a state fingerprint vector, calculate the short-time difference, long-time difference, and cross-parameter coupling term of the synchronous sampled values, and concatenate them with the basic state features to form a state fingerprint vector. The state expression simultaneously includes instantaneous changes, trend changes, and linkage structures.

[0022] Step S33: Calculate the comprehensive residual. Establish short-window baseline, long-window baseline and mode baseline for the state fingerprint vector. The short-window baseline reflects the local normal structure of the most recent running segment, the long-window baseline reflects the long-term stable operating characteristics, and the mode baseline reflects the benchmark structure under different operating conditions. Then, the deviation between the current state and the three types of baselines is calculated as the comprehensive residual.

[0023] Step S34: Generate a residual heat map. Use the monitoring topology map established in step S1 to perform graph propagation on the comprehensive residuals to form a residual heat map, so that local anomalies can be diffused and analyzed along the elevator functional topology.

[0024] Step S35: Output the abnormal state. Further time-progression is performed on the integrated residual and the residual after topology propagation to form an abnormal evidence value. The abnormal state is output through a double threshold hysteresis method. When the abnormal evidence value is higher than the entry threshold, the abnormality is marked as valid. When the abnormal evidence value is lower than the exit threshold, the abnormality is marked as invalid. When it is between the two thresholds, the state of the previous time step remains unchanged.

[0025] Further, in step S4, the working condition decoding specifically includes the following steps:

[0026] Step S41: Decode the current operating condition based on the state fingerprint, input the state fingerprint vector and the state fingerprint of the previous moment into the operating condition decoding model, output the category probability of the current elevator operating condition, and take the category corresponding to the highest probability as the label of the current operating condition.

[0027] Step S42: Decompose the abnormal residuals into fault sources. Define the difference between the current state fingerprint and the short window baseline as the abnormal residual. Then, combine the residuals after topology propagation and the residual change rate into the fault source decomposition model to obtain the probability of fault categories such as door system faults, traction system faults, braking system faults, leveling system faults, sensor drift, and communication anomalies. At the same time, introduce the working condition compatibility coefficient to suppress fault categories that do not conform to the current working conditions.

[0028] Step S43: Generate a risk score by combining abnormal evidence values, maximum fault correction probability, and inconsistency items in operating conditions to form a risk score, and map the risk score to a five-level risk level; the higher the risk level, the more likely the fault is to affect the normal operation of the elevator and the higher the priority it needs to be handled.

[0029] Step S44: Output the main fault category. Select the fault category corresponding to the maximum value from the fault correction probability as the main fault category, and generate an alarm direction by combining the clustering location of the fault in the monitoring topology map.

[0030] Furthermore, in step S5, the alarm output specifically includes the following steps:

[0031] Step S51: Generate an alarm event package. When the risk level reaches the preset alarm conditions, the elevator number, node identifier, monitoring time, current working condition label, main fault category, risk level, fault location index and status fingerprint are encapsulated together into an alarm event package.

[0032] Step S52: Parameter update. When manual inspection or subsequent confirmation results determine that a certain period is a normal operation sample, the short window baseline, long window baseline and threshold parameters are incrementally updated using the status fingerprint of that period. When the confirmation result determines that it is a valid fault sample, the compatibility relationship between the operating condition and the fault is corrected, and the credibility of the corresponding node is adjusted accordingly.

[0033] The present invention provides an IoT-based real-time elevator operation status monitoring system, which includes a monitoring node registration module, an operation parameter acquisition module, a residual propagation modeling module, an operating condition decoding module, and an alarm output module.

[0034] The monitoring node registration module is used to generate unique identifiers for each participating node and establish a mapping relationship with elevator equipment and installation location to form a node object table; to evaluate the credibility of each node and select valid monitoring nodes; and to construct an elevator monitoring topology map based on node spatial location, latency consistency and link quality.

[0035] The operation parameter acquisition module is connected to the monitoring node registration module and is used to acquire multi-source operation parameters of the elevator through car sensor nodes, landing sensor nodes, machine room sensor nodes and control cabinet sensor nodes to form an original observation sequence; generate fusion weights by combining observation time offset, link quality and node reliability; and calculate synchronous sampling values ​​according to the fusion weights after compensating for missing or distorted observation values ​​in the sampling window.

[0036] The residual propagation modeling module is connected to the operating parameter acquisition module and is used to standardize the synchronous sampled values ​​to generate basic state features; combine differential features and cross-parameter coupling features with the basic state features to form a state fingerprint vector; establish short-window baselines, long-window baselines, and pattern baselines and calculate the deviation between the state fingerprint vector and each baseline to obtain the comprehensive residual; perform graph propagation on the comprehensive residual according to the elevator monitoring topology map to obtain a residual heat map; perform time recursion on the residual to obtain anomaly evidence values, and output the abnormal state through a double-threshold hysteresis method; the residual propagation modeling module includes a baseline construction unit and a graph propagation unit. The baseline construction unit is used to construct the short-window baseline, long-window baseline, and pattern baseline and calculate the comprehensive residual. The graph propagation unit is used to propagate the comprehensive residual according to the adjacency matrix of the elevator monitoring topology map to obtain the residual heat map.

[0037] The operating condition decoding module is connected to the residual propagation modeling module. It is used to input the state fingerprint vector into the operating condition decoding model to obtain the current elevator operating condition label; to decompose the abnormal residual into fault sources to obtain the probability of various faults, and to correct the probability by combining the operating condition compatibility coefficient to obtain the fault correction probability; to generate a risk score by fusing the abnormal evidence value, the fault correction probability, and the operating condition inconsistency item, and to map it to a risk level; to determine the main fault category based on the fault correction probability, and to determine the fault location by combining the elevator monitoring topology map.

[0038] The alarm output module is connected to the operating condition decoding module and is used to encapsulate the elevator number, node identifier, monitoring time, operating condition label, main fault category, risk level, fault location, and status fingerprint into an alarm event package and output it when the risk level reaches the preset alarm conditions. The alarm output module is also connected to the monitoring node registration module and the residual propagation modeling module respectively, and is used to update the baseline, threshold, compatibility relationship between operating conditions and faults, and node credibility based on the confirmation result of the alarm event package.

[0039] The beneficial effects achieved by the present invention using the above solution are as follows:

[0040] (1) In view of the problem that the data sources of multiple sensor nodes in the existing elevator Internet of Things monitoring technology lack a unified identity identification and credibility assessment mechanism, and that abnormal data is difficult to trace to specific nodes and locations, this solution generates a unique identifier for each monitoring node and establishes a mapping relationship between it and the elevator and its installation location. At the same time, it calculates the node credibility and selects effective monitoring objects by combining signal strength, reporting continuity and location consistency. Then, it constructs an elevator monitoring topology map based on spatial distance, time delay consistency and link quality, thereby realizing the traceability of monitoring data sources and the accurate construction of monitoring topology, providing a reliable spatial correlation basis for subsequent residual propagation and fault attribution.

[0041] (2) In view of the problems that most existing elevator operation status monitoring methods rely on a single threshold or isolated local features to judge anomalies, fail to reflect the propagation and diffusion law of faults among various functional components of the elevator, and result in inaccurate anomaly location and a large number of false alarms and missed alarms, this solution calculates the comprehensive residual by establishing a short time window baseline, a long-term operation baseline and a mode baseline corresponding to different operating conditions, and uses the monitoring topology map to perform graph propagation of the residual to form a residual heat map, so that local anomalies can be propagated along the elevator functional topology. Combined with time recursion and double threshold hysteresis mechanism, the abnormal status is output, thereby improving the accuracy and stability of anomaly detection and effectively suppressing false alarms and missed alarms caused by threshold jitter.

[0042] (3) In view of the problems that existing elevator fault diagnosis technologies have independent operating condition identification and fault diagnosis, are prone to misjudging normal operating condition switching as fault, and are difficult to comprehensively quantify and evaluate fault type, fault location and risk level, this solution decodes the current operating condition through state fingerprint vector, and introduces operating condition compatibility coefficient in the fault source decomposition process to suppress and correct fault categories that do not conform to the current operating condition. At the same time, it integrates abnormal evidence value, fault probability and operating condition inconsistency item to generate comprehensive risk score and risk level, thereby realizing fault source tracing and risk quantification under operating condition perception conditions, which significantly improves the accuracy of fault diagnosis and the pertinence of alarms. Attached Figure Description

[0043] Figure 1 A schematic diagram of the real-time monitoring method for elevator operation status based on the Internet of Things provided by the present invention;

[0044] Figure 2 A schematic diagram of the real-time monitoring system for elevator operation status based on the Internet of Things provided by the present invention.

[0045] Figure 3 A detailed flowchart for monitoring node registration in step S1;

[0046] Figure 4 A detailed flowchart for modeling residual propagation in step S3;

[0047] Figure 5 This is a diagram showing the elevator monitoring topology and residual heat propagation of the present invention.

[0048] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0049] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0050] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0051] Example 1, see Figure 1 The present invention provides a method for real-time monitoring of elevator operating status based on the Internet of Things, the method comprising the following steps:

[0052] Step S1: Register monitoring nodes. Generate a unique identifier for each node participating in the monitoring and establish its mapping relationship with the elevator and installation location to form a node object table. At the same time, jointly evaluate the signal strength, reporting continuity and location consistency of each node to calculate the credibility and screen out the effective monitoring nodes. Then, based on the spatial location relationship of the nodes, communication delay characteristics and link quality indicators, construct the elevator monitoring topology map of the screened nodes.

[0053] Step S2: Acquire operating parameters. Collect multi-source operating parameters such as door operator opening and closing status, door lock status, car speed, traction current, and braking temperature to form the original observation sequence. Combine time offset, link quality, and node reliability to generate fusion weights. Based on the state change trend in historical continuous sampling data, recursively compensate for missing or abnormally deviated observations within the sampling window, and then calculate the synchronous sampling value according to the fusion weights.

[0054] Step S3: Residual propagation modeling, robustly standardize the synchronous sampled values ​​and compress extreme fluctuations to generate basic state features; combine short-time difference, long-time difference and cross-parameter coupling terms to form a state fingerprint vector; establish short-time window baseline, long-term operating baseline and mode baseline corresponding to different operating conditions, calculate the deviation between the state fingerprint and the baseline to obtain the comprehensive residual; use the monitoring topology map to propagate the residual to form a residual heat map; then perform time recursion on the residual to obtain anomaly evidence values, and output the abnormal state through a double threshold hysteresis method;

[0055] Step S4: Operating condition decoding. Input the current and previous time-state fingerprint vectors into the operating condition decoding model and output the operating condition category label; combine the topology propagation residual and the residual change rate to decompose the fault source of the abnormal residual, and introduce the operating condition compatibility coefficient to correct the fault category probability; fuse the abnormal evidence value, fault probability and operating condition inconsistency term to generate a risk score and map it to a risk level; select the fault category with the highest corrected probability as the main fault category, and generate an alarm direction based on the topology cluster location;

[0056] Step S5: Alarm output. When the risk level reaches the preset conditions, the elevator number, node identifier, monitoring time, working condition label, main fault category, risk level, fault location and status fingerprint are encapsulated into an alarm event package. Based on the results of manual inspection or subsequent confirmation, the short window baseline, long window baseline, abnormal threshold, working condition fault compatibility relationship and node credibility are incrementally updated.

[0057] Example 2, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S1, the registration of the monitoring node specifically includes the following steps:

[0058] Step S11: Generate unique identifiers for monitoring nodes and establish location mappings. For each node participating in monitoring, read its physical address, elevator number, installation location code, node type, and power-on timestamp. Concatenate this information and perform a hash operation to generate a unique node identifier. Then, establish a mapping relationship between this unique identifier and the corresponding elevator equipment and installation location to form a node object table. This ensures that all subsequently collected data can be traced back to a specific node and location. The formula is as follows:

[0059] ;

[0060] in, Indicates the first A unique identifier for each node. Indicates the first The physical address of each node. This indicates the elevator number corresponding to this node. This indicates the installation location code of the node. Indicates the type of the node. Indicates the node's power-on timestamp. Represents the hash operation function. This indicates the truncation of the hash result before... One character, This indicates field concatenation operations; it is used to uniformly map node hardware information and spatial information into a stable monitoring identity.

[0061] Step S12: Calculate node credibility and filter valid monitoring targets. Perform a joint evaluation of each node's reception strength, reporting continuity, and location consistency during the registration phase to form node credibility. Nodes with credibility higher than a preset threshold are included in the formal monitoring target list, while nodes with insufficient credibility are placed in a pending review state. This limits the data sources entering the algorithm's subsequent path. The formula is as follows:

[0062] ;

[0063] in, Indicates the first The credibility of each node This indicates the current received signal strength of the node. and These represent the lower and upper boundaries of the received signal strength, respectively. This indicates the number of valid reports submitted by the node within the statistics window. This represents the theoretical upper limit of the number of valid reports. This represents the consistency score of a node's location. This indicates the upper limit of the positional consistency score. This indicates the clock synchronization deviation at that node. Indicates the maximum permissible synchronization deviation. , , and These represent the weight coefficients of the three evaluation factors, and This is used to transform whether a node can be included in the monitoring system into a quantifiable result;

[0064] Step S13: Construct an elevator monitoring topology. Nodes that have passed the trust screening are used to construct the elevator monitoring topology based on their installation location, functional association, and communication link relationships. The adjacency weights between nodes simultaneously consider spatial distance, latency consistency, and link quality, enabling subsequent residual propagation and fault attribution to be performed under topological constraints, as shown below:

[0065] ;

[0066] in, Represents a node With nodes Adjacency weights between them and These represent the credibility of the two nodes, respectively. This represents the spatial distance between two nodes. This indicates the difference in communication latency between two nodes. and These represent the distance attenuation coefficient and the time delay attenuation coefficient, respectively.

[0067] By performing the above operations, this solution addresses the problems in existing elevator IoT monitoring technologies, such as the lack of unified identification and credibility assessment mechanisms for data sources from multiple sensor nodes, and the difficulty in tracing specific nodes and locations of abnormal data. It generates a unique identifier for each monitoring node and establishes a mapping relationship between it and the elevator and its installation location. Simultaneously, it calculates node credibility and filters valid monitoring targets by combining signal strength, reporting continuity, and location consistency. Furthermore, it constructs an elevator monitoring topology map based on spatial distance, latency consistency, and link quality. This achieves traceability of monitoring data sources and accurate construction of the monitoring topology, providing a reliable spatial correlation foundation for subsequent residual propagation and fault attribution.

[0068] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, the acquisition of operating parameters specifically includes the following steps:

[0069] Step S21: Collect multi-source operating parameters. The car sensor node, landing sensor node, machine room sensor node, and control cabinet sensor node respectively collect the door operator opening and closing status, door lock status, car speed, traction current, braking temperature, vibration amplitude, leveling error, communication delay, load status, and floor stopping status, and form an original observation sequence according to a unified sampling period, as shown in the following formula:

[0070] ;

[0071] in, Indicates the first The sampling time of the first sampling moment The first class parameter One original observation value, Indicates the first The monitoring node at the first The original state quantity output at each sampling time. Indicates that for the first Observation mapping function for class parameters;

[0072] Step S22: Generate fusion weights. For each original observation, adjust the weights based on its time offset, link quality, and observation reliability to obtain the final fusion weights for synchronous computation. The closer the time is to the window center, the more stable the link, and the more reliable the observation, the larger the weight, as shown below:

[0073] ;

[0074] in, Indicates the first The timestamp of each original observation. Indicates the first The center timestamp of each sampling window Indicates the time decay coefficient. This represents the link confidence coefficient of the observed value. This indicates the confidence level of the node that outputs the observation. Indicates the first The node index corresponding to each observation value;

[0075] Step S23: Output synchronized sampled values. For missing or distorted observations within the current sampling window, replace them with recursive compensation values ​​from nearby historical times. Then, calculate the synchronized sampled values ​​according to the fusion weights, as shown below:

[0076] ;

[0077] in, Indicates the compensation after the first One observation value, Represents the historical recurrence coefficient. Indicates the first The sampling time of the first sampling moment Synchronized sampled values ​​of class parameters, Indicates the first The class parameter represents the number of valid observations within the current window. Indicates the first Within the sampling window, the first Class parameter number The fusion weights corresponding to each observation value This indicates that the observation is valid. This indicates a lack of observations. Indicates the first Time of the first Synchronized sampled values ​​of class parameters.

[0078] Example 4, see Figure 1 , Figure 4 and Figure 5 This embodiment is based on the above embodiment. In step S3, the residual propagation modeling specifically includes the following steps:

[0079] Step S31: Generate basic state features. Robustly standardize the synchronous sampled values ​​according to the historical normal distribution, and use the hyperbolic tangent function to compress extreme fluctuations, so that the operating parameters of different dimensions are mapped to a unified feature space to form basic state features. ;in, Indicates the first The sampling time of the first sampling moment One basic state characteristic, Indicates the first Synchronous sampling values, This represents the median of the parameter in historical normal samples. This indicates the absolute deviation of the parameter from the median in historical normal samples. This indicates the smallest positive number whose denominator is zero. Represents the hyperbolic tangent function;

[0080] Step S32: Generate a state fingerprint vector. Calculate the short-time difference, long-time difference, and cross-parameter coupling term of the synchronously sampled values. Concatenate these with the basic state features to form the state fingerprint vector. The state representation simultaneously includes instantaneous changes, trend changes, and linkage structures, as shown below: , , ;in, Indicates the first Short-time difference characteristics of class parameters, Indicates the first A cross-parameter coupling feature, and They respectively represent the construction of the first The two synchronously sampled values ​​corresponding to each coupling feature This represents the adjustment parameter used to suppress the amplification of the coupling term. They represent the first The parameter index corresponding to each coupling feature Indicates the first The state fingerprint vector at each sampling time. Indicates the number of basic state features. Indicates the number of coupling features;

[0081] Step S33: Calculate the comprehensive residual. Establish short-window baseline, long-window baseline, and mode baseline for the state fingerprint vector. The short-window baseline reflects the local normal structure of the most recent operating segment, the long-window baseline reflects the long-term stable operating characteristics, and the mode baseline reflects the baseline structure under different operating conditions. Then, calculate the comprehensive residual by combining the deviation of the current state from the three types of baselines, as follows: , , , ;in, Indicates the first The short-window baseline vector at each time step. Represents the long window baseline vector. Indicates the current operating condition The corresponding mode baseline vector, Indicates the short window recursion coefficient. This indicates the number of historical samples used for long-window aggregation. This indicates the number of samples used for pattern baseline aggregation. Indicates the first under the current working condition category A historical state fingerprint, Indicates the first The composite residual vector at each time step. , and These represent the weighting coefficients of the three types of baseline deviation terms, and the sum of the three is 1;

[0082] Step S34: Generate a residual heatmap. Using the monitoring topology map established in step S1, perform graph propagation on the comprehensive residuals to form a residual heatmap, enabling local anomalies to be analyzed and diffused along the elevator functional topology, as shown below: ;in, This represents the residual vector after topological propagation. This represents the adjacency matrix of the monitoring topology graph. Indicates the graph propagation smoothing coefficient. Represents the composite residual vector;

[0083] Step S35: Output the abnormal state. Further time-progress the integrated residual and the residual after topology propagation to form an anomaly evidence value. Output the abnormal state using a dual-threshold hysteresis method. When the anomaly evidence value is higher than the entry threshold, mark the anomaly as valid; when the anomaly evidence value is lower than the exit threshold, mark the anomaly as invalid; when it is between the two thresholds, maintain the previous state. This is represented as: ; ;in, Indicates the first The value of anomalous evidence at each moment. Represents the recursive smoothing coefficient. This indicates the abnormal state indicator at the current moment. This indicates the threshold for abnormal entry. Indicates the threshold for abnormal exit. This indicates the abnormal state at the previous moment.

[0084] By performing the above operations, this solution addresses the problems of existing elevator operation status monitoring methods that mostly rely on single thresholds or isolated local features to judge anomalies, fail to reflect the propagation and diffusion patterns of faults among various functional components of the elevator, and result in inaccurate anomaly localization and a large number of false alarms and missed alarms. This solution establishes a short-time window baseline, a long-term operating baseline, and mode baselines corresponding to different operating conditions to jointly calculate the comprehensive residual. It then uses the monitoring topology map to perform graph propagation of the residual to form a residual heat map, enabling local anomalies to be analyzed and diffused along the elevator's functional topology. Combined with time recursion and a dual-threshold hysteresis mechanism, the solution outputs the anomaly status, thereby improving the accuracy and stability of anomaly detection and effectively suppressing false alarms and missed alarms caused by threshold jitter.

[0085] Example 5, see Figure 1 This embodiment is based on the above embodiment. In step S4, the working condition decoding specifically includes the following steps:

[0086] Step S41: Decode the current operating condition based on the state fingerprint. Input the state fingerprint vector and its previous state fingerprint together into the operating condition decoding model, output the category probability of the current elevator operating condition, and take the category corresponding to the highest probability as the label of the current operating condition, represented as: ; ;in, Indicates the first At the [time]th moment The probability of a given operating condition. and They represent the first Current state discrimination parameters and historical state discrimination parameters for similar working conditions. Represents the fingerprint vector of the current state. This represents the fingerprint vector of the state at the previous time step. Indicates the total number of operating condition categories. Indicates the current operating condition label; identifies the specific operating condition of the elevator, such as starting, running, decelerating, leveling, opening, or closing the door.

[0087] Step S42: Decompose the abnormal residuals into fault sources. Define the difference between the current state fingerprint and the short-window baseline as the abnormal residual. Then, combine the residuals after topology propagation and the residual change rate into the fault source decomposition model to obtain the probabilities of fault categories such as door system faults, traction system faults, braking system faults, leveling system faults, sensor drift, and communication anomalies. At the same time, introduce a working condition compatibility coefficient to suppress fault categories that do not conform to the current working conditions, expressed as: ; ; ;in, Indicates the rate of change of abnormal residuals. Indicates the first At the [time]th moment The original probability of the type of fault, , and They represent the first The residual discrimination parameters, topological residual discrimination parameters, and rate of change discrimination parameters corresponding to the fault type. This represents the residual vector after topological propagation. Indicates the total number of fault categories. This represents the fault correction probability after introducing the operating condition compatibility coefficient. Indicates the current operating condition With the Compatibility coefficients between different types of faults;

[0088] Step S43: Generate a risk score by combining abnormal evidence values, maximum fault correction probability, and inconsistency items in operating conditions to form a risk score, which is then mapped to a five-level risk rating. A higher risk rating indicates a greater likelihood that the fault will affect the normal operation of the elevator, requiring higher priority processing, as shown below:

[0089] ;

[0090] ;

[0091] in, Indicates the first The overall risk score at each moment, Representation Logic function, , , and This represents the fusion weight of the abnormal evidence term, the failure probability term, the inconsistency in operating conditions term, and the residual magnitude term. This indicates the probability of the current operating condition corresponding to the current operating condition label. Indicates the risk level, with a value between 1 and 5 (integer). This indicates the floor function;

[0092] Step S44: Output the main fault category. Select the fault category corresponding to the maximum value from the fault correction probabilities as the main fault category, and generate an alarm direction by combining the clustering location of the fault in the monitoring topology map, as shown below:

[0093] ;

[0094] ;

[0095] in, Indicates the first Index of primary fault categories at each time point Indicates the first At the [time]th moment Correction probability of class-specific faults Indicates the total number of fault categories. This indicates the location index of the primary fault on the topology. This indicates the total number of preset part categories. Indicates the first Class of faults and the first Correspondence coefficients between different types of parts.

[0096] By performing the above operations, this solution addresses the problems in existing elevator fault diagnosis technologies, such as the independence of operating condition identification and fault diagnosis, the tendency to misjudge normal operating condition transitions as faults, and the difficulty in comprehensively and quantitatively assessing fault types, fault locations, and risk levels. This solution decodes the current operating condition using state fingerprint vectors and introduces an operating condition compatibility coefficient during fault source decomposition to suppress and correct fault categories that do not conform to the current operating condition. Simultaneously, it integrates abnormal evidence values, fault probabilities, and operating condition inconsistencies to generate a comprehensive risk score and risk level. This enables fault tracing and risk quantification under operating condition perception conditions, significantly improving the accuracy of fault diagnosis and the relevance of alarms.

[0097] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S5, the alarm output specifically includes the following steps:

[0098] Step S51: Generate an alarm event package. When the risk level reaches the preset alarm conditions, the elevator number, node identifier, monitoring time, current operating condition label, main fault category, risk level, fault location index, and status fingerprint are encapsulated into an alarm event package, represented as: ;in, Indicates the first Alarm event packets generated at specific times. This represents a unique identifier for the corresponding node. Indicates the first Each monitoring moment, Indicates the current operating condition label. Indicates the main fault category. Indicates the fault location index. Indicates the risk level. Represents the fingerprint vector of the current state;

[0099] Step S52: Parameter Update. When manual inspection or subsequent confirmation results determine that a certain period is a normal operating sample, the short window baseline, long window baseline, and threshold parameters are incrementally updated using the state fingerprint of that period. When the confirmation result determines that it is a valid fault sample, the compatibility relationship between operating conditions and faults is corrected, and the credibility of the corresponding node is adjusted accordingly, as shown below:

[0100] ;

[0101] in, This represents the updated short-window baseline vector. This represents the short-window baseline vector before the update. This indicates the short-window baseline update gating coefficient. This represents the updated long window baseline vector. This represents the baseline vector of the long window before the update. Indicates the long window baseline update coefficient. This indicates the threshold for anomaly entry at the current moment. This indicates the updated exception entry threshold. Indicates the threshold update step size. This represents the mean value of abnormal evidence within the confirmed normal sample window. This indicates the compatibility coefficient between the updated operating conditions and faults. This indicates the compatibility coefficient between the operating conditions and faults before the update. This represents the compatibility relationship update coefficient. This indicates an indicator function; it takes the value 1 when both the operating condition category and the fault category match the confirmation result, and 0 otherwise. This indicates the updated node trustworthiness. This indicates the node's trustworthiness before the update. This indicates the step size for node credibility feedback. Indicates the relationship with the first The confirmation quality label corresponding to each node.

[0102] Example 7, see Figure 2Based on the above embodiments, the present invention provides an IoT-based real-time elevator operation status monitoring system, which includes a monitoring node registration module, an operation parameter acquisition module, a residual propagation modeling module, an operating condition decoding module, and an alarm output module.

[0103] The monitoring node registration module is used to generate unique identifiers for each participating node and establish a mapping relationship with elevator equipment and installation location to form a node object table; to evaluate the credibility of each node and select valid monitoring nodes; and to construct an elevator monitoring topology map based on node spatial location, latency consistency and link quality.

[0104] The operation parameter acquisition module is connected to the monitoring node registration module and is used to acquire multi-source operation parameters of the elevator through car sensor nodes, landing sensor nodes, machine room sensor nodes and control cabinet sensor nodes to form an original observation sequence; generate fusion weights by combining observation time offset, link quality and node reliability; and calculate synchronous sampling values ​​according to the fusion weights after compensating for missing or distorted observation values ​​in the sampling window.

[0105] The residual propagation modeling module is connected to the operating parameter acquisition module and is used to standardize the synchronous sampled values ​​to generate basic state features; combine differential features and cross-parameter coupling features with the basic state features to form a state fingerprint vector; establish short-window baselines, long-window baselines, and pattern baselines and calculate the deviation between the state fingerprint vector and each baseline to obtain the comprehensive residual; perform graph propagation on the comprehensive residual according to the elevator monitoring topology map to obtain a residual heat map; perform time recursion on the residual to obtain anomaly evidence values, and output the abnormal state through a double-threshold hysteresis method; the residual propagation modeling module includes a baseline construction unit and a graph propagation unit. The baseline construction unit is used to construct the short-window baseline, long-window baseline, and pattern baseline and calculate the comprehensive residual. The graph propagation unit is used to propagate the comprehensive residual according to the adjacency matrix of the elevator monitoring topology map to obtain the residual heat map.

[0106] The operating condition decoding module is connected to the residual propagation modeling module. It is used to input the state fingerprint vector into the operating condition decoding model to obtain the current elevator operating condition label; to decompose the abnormal residual into fault sources to obtain the probability of various faults, and to correct the probability by combining the operating condition compatibility coefficient to obtain the fault correction probability; to generate a risk score by fusing the abnormal evidence value, the fault correction probability, and the operating condition inconsistency item, and to map it to a risk level; to determine the main fault category based on the fault correction probability, and to determine the fault location by combining the elevator monitoring topology map.

[0107] The alarm output module is connected to the operating condition decoding module and is used to encapsulate the elevator number, node identifier, monitoring time, operating condition label, main fault category, risk level, fault location, and status fingerprint into an alarm event package and output it when the risk level reaches the preset alarm conditions. The alarm output module is also connected to the monitoring node registration module and the residual propagation modeling module respectively, and is used to update the baseline, threshold, compatibility relationship between operating conditions and faults, and node credibility based on the confirmation result of the alarm event package.

[0108] It should be noted that, in this document, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0109] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0110] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A method for real-time monitoring of elevator operating status based on the Internet of Things, characterized in that: The method includes the following steps: Step S1: Register monitoring nodes. Generate a unique identifier for each node participating in the monitoring and establish its mapping relationship with the elevator and installation location to form a node object table. At the same time, jointly evaluate the signal strength, reporting continuity and location consistency of each node to calculate the credibility and screen out the effective monitoring nodes. Then, based on the spatial location relationship of the nodes, communication delay characteristics and link quality indicators, construct the elevator monitoring topology map of the screened nodes. Step S2: Acquire operating parameters. Collect multi-source operating parameters such as door operator opening and closing status, door lock status, car speed, traction current, and braking temperature to form the original observation sequence. Combine time offset, link quality, and node reliability to generate fusion weights. Based on the state change trend in historical continuous sampling data, recursively compensate for missing or abnormally deviated observations within the sampling window, and then calculate the synchronous sampling value according to the fusion weights. Step S3: Residual propagation modeling, robustly standardize the synchronous sampled values ​​and compress extreme fluctuations to generate basic state features; combine short-time difference, long-time difference and cross-parameter coupling terms to form a state fingerprint vector; establish short-time window baseline, long-term operating baseline and mode baseline corresponding to different operating conditions, calculate the deviation between the state fingerprint and the baseline to obtain the comprehensive residual; use the monitoring topology map to propagate the residual to form a residual heat map; then perform time recursion on the residual to obtain anomaly evidence values, and output the abnormal state through a double threshold hysteresis method; Step S4: Operating condition decoding. Input the current and previous time-state fingerprint vectors into the operating condition decoding model and output the operating condition category label; combine the topology propagation residual and the residual change rate to decompose the fault source of the abnormal residual, and introduce the operating condition compatibility coefficient to correct the fault category probability; fuse the abnormal evidence value, fault probability and operating condition inconsistency term to generate a risk score and map it to a risk level; select the fault category with the highest corrected probability as the main fault category, and generate an alarm direction based on the topology cluster location; Step S5: Alarm output. When the risk level reaches the preset conditions, the elevator number, node identifier, monitoring time, working condition label, main fault category, risk level, fault location and status fingerprint are encapsulated into an alarm event package. Based on the results of manual inspection or subsequent confirmation, the short window baseline, long window baseline, abnormal threshold, working condition fault compatibility relationship and node credibility are incrementally updated.

2. The method for real-time monitoring of elevator operation status based on the Internet of Things according to claim 1, characterized in that: The monitoring node registration in step S1 includes: Read the physical address, elevator number, installation location code, node type, and power-on timestamp of each monitoring node, combine the above information, and perform a hash operation to generate a unique node identifier; The unique identifier of each node is associated with the corresponding elevator equipment and installation location to form a node object table for data tracking.

3. The method for real-time monitoring of elevator operation status based on the Internet of Things according to claim 1, characterized in that: The node credibility assessment in step S1 includes: The node reliability is obtained by jointly calculating the received signal strength, data reporting continuity, and node location consistency of the monitoring nodes. Monitoring nodes with a confidence level higher than a preset threshold are designated as valid monitoring nodes, while monitoring nodes with a confidence level lower than the preset threshold are marked as nodes to be reviewed. An elevator monitoring topology map is established based on the installation location, functional relationships, and communication link relationships of effective monitoring nodes, and the topology map is used as a constraint structure for subsequent anomaly propagation analysis.

4. The method for real-time monitoring of elevator operation status based on the Internet of Things according to claim 1, characterized in that: The multi-source state parameters in step S2 include door operator opening and closing status, door lock status, car speed, traction current, braking temperature, vibration amplitude, leveling error, communication delay, load status, and floor stopping status. The process of generating the synchronous sampled values ​​includes: The fusion weights for each observation data are determined based on the time offset of the observation data, the quality of the corresponding communication link, and the reliability of the monitoring nodes. For missing or distorted observations within the sampling window, recursive compensation is performed based on the state change trends of nearby historical moments. The compensated observations are fused according to the fusion weights to obtain synchronous sampled values ​​under a unified time reference.

5. The method for real-time monitoring of elevator operation status based on the Internet of Things according to claim 1, characterized in that: The state fingerprint vector in step S3 includes: Robust normalization is performed based on the historical normal operation distribution of synchronous sampled values, and nonlinear compression is used to reduce the impact of extreme fluctuations to obtain basic state characteristics. The short-time difference features, long-time difference features, and coupling features between different operating parameters of the synchronous sampled values ​​are calculated and combined with the basic state features to form a state fingerprint vector that characterizes the change law of elevator operating state.

6. The method for real-time monitoring of elevator operation status based on the Internet of Things according to claim 1, characterized in that: The abnormal states output in step S3 include: The elevator’s recent operating status, long-term stable status, and normal status under different operating conditions are described based on the short window baseline, long window baseline, and mode baseline, respectively. The deviation between the state fingerprint vector and each baseline is calculated to obtain the comprehensive residual; Based on the node connection relationship in the elevator monitoring topology diagram, a propagation analysis is performed on the comprehensive residual to form a residual heat map representing the scope of anomaly influence. The abnormal evidence value is obtained by time recursion of the residual after propagation. When the abnormal evidence value exceeds the entry threshold, the abnormal state is output. When the abnormal evidence value is lower than the exit threshold, the abnormal state is lifted. When the abnormal evidence value is between the entry threshold and the exit threshold, the previous state is maintained.

7. The method for real-time monitoring of elevator operation status based on the Internet of Things according to claim 1, characterized in that: The fault source decomposition in step S4 includes: The difference between the current state fingerprint vector and the short window baseline is calculated to obtain the abnormal residual; By combining the abnormal residuals, the residuals after topology propagation, and the residual change rate into the fault source decomposition model, the fault probabilities corresponding to door system faults, traction system faults, braking system faults, leveling system faults, sensor drift, and communication anomalies are obtained. By using the operating condition compatibility coefficient, the probability of fault categories that do not conform to the current operating conditions is suppressed, and the fault correction probability is obtained.

8. An IoT-based real-time elevator operation status monitoring system, used to implement the IoT-based real-time elevator operation status monitoring method as described in any one of claims 1-7, characterized in that: It includes a monitoring node registration module, an operating parameter acquisition module, a residual propagation modeling module, an operating condition decoding module, and an alarm output module.

9. The real-time monitoring system for elevator operation status based on the Internet of Things according to claim 8, characterized in that: The monitoring node registration module is used to generate a unique identifier for each node participating in the monitoring and establish a mapping relationship with the elevator equipment and installation location, forming a node object table; The credibility of each node is assessed and the effective monitoring nodes are selected. Based on the spatial location of nodes, latency consistency, and link quality, the selected nodes are used to construct an elevator monitoring topology map. The operating parameter acquisition module is connected to the monitoring node registration module and is used to collect multi-source operating parameters of the elevator through car sensor nodes, landing sensor nodes, machine room sensor nodes and control cabinet sensor nodes to form an original observation sequence. The fusion weights are generated by combining observation time offset, link quality, and node credibility. After compensating for missing or distorted observations within the sampling window, the synchronous sampled values ​​are calculated according to the fusion weights. The residual propagation modeling module is connected to the running parameter acquisition module and is used to standardize the synchronous sampled values ​​to generate basic state features. A state fingerprint vector is formed by concatenating differential features and cross-parameter coupling features with the basic state features; a short-window baseline, a long-window baseline, and a pattern baseline are established, and the deviation between the state fingerprint vector and each baseline is calculated to obtain a comprehensive residual; the comprehensive residual is propagated through graph propagation according to the elevator monitoring topology map to obtain a residual heat map; the residual is recursively extrapolated over time to obtain anomaly evidence values, and the abnormal state is output through a dual-threshold hysteresis method; The residual propagation modeling module includes a baseline construction unit and a graph propagation unit. The baseline construction unit is used to construct the short window baseline, long window baseline and pattern baseline and calculate the comprehensive residual. The graph propagation unit is used to propagate the comprehensive residual according to the adjacency matrix of the elevator monitoring topology graph to obtain the residual heat map. The operating condition decoding module is connected to the residual propagation modeling module and is used to input the state fingerprint vector into the operating condition decoding model to obtain the current elevator operating condition label. The abnormal residuals are decomposed into fault sources to obtain the probability of various faults, and the probability is corrected by combining the operating condition compatibility coefficient to obtain the fault correction probability. A risk score is generated by integrating the abnormal evidence value, fault correction probability, and inconsistency item of operating condition, and then mapped to a risk level; The main fault category is determined based on the fault correction probability, and the fault location is determined in conjunction with the elevator monitoring topology map; The alarm output module is connected to the working condition decoding module and is used to encapsulate the elevator number, node identifier, monitoring time, working condition label, main fault category, risk level, fault location and status fingerprint into an alarm event package and output it when the risk level reaches the preset alarm conditions. The alarm output module is also connected to the monitoring node registration module and the residual propagation modeling module, respectively, and is used to update the baseline, threshold, compatibility relationship between operating conditions and faults, and node credibility based on the confirmation results of the alarm event packets.

Images