An elevator brake abnormality diagnosis system and method based on multi-modal perception

Abstract

The application discloses an elevator brake abnormality diagnosis system and method based on multi-modal perception, relates to the technical field of elevator brakes, and collects longitudinal acceleration and deceleration data of an elevator in a full working condition in real time, compares normal benchmark values to determine acceleration abnormalities, and associates corresponding brake types according to working conditions; for the brake, a high-speed camera is started to collect trajectory images, vibration, temperature and braking force data are synchronously collected to form a multi-modal data set; trajectory images are processed to extract motion features and match faults, misjudgments are excluded through cross verification of auxiliary data; parameters are adjusted according to faults, successful adjustment is achieved when the abnormality disappears, and double brakes are started to relay when the mechanical limit is invalid; a complementary relay brake is selected, the required capacity, the optimal relay time period and distance of the complementary relay brake are calculated, and the relay control is executed according to the parameters.

Description

Technical Field

[0001] This invention relates to the field of elevator brake technology, specifically to an elevator brake anomaly diagnosis system and method based on multimodal perception. Background Technology

[0002] As a core component ensuring the safe operation of elevators, the performance stability of elevator brakes is of paramount importance. With the increasing service life and complexity of elevators, brakes are prone to failure, leading to safety accidents. Therefore, reliable diagnostic and control technologies are needed.

[0003] Existing technologies mostly collect acceleration data for single operating conditions, ignoring high-risk fault phases such as start-up acceleration and deceleration. After anomaly detection, they rely on manual methods or fuzzy correlation of brake types, failing to distinguish the differences in brake faults under different operating conditions, resulting in low troubleshooting efficiency. Existing technologies often rely solely on single sensor diagnosis, ignoring motion trajectory characteristics, leading to a high false alarm rate, and do not consider combining trajectory images with multimodal data to determine faults. Even when a fault is identified, there is a lack of targeted parameter adjustment schemes, or the anomaly is not verified after adjustment, forming an open-loop process. Traditional methods often directly shut down the machine when parameter adjustments to the limits are ineffective, affecting operation; a few use backup brakes, but they do not consider relay braking situations, nor do they calculate the required braking force and handover parameters. Summary of the Invention

[0004] The purpose of this invention is to provide an elevator brake anomaly diagnosis system and method based on multimodal perception, so as to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for diagnosing elevator brake anomalies based on multimodal sensing, comprising: Real-time acquisition of longitudinal acceleration and deceleration data for the entire elevator's ascent and descent; comparison of the real-time acquired data with normal baseline values ​​to determine acceleration anomalies and associate the corresponding brake type according to the operating condition; For the associated brake type, a high-speed camera is activated to acquire motion trajectory images, which are time-aligned with the moment of acceleration anomaly; vibration data, temperature data, and braking force data are collected to form a multimodal verification dataset; The system processes trajectory images to extract motion features and matches fault types; it also uses cross-validation with auxiliary modal data to eliminate false positives. Adjust parameters according to the type of fault, monitor acceleration and trajectory images in real time, and determine that the adjustment is successful when the abnormality disappears after parameter adjustment. If the adjustment is still ineffective after reaching the mechanical limit, activate the dual brake relay control. Based on the type of fault brake and the elevator configuration, select a relay brake with complementary capabilities, calculate the required capacity, optimal handover time and distance of the relay brake, and execute relay control according to the calculated parameters.

[0006] In conjunction with the first aspect, in the first implementation of the first aspect of this application, the real-time acquisition of longitudinal acceleration and deceleration data of the elevator during all operating conditions of ascent and descent includes: Using a triaxial accelerometer, data is collected for both ascent and descent, divided into three sub-stages: acceleration at start-up, constant speed operation, and deceleration near the target layer. Specifically, longitudinal acceleration data is collected for the acceleration at start-up sub-stage, longitudinal acceleration fluctuation data is collected for the constant speed operation sub-stage, longitudinal deceleration data is collected for the deceleration near the target layer sub-stage, and lateral acceleration fluctuation values ​​are collected for all three sub-stages. The data from each acceleration sensor are timestamped to generate acceleration change curves for each sub-stage, and the acceleration reference range for each sub-stage under normal operating conditions is recorded.

[0007] In conjunction with the first aspect, in the second implementation of the first aspect of this application, the step of comparing the real-time collected data with a normal reference value to determine an acceleration anomaly and associating it with the corresponding brake type according to the operating condition includes: Based on the acceleration reference range of each sub-stage under normal operating conditions, the acceleration peak threshold range, acceleration fluctuation threshold range, and deceleration peak threshold range are set for the starting acceleration, constant speed operation, and deceleration sub-stages near the target layer. When the real-time collected data exceeds the corresponding threshold range, the sub-stage is determined to be abnormal. The load is compared with the preset light load, medium load and heavy load thresholds, and divided into light load condition, medium load condition and heavy load condition. The analysis of the startup acceleration sub-stage correlation logic shows that during upward startup acceleration, abnormalities under light load conditions are associated with electromechanical brakes, while abnormalities under medium or heavy load conditions are associated with disc brakes. During downward startup acceleration, abnormalities under light load conditions are associated with electromechanical brakes, abnormalities under medium load conditions are preferentially associated with electromechanical brakes, and if the electromechanical brakes do not exhibit abnormal characteristics, they are associated with disc brakes. Abnormalities under heavy load conditions are associated with disc brakes. Analyze the logic associated with the uniform speed operation sub-stage and identify abnormal associations with the brakes currently involved in maintaining braking force. Analyze the correlation logic of the deceleration sub-stage near the target layer. When decelerating upwards, the abnormality is associated with the drum brake. When decelerating downwards, the abnormality is associated with the disc brake first. If the disc brake parameters are normal, the abnormality is associated with the electromechanical brake. After identifying an anomaly, record the structured information of the anomaly sub-stage, anomaly type, real-time value, baseline range, and associated brake type.

[0008] In conjunction with the first aspect, in the third implementation of the first aspect of this application, the step of activating a high-speed camera to acquire motion trajectory images for the associated brake type and aligning them with the time of acceleration anomalies includes: Based on the associated brake type, a corresponding ROI is set. When an acceleration anomaly is determined and associated with the brake type, the high-speed camera acquires trajectory images for preset durations before and after the anomaly moment. The image sequence is aligned with the acceleration anomaly moment using an interpolation algorithm. The original image is cropped according to the preset ROI, and the image data is stored in the format of associated brake type, anomaly timestamp, frame number, and ROI parameter.

[0009] In conjunction with the first aspect, in the fourth implementation of the first aspect of this application, the collection of vibration data, temperature data, and braking force data to form a multimodal verification dataset includes: Vibration and temperature data of three types of brakes are collected using vibration sensors and infrared thermometers; pressure data of disc brakes are collected using pressure sensors and converted into braking force data; current data of electromechanical brakes are collected using current sensors and real-time braking force data is calculated based on existing current-electromagnetic force conversion models; and pressure values ​​of the contact surface between the brake shoe and the brake drum of drum brakes are collected using pressure sensors and used as braking force data. Using the moment of acceleration anomaly as the reference time point, vibration data, temperature data, and braking force data are aligned according to the timestamp difference to form a six-dimensional associated dataset of time, acceleration, trajectory image, vibration, temperature, and braking force, which is then stored in a structured manner.

[0010] In conjunction with the first aspect, in the fifth implementation of the first aspect of this application, the step of processing the trajectory image to extract motion features and match fault types includes: The trajectory image is preprocessed, including noise suppression, edge enhancement, and dynamic ROI locking; For drum brakes, identify stagnation and abrupt movement features where the duration of stagnation on one side of the brake shoe exceeds a preset stagnation threshold and the abrupt movement displacement exceeds a preset abrupt movement threshold, and extract the time difference between the contact of the brake shoes on both sides; for disc brakes, extract the piston-brake disc clearance and piston speed fluctuation patterns; for electromechanical brakes, extract the armature trajectory offset from the preset axis and the armature response delay. Analyzing the fault types of drum brakes, when one side of the brake shoe exhibits stagnation and sudden movement characteristics, and the impact amplitude in the vibration data at the corresponding moment exceeds the preset vibration impact threshold, while the temperature rise rate of the temperature data is less than the preset normal temperature rise rate, it is determined to be a local jamming of the brake shoe; when the time difference between the contact of the two brake shoes exceeds the preset contact synchronization allowable value and the lateral acceleration fluctuation value exceeds the preset lateral fluctuation threshold, it is determined to be an abnormal adjustment of the brake shoe gap. Analyzing the fault types of disc brakes, when the piston movement speed fluctuates beyond the preset stable speed range and the pressure data fluctuation amplitude exceeds the preset pressure fluctuation threshold, it is determined that the piston seal ring is aged and stuck; when the gap between the piston and the brake disc is greater than the preset upper limit of normal gap and the braking force data is less than the preset lower limit of rated braking force, it is determined that the hydraulic system pressure is abnormal. Analyzing the fault types of electromechanical brakes, when the armature response delay exceeds the preset delay threshold and the current data is within the preset rated working range of the electromagnet, it is determined to be an electromagnet residual magnetism fault; when the armature trajectory deviates from the preset axis by more than the preset deviation threshold and the vibration data shows noise consistent with the trajectory deviation direction, it is determined to be an armature guide shaft wear. The extracted motion features, matched fault types, and associated logic are stored in the format of brake type, abnormal timestamp, motion feature, fault type, and associated logic.

[0011] In conjunction with the first aspect, in the sixth implementation of the first aspect of this application, the step of eliminating false positives through cross-validation of auxiliary modal data includes: When it is initially determined that the brake shoe is partially stuck, verify whether the time difference between the time node of the brake shoe stagnation and sudden movement on one side and the impact signal timestamp of the vibration data is less than the preset time synchronization threshold; when it is initially determined that the brake shoe gap adjustment is abnormal, verify whether the correlation coefficient between the time difference of the brake shoe contact on both sides and the lateral acceleration fluctuation value is greater than the preset correlation threshold. When the initial diagnosis is that the piston seal ring is aging and stuck, verify whether the synchronization deviation of the period and peak time of the piston speed fluctuation and pressure fluctuation is less than the preset period synchronization threshold; when the initial diagnosis is that the hydraulic system pressure is abnormal, verify whether the negative correlation coefficient between the piston and brake disc clearance and the braking force data is less than the preset negative correlation threshold. When it is initially determined that the electromagnet is remanent, verify whether the current data is within the preset rated current range during the armature response delay period; when it is initially determined that the armature guide shaft is worn, verify whether the direction of the armature trajectory offset and the direction of vibration noise meet the preset direction matching threshold. When the core correlation is verified to meet the threshold requirements, the verification is marked as passed. When the core correlation is not verified, the failed item is recorded, trajectory images and auxiliary data are re-acquired, and fault matching is performed.

[0012] In conjunction with the first aspect, in the seventh implementation of the first aspect of this application, the step of adjusting parameters specifically according to the fault type, monitoring acceleration and trajectory images in real time, determining successful adjustment when the abnormality disappears after parameter adjustment, and activating dual-brake relay control when adjustment to the mechanical limit is still ineffective, includes: To address localized brake shoe jamming, adjust the brake shoe clearance compensation parameters; to address abnormal brake shoe clearance adjustment, initiate a synchronized calibration program for both brake shoes, dynamically adjusting the stroke limit parameters of a single brake shoe to achieve the target of synchronized fit; to address piston seal aging and jamming, adjust the hydraulic system pressure compensation coefficient and correct the allowable threshold for piston movement speed fluctuations; to address abnormal hydraulic system pressure, adjust the hydraulic pump output pressure setting; to address electromagnet residual magnetism, introduce reverse current pulse parameters to shorten the armature engagement and release interval; to address armature guide shaft wear, adjust the guide shaft lubrication cycle parameters and correct the allowable offset threshold for the armature trajectory. During the adjustment process, acceleration data and trajectory images are collected for each sub-stage. After each adjustment, the acceleration recovery rate and trajectory feature recovery rate are calculated. When both reach the preset recovery threshold, the adjustment is considered successful, and the final adjustment parameters and recovery process data are recorded. When the parameter adjustment reaches the mechanical limit and both the acceleration recovery rate and trajectory feature recovery rate are less than the preset recovery threshold, the adjustment is considered invalid, triggering dual-brake relay control, and recording the adjustment limit value and unrecovered abnormal features.

[0013] In conjunction with the first aspect, in the eighth implementation of the first aspect of this application, the step of selecting a relay brake with complementary capabilities based on the type of fault brake and the elevator configuration, calculating the required capacity, optimal handover time period and distance of the relay brake, and executing relay control according to the calculated parameters includes: When the faulty brake is a drum brake, a disc brake should be selected as the replacement brake; when the faulty brake is a disc brake, a drum brake should be selected as the replacement brake; when the faulty brake is an electromechanical brake, a disc brake should be selected as the replacement brake. Calculate the current remaining braking force of the faulty brake based on real-time braking force data. The formula is: ; in, Dedicated power to the faulty brake. For a faulty brake, the coefficient of performance degradation is: for a drum brake, it is the ratio of the actual travel of the brake shoe to the rated travel of the brake shoe; for a disc brake, it is the ratio of the actual piston pressure to the rated piston pressure; and for an electromechanical brake, it is the ratio of the actual current to the rated current of the electromagnet. Calculate the total braking force required for the elevator based on the current operating conditions. The calculation formula for the uniform motion sub-stage is: ; The calculation formulas for the acceleration and deceleration sub-stages near the target layer are as follows: ; in, Given the current elevator load, To accelerate or decelerate the current sub-stage objective. It is the acceleration due to gravity. The coefficient of friction between the car and the guide rail; Calculate the required capacity of the relay brake The formula is: ; To calculate the optimal handover period, based on the acceleration change curve, candidate periods with acceleration fluctuation values ​​less than the acceleration fluctuation threshold within the current sub-stage are selected, and the period with the longest duration among the candidate periods is chosen as the optimal handover period. When the deceleration sub-stage is near the target floor, the distance between the elevator position and the target floor corresponding to the candidate time period must be greater than the safety distance. The formula is: ; in, For real-time running speed; Calculate the optimal junction distance The formula is: ; in, The known response time of the relay brake; according to Send control command, disc brake adjusts pressure to The calculation formula is: ; in, The piston's surface area under stress; Electromechanical brake adjustment current to The calculation formula is: ; in, The conversion coefficient between electric current and electromagnetic force; Drum brake adjustment: The pressure between the brake shoe and the brake drum contact surface is adjusted to... = Send a discharge command to the faulty brake, discharge rate The calculation formula is: ; During the control process, acceleration data, trajectory images, and vibration data are monitored in real time. When all of them meet the requirements, the handover is deemed complete, the control parameters and monitoring data are recorded, and a relay control log is generated.

[0014] Secondly, the present invention provides an elevator brake anomaly diagnosis system based on multimodal perception, comprising: Multimodal data acquisition module: Among them, the acceleration acquisition unit collects acceleration data of the elevator in all operating conditions, divides it into sub-stages, aligns the timestamps, generates curves and records the normal baseline; the visual trajectory acquisition unit sets the ROI according to the associated brake, collects trajectory images before and after abnormal moments, aligns them and then crops and stores them; the auxiliary physical quantity acquisition unit collects brake vibration, temperature and braking force data, and forms a six-dimensional structured dataset after alignment. Anomaly detection and fault matching module: The anomaly detection unit sets an acceleration threshold, divides the load conditions, compares data to determine sub-stage anomalies and associates them with the corresponding brakes; the trajectory processing and feature extraction unit preprocesses the trajectory image and extracts the core motion features of the brakes; the fault matching unit combines motion features and auxiliary data to determine the specific fault type and stores the information. Multimodal cross-validation module: The core correlation validation unit verifies whether the key correlation meets the threshold requirements for the initial fault; the validation result processing unit confirms the fault when the standard is met, and records the problem if the standard is not met, and re-collects data and matches the fault a second time. Fault handling and relay control module: The parameter adjustment unit formulates adjustment plans according to the fault type and monitors the adjustment process in real time; the effect evaluation unit calculates the acceleration and trajectory feature recovery degree, and if the standard is met, the adjustment is judged to be successful; if it is still ineffective after reaching the limit, the relay is triggered; the relay control unit selects the complementary relay brake, calculates the force value and handover parameters, generates control commands and monitors the handover process, and switches to individual control after completion. System Log and Information Management Unit: The information recording unit records structured information throughout the entire process; the log generation unit integrates the recorded data and generates a full-process log.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention collects acceleration and deceleration data of elevators under all operating conditions, determines acceleration anomalies, and associates them with brake types according to operating conditions.

[0016] 2. This invention extracts motion features from the motion trajectory image of the associated brake and collects multimodal datasets to match fault types and adjust parameters until the anomaly disappears.

[0017] 3. This invention has a dual-brake relay control function. When the parameters are adjusted to the mechanical limit and still ineffective, the required capacity of the relay brake, the optimal handover time and distance are calculated, and relay control is performed. Attached Figure Description

[0018] Figure 1 This is a schematic diagram illustrating the steps of an elevator brake anomaly diagnosis method based on multimodal perception according to the present invention. Figure 2 This is a system structure diagram of an elevator brake anomaly diagnosis system based on multimodal perception according to the present invention. Detailed Implementation

[0019] 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.

[0020] Example: Figures 1-2 As shown, the present invention provides a technical solution. like Figure 1 A schematic diagram illustrating the steps of an elevator brake anomaly diagnosis method based on multimodal perception is provided in this invention. The method includes: Step S100: Collect longitudinal acceleration and deceleration data of the elevator during all operating conditions of rising and falling in real time; compare the real-time collected data with the normal baseline value, determine the acceleration anomaly, and associate the corresponding brake type according to the operating condition; Specifically, data from both the ascent and descent phases are collected using a triaxial accelerometer, divided into three sub-phases: acceleration at startup, constant speed operation, and deceleration near the target layer. Among these, longitudinal acceleration data from the acceleration at startup sub-phase, longitudinal acceleration fluctuation data from the constant speed operation sub-phase, longitudinal deceleration data from the deceleration near the target layer sub-phase, and lateral acceleration fluctuation values ​​from all three sub-phases are collected. The data from each acceleration sensor are timestamped to generate acceleration change curves for each sub-stage, and the acceleration reference range for each sub-stage under normal operating conditions is recorded.

[0021] Based on the acceleration reference range of each sub-stage under normal operating conditions, the acceleration peak threshold range, acceleration fluctuation threshold range, and deceleration peak threshold range are set for the starting acceleration, constant speed operation, and deceleration sub-stages near the target layer. When the real-time collected data exceeds the corresponding threshold range, the sub-stage is determined to be abnormal. The load is compared with the preset light load, medium load and heavy load thresholds, and divided into light load condition, medium load condition and heavy load condition. The analysis of the startup acceleration sub-stage correlation logic shows that during upward startup acceleration, abnormalities under light load conditions are associated with electromechanical brakes, while abnormalities under medium or heavy load conditions are associated with disc brakes. During downward startup acceleration, abnormalities under light load conditions are associated with electromechanical brakes, abnormalities under medium load conditions are preferentially associated with electromechanical brakes, and if the electromechanical brakes do not exhibit abnormal characteristics, they are associated with disc brakes. Abnormalities under heavy load conditions are associated with disc brakes. Analyze the logic associated with the uniform speed operation sub-stage and identify abnormal associations with the brakes currently involved in maintaining braking force. Analyze the correlation logic of the deceleration sub-stage near the target layer. When decelerating upwards, the abnormality is associated with the drum brake. When decelerating downwards, the abnormality is associated with the disc brake first. If the disc brake parameters are normal, the abnormality is associated with the electromechanical brake. After identifying an anomaly, record the structured information of the anomaly sub-stage, anomaly type, real-time value, baseline range, and associated brake type.

[0022] In one specific embodiment, an office building elevator has a rated load of 1600kg. A 100Hz triaxial accelerometer executes step S100, dividing the load into light load (0-500kg), medium load (501-1000kg), and heavy load (1001-1600kg). Each sub-stage sets a benchmark and threshold, such as a starting acceleration longitudinal acceleration benchmark of 0.8-1.2m / s² and a peak threshold of ≤1.5m / s².

[0023] During the upward acceleration, the longitudinal acceleration under a light load of 300kg is 1.6m / s², exceeding the threshold, indicating an anomaly and linking it to the electromechanical brake. During the downward acceleration under a medium load of 700kg, the longitudinal acceleration is 1.7m / s², exceeding the threshold, and the electromechanical brake is linked first. During the upward deceleration under a heavy load of 1200kg, the longitudinal deceleration is 1.8m / s², exceeding the threshold, and the drum brake is linked. All anomaly information is recorded in a structured format, with the following content: "Anomaly sub-stage: upward acceleration; anomaly type: longitudinal acceleration peak exceeds limit; real-time value: 1.6m / s²; reference range: 0.8-1.2m / s²; linked brake type: electromechanical brake."

[0024] Step S200: For the associated brake type, start the high-speed camera to acquire motion trajectory images, aligning them with the time of acceleration anomalies; acquire vibration data, temperature data, and braking force data to form a multimodal verification dataset; Specifically, a corresponding ROI is set according to the associated brake type. When an acceleration anomaly is determined and associated with the brake type, the high-speed camera acquires trajectory images for preset durations before and after the anomaly moment. The image sequence is aligned with the acceleration anomaly moment through an interpolation algorithm. The original image is cropped according to the preset ROI, and the image data is stored in the format of associated brake type, anomaly timestamp, frame number, and ROI parameter.

[0025] Vibration and temperature data of three types of brakes are collected using vibration sensors and infrared thermometers; pressure data of disc brakes are collected using pressure sensors and converted into braking force data; current data of electromechanical brakes are collected using current sensors and real-time braking force data is calculated based on existing current-electromagnetic force conversion models; and pressure values ​​of the contact surface between the brake shoe and the brake drum of drum brakes are collected using pressure sensors and used as braking force data. Using the moment of acceleration anomaly as the reference time point, vibration data, temperature data, and braking force data are aligned according to the timestamp difference to form a six-dimensional associated dataset of time, acceleration, trajectory image, vibration, temperature, and braking force, which is then stored in a structured manner.

[0026] In one specific embodiment, a high-speed camera with a resolution of 1920×1080 and a frame rate of 200fps is used. ROIs are set according to the associated brake type: electromechanical brakes correspond to the armature movement area, disc brakes correspond to the piston propulsion area, and drum brakes correspond to the brake shoe contact area. When an associated electromechanical brake is detected during the acceleration phase, the camera acquires trajectory images 2 seconds before and 3 seconds after the abnormal moment. The image sequence is aligned with the abnormal moment using a linear interpolation algorithm, cropped according to the ROI, and stored in the specified format.

[0027] Simultaneous acquisition of multimodal data: vibration sensors measured vibrations of 0.8g for electromechanical brakes, 1.2g for disc brakes, and 1.5g for drum brakes; infrared thermometers collected temperatures every 100ms, with the electromechanical armature at 42℃, the disc brake piston at 45℃, and the drum brake shoe at 50℃; regarding braking force data, the disc brake pressure sensor collected 1.8MPa, which was converted to 36kN based on a piston area of ​​0.02m²; the electromechanical current sensor collected 8A, which was calculated to 4kN using the conversion model K=0.5kN / A; and the pressure sensor collected the pressure value of 25kN between the brake shoe and the brake drum of the drum brake, which was used as braking force data.

[0028] Using the abnormal time 2025.06.15 09:32:45.120 as the baseline, vibration, temperature, and braking force data are aligned according to a 10ms timestamp difference to form a six-dimensional associated dataset. For example, at the baseline time +0ms, the corresponding acceleration is 1.6m / s², trajectory image frame 0500, vibration is 0.8g, temperature is 42℃, and braking force is 4kN. All data are stored in a structured CSV format.

[0029] Step S300: Process the trajectory image to extract motion features and match the fault type; eliminate false judgments through cross-validation using auxiliary modal data; Specifically, the trajectory image is preprocessed, including noise suppression, edge enhancement, and dynamic ROI locking; For drum brakes, identify stagnation and abrupt movement features where the duration of stagnation on one side of the brake shoe exceeds a preset stagnation threshold and the abrupt movement displacement exceeds a preset abrupt movement threshold, and extract the time difference between the contact of the brake shoes on both sides; for disc brakes, extract the piston-brake disc clearance and piston speed fluctuation patterns; for electromechanical brakes, extract the armature trajectory offset from the preset axis and the armature response delay. Analyzing the fault types of drum brakes, when one side of the brake shoe exhibits stagnation and sudden movement characteristics, and the impact amplitude in the vibration data at the corresponding moment exceeds the preset vibration impact threshold, while the temperature rise rate of the temperature data is less than the preset normal temperature rise rate, it is determined to be a local jamming of the brake shoe; when the time difference between the contact of the two brake shoes exceeds the preset contact synchronization allowable value and the lateral acceleration fluctuation value exceeds the preset lateral fluctuation threshold, it is determined to be an abnormal adjustment of the brake shoe gap. Analyzing the fault types of disc brakes, when the piston movement speed fluctuates beyond the preset stable speed range and the pressure data fluctuation amplitude exceeds the preset pressure fluctuation threshold, it is determined that the piston seal ring is aged and stuck; when the gap between the piston and the brake disc is greater than the preset upper limit of normal gap and the braking force data is less than the preset lower limit of rated braking force, it is determined that the hydraulic system pressure is abnormal. Analyzing the fault types of electromechanical brakes, when the armature response delay exceeds the preset delay threshold and the current data is within the preset rated working range of the electromagnet, it is determined to be an electromagnet residual magnetism fault; when the armature trajectory deviates from the preset axis by more than the preset deviation threshold and the vibration data shows noise consistent with the trajectory deviation direction, it is determined to be an armature guide shaft wear. The extracted motion features, matched fault types, and associated logic are stored in the format of brake type, abnormal timestamp, motion feature, fault type, and associated logic.

[0030] When it is initially determined that the brake shoe is partially stuck, verify whether the time difference between the time node of the brake shoe stagnation and sudden movement on one side and the impact signal timestamp of the vibration data is less than the preset time synchronization threshold; when it is initially determined that the brake shoe gap adjustment is abnormal, verify whether the correlation coefficient between the time difference of the brake shoe contact on both sides and the lateral acceleration fluctuation value is greater than the preset correlation threshold. When the initial diagnosis is that the piston seal ring is aging and stuck, verify whether the synchronization deviation of the period and peak time of the piston speed fluctuation and pressure fluctuation is less than the preset period synchronization threshold; when the initial diagnosis is that the hydraulic system pressure is abnormal, verify whether the negative correlation coefficient between the piston and brake disc clearance and the braking force data is less than the preset negative correlation threshold. When it is initially determined that the electromagnet is remanent, verify whether the current data is within the preset rated current range during the armature response delay period; when it is initially determined that the armature guide shaft is worn, verify whether the direction of the armature trajectory offset and the direction of vibration noise meet the preset direction matching threshold. When the core correlation is verified to meet the threshold requirements, the verification is marked as passed. When the core correlation is not verified, the failed item is recorded, trajectory images and auxiliary data are re-acquired, and fault matching is performed.

[0031] In one specific embodiment, taking the electromechanical brake associated in step S200 as an example, step S300 is executed to preprocess the trajectory image acquired by the high-speed camera, suppress noise by Gaussian filtering, strengthen the armature edge by Sobel operator, and lock the armature motion area by dynamic ROI.

[0032] When extracting motion features, the frame difference method was used to calculate that the response delay of the armature from receiving the action command to starting the movement was 60ms, and the maximum offset of the armature trajectory from the preset axis was 0.8mm. Combining the preset threshold response delay threshold of 50ms and the offset threshold of 0.5mm, the preliminary analysis indicated that since the response delay of 60ms was greater than 50ms, and the 8A current collected by the current sensor was within the preset rated range of 5-10A, the preliminary judgment was that the problem was a residual magnetism fault in the electromagnet.

[0033] Cross-validation was performed to verify the current data during the armature response delay period from 09:32:45.120 to 09:32:45.180. During this period, the current was stable at 7.8-8.2A, which is within the rated range of 5-10A, meeting the preset verification threshold for current within the rated range, with a deviation of ≤±10%. Core correlation verification passed, and marker verification passed.

[0034] Finally, the information is stored in the following format: "Brake type: electromechanical; Abnormal timestamp: 20240615093245120; Motion characteristics: response delay 60ms, offset 0.8mm; Fault type: electromagnet residual magnetism; Associated logic: response delay exceeds threshold + current within rated range and verification passed."

[0035] Step S400: Adjust parameters according to the fault type, monitor acceleration and trajectory images in real time. When the abnormality disappears after parameter adjustment, the adjustment is considered successful. If the adjustment is still ineffective after reaching the mechanical limit, activate the dual brake relay control. Specifically, for localized brake shoe jamming, adjust the brake shoe clearance compensation parameters; for abnormal brake shoe clearance adjustment, initiate a synchronous calibration program for both brake shoes, dynamically adjusting the stroke limit parameters of a single brake shoe to achieve the target of synchronized fit; for piston seal aging and jamming, adjust the hydraulic system pressure compensation coefficient to correct the allowable threshold for piston movement speed fluctuations; for abnormal hydraulic system pressure, adjust the hydraulic pump output pressure setting; for residual magnetism in the electromagnet, introduce reverse current pulse parameters to shorten the armature engagement and release interval; for wear on the armature guide shaft, adjust the guide shaft lubrication cycle parameters to correct the allowable offset threshold for the armature trajectory. During the adjustment process, acceleration data and trajectory images are collected for each sub-stage. After each adjustment, the acceleration recovery rate and trajectory feature recovery rate are calculated. When both reach the preset recovery threshold, the adjustment is considered successful, and the final adjustment parameters and recovery process data are recorded. When the parameter adjustment reaches the mechanical limit and both the acceleration recovery rate and trajectory feature recovery rate are less than the preset recovery threshold, the adjustment is considered invalid, triggering dual-brake relay control, and recording the adjustment limit value and unrecovered abnormal features.

[0036] In one specific embodiment, step S400 is executed for the residual magnetism fault of the electromechanical brake electromagnet confirmed in step S300. Based on the fault type, an adjustment strategy of introducing a reverse current pulse is adopted. The initial reverse current pulse parameter is set to -2A, lasting 100ms, with the interval period synchronized with the armature action command. During the adjustment process, longitudinal acceleration data and armature trajectory images are collected in real time during the upward acceleration phase. After the first adjustment, the longitudinal acceleration decreased from an abnormal value of 1.6m / s² to 1.3m / s², within the baseline range of 0.8-1.2m / s², with a calculated acceleration recovery rate of 75%. The armature response delay was shortened from 60ms to 55ms, with a threshold of 50ms, and the trajectory feature recovery rate was 50%, all of which failed to reach the preset recovery threshold.

[0037] The second adjustment increased the reverse current pulse amplitude to -2.5A and lasted for 120ms. After the adjustment, the longitudinal acceleration decreased to 1.1m / s², with an acceleration recovery rate of 125%; the armature response delay was shortened to 45ms, and the trajectory characteristic recovery rate was 150%, both meeting the recovery threshold. A stable observation period of 3 braking cycles was then established, during which the acceleration stabilized at 1.0-1.2m / s², and the response delay stabilized at 43-47ms, with no abnormal fluctuations, indicating the adjustment was successful.

[0038] The final recorded adjustment parameters are: reverse current pulse amplitude -2.5A, duration 120ms. The recovery process data includes acceleration and response delay change curves after two adjustments. If, when a fault is adjusted to the mechanical limit, the acceleration recovery and trajectory characteristic recovery are still below the threshold, the adjustment is deemed invalid, and dual-brake relay control is triggered.

[0039] Step S500: Based on the type of fault brake and the elevator configuration, select a relay brake with complementary capabilities, calculate the required capacity, optimal handover time and distance of the relay brake, and execute relay control according to the calculated parameters.

[0040] Specifically, when the faulty brake is a drum brake, a disc brake should be selected as the replacement brake; when the faulty brake is a disc brake, a drum brake should be selected as the replacement brake; when the faulty brake is an electromechanical brake, a disc brake should be selected as the replacement brake. Calculate the current remaining braking force of the faulty brake based on real-time braking force data. The formula is: ; in, Dedicated power to the faulty brake. For a fault brake, the coefficient of performance degradation is: for a drum brake, it is the ratio of the actual travel of the brake shoe to the rated travel of the brake shoe; for a disc brake, it is the ratio of the actual piston pressure to the rated piston pressure; and for an electromechanical brake, it is the ratio of the actual current to the rated current of the electromagnet. Calculate the total braking force required for the elevator based on the current operating conditions. The calculation formula for the uniform motion sub-stage is: ; The calculation formulas for the acceleration and deceleration sub-stages near the target layer are as follows: ; in, Given the current elevator load, To accelerate or decelerate the current sub-stage objective. It is the acceleration due to gravity. The coefficient of friction between the car and the guide rail; Calculate the required capacity of the relay brake The formula is: ; To calculate the optimal handover period, based on the acceleration change curve, candidate periods with acceleration fluctuation values ​​less than the acceleration fluctuation threshold within the current sub-stage are selected, and the period with the longest duration among the candidate periods is chosen as the optimal handover period. When the deceleration sub-stage is near the target floor, the distance between the elevator position and the target floor corresponding to the candidate time period must be greater than the safety distance. The formula is: ; in, For real-time running speed; Calculate the optimal junction distance The formula is: ; in, The known response time of the relay brake; according to Send control command, disc brake adjusts pressure to The calculation formula is: ; in, The piston's surface area under stress; Electromechanical brake adjustment current to The calculation formula is: ; in, The conversion coefficient between electric current and electromagnetic force; Drum brake adjustment: The pressure between the brake shoe and the brake drum contact surface is adjusted to... = Send a discharge command to the faulty brake, discharge rate The calculation formula is: ; During the control process, acceleration data, trajectory images, and vibration data are monitored in real time. When all of them meet the requirements, the handover is deemed complete, the control parameters and monitoring data are recorded, and a relay control log is generated.

[0041] In one specific embodiment, when the armature guide shaft of the electromechanical brake fails due to wear, and adjusting the parameters to the mechanical limit is ineffective, step S500 is executed. According to the rules, in the event of a fault in the electromechanical brake, the disc brake is preferentially selected as the replacement. The rated power of this elevator disc brake is 30kN, and the piston force-bearing area is... =0.01m², response time =0.4s; rated power of faulty electromechanical brake =5kN, actual current 2A, rated current 5A, therefore performance degradation coefficient =0.4, remaining braking force =2kN.

[0042] The elevator is currently in the upward acceleration phase, with a load capacity of [missing information]. =2000kg, target acceleration =2.0 m / s², gravitational acceleration =9.8m / s², coefficient of friction between car and guide rail =0.03, total braking force =4.588kN. Based on this, the relay demand capacity can be calculated. =2.588kN.

[0043] The current acceleration fluctuation threshold for the initial acceleration sub-phase is ±0.15 m / s², and candidate time periods with a duration of 2 seconds are selected. =2s; Real-time running speed =1.0m / s, optimal handover distance =0.4m. Disc brakes require pressure adjustment. =0.2588MPa; Send a discharge command to the faulty electromechanical brake, discharge rate =1000N / s.

[0044] During the control process, real-time monitoring showed that the longitudinal acceleration fluctuation was stable within ±0.12 m / s², the disc piston trajectory image showed normal motion, and the peak vibration data was 0.25 g. After three consecutive acquisition cycles of 120 ms, the disc output force stabilized at 2.46-2.72 kN, and the output force of the faulty electromechanical brake dropped to 0.3 kN. The handover was deemed complete, and the system entered the individual control phase for the disc brake. Finally, a relay control log containing force calculations, handover parameters, and monitoring data was generated.

[0045] like Figure 2 The system structure diagram of an elevator brake anomaly diagnosis system based on multimodal perception is shown. This invention provides an elevator brake anomaly diagnosis system based on multimodal perception, comprising: Multimodal data acquisition module: Among them, the acceleration acquisition unit collects acceleration data of the elevator in all operating conditions, divides it into sub-stages, aligns the timestamps, generates curves and records the normal baseline; the visual trajectory acquisition unit sets the ROI according to the associated brake, collects trajectory images before and after abnormal moments, aligns them and then crops and stores them; the auxiliary physical quantity acquisition unit collects brake vibration, temperature and braking force data, and forms a six-dimensional structured dataset after alignment. Anomaly detection and fault matching module: The anomaly detection unit sets an acceleration threshold, divides the load conditions, compares data to determine sub-stage anomalies and associates them with the corresponding brakes; the trajectory processing and feature extraction unit preprocesses the trajectory image and extracts the core motion features of the brakes; the fault matching unit combines motion features and auxiliary data to determine the specific fault type and stores the information. Multimodal cross-validation module: The core correlation validation unit verifies whether the key correlation meets the threshold requirements for the initial fault; the validation result processing unit confirms the fault when the standard is met, and records the problem if the standard is not met, and re-collects data and matches the fault a second time. Fault handling and relay control module: The parameter adjustment unit formulates adjustment plans according to the fault type and monitors the adjustment process in real time; the effect evaluation unit calculates the acceleration and trajectory feature recovery degree, and if the standard is met, the adjustment is judged to be successful; if it is still ineffective after reaching the limit, the relay is triggered; the relay control unit selects the complementary relay brake, calculates the force value and handover parameters, generates control commands and monitors the handover process, and switches to individual control after completion. System Log and Information Management Unit: The information recording unit records structured information throughout the entire process; the log generation unit integrates the recorded data and generates a full-process log.

[0046] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for diagnosing elevator brake anomalies based on multimodal sensing, characterized in that, include: Real-time acquisition of longitudinal acceleration and deceleration data for the entire elevator operation, including both ascending and descending. The real-time collected data is compared with the normal baseline value to determine the acceleration anomaly. The corresponding brake type is associated with the working condition. Specifically, based on the acceleration baseline range of each sub-stage under normal working conditions, the acceleration peak threshold range, acceleration fluctuation threshold range, and deceleration peak threshold range are set for the starting acceleration, constant speed operation, and deceleration near the target layer sub-stages. When the real-time collected data exceeds the corresponding threshold range, the sub-stage is determined to be abnormal. The load is compared with preset light, medium, and heavy load thresholds to classify the operating conditions into light, medium, and heavy load conditions. The logic of the acceleration sub-stage is analyzed: when accelerating upwards, anomalies in the light load condition are associated with the electromechanical brake; anomalies in the medium or heavy load condition are associated with the disc brake. When accelerating downwards, anomalies in the light load condition are associated with the electromechanical brake; anomalies in the medium load condition are preferentially associated with the electromechanical brake; if the electromechanical brake has no abnormal characteristics, it is associated with the disc brake; and anomalies in the heavy load condition are associated with the disc brake. The logic of the constant speed operation sub-stage is analyzed, and anomalies are associated with the brake currently involved in maintaining braking force. The logic of the deceleration sub-stage near the target layer is analyzed: when decelerating upwards, anomalies are associated with the drum brake; when decelerating downwards, anomalies are preferentially associated with the disc brake; if the disc brake parameters are normal, it is associated with the electromechanical brake. After an anomaly is determined, structured information such as the anomaly sub-stage, anomaly type, real-time value, reference range, and associated brake type is recorded. For the associated brake type, a high-speed camera is activated to acquire motion trajectory images, which are time-aligned with the moment of acceleration anomaly; vibration data, temperature data, and braking force data are collected to form a multimodal verification dataset; The system processes trajectory images to extract motion features and matches fault types; it also uses cross-validation with auxiliary modal data to eliminate false positives. Adjust parameters according to the type of fault, monitor acceleration and trajectory images in real time, and determine that the adjustment is successful when the abnormality disappears after parameter adjustment. If the adjustment is still ineffective after reaching the mechanical limit, activate the dual brake relay control. Based on the type of fault brake and the elevator configuration, select a relay brake with complementary capabilities, calculate the required capacity, optimal handover time and distance of the relay brake, and execute relay control according to the calculated parameters.

2. The elevator brake anomaly diagnosis method based on multimodal perception according to claim 1, characterized in that, The real-time acquisition of longitudinal acceleration and deceleration data for the entire elevator's ascent and descent includes: Using a triaxial accelerometer, data is collected for both ascent and descent, divided into three sub-stages: acceleration at start-up, constant speed operation, and deceleration near the target layer. Specifically, longitudinal acceleration data is collected for the acceleration at start-up sub-stage, longitudinal acceleration fluctuation data is collected for the constant speed operation sub-stage, longitudinal deceleration data is collected for the deceleration near the target layer sub-stage, and lateral acceleration fluctuation values ​​are collected for all three sub-stages. The data from each acceleration sensor are timestamped to generate acceleration change curves for each sub-stage, and the acceleration reference range for each sub-stage under normal operating conditions is recorded.

3. The elevator brake anomaly diagnosis method based on multimodal perception according to claim 1, characterized in that, For the associated brake type, the process of activating a high-speed camera to acquire motion trajectory images, and aligning them with the time of acceleration anomalies, includes: Based on the associated brake type, a corresponding ROI is set. When an acceleration anomaly is determined and associated with the brake type, the high-speed camera acquires trajectory images for preset durations before and after the anomaly moment. The image sequence is aligned with the acceleration anomaly moment using an interpolation algorithm. The original image is cropped according to the preset ROI, and the image data is stored in the format of associated brake type, anomaly timestamp, frame number, and ROI parameter.

4. The elevator brake anomaly diagnosis method based on multimodal perception according to claim 1, characterized in that, The collected vibration data, temperature data, and braking force data form a multimodal verification dataset, including: Vibration and temperature data of three types of brakes are collected using vibration sensors and infrared thermometers; pressure data of disc brakes are collected using pressure sensors and converted into braking force data; current data of electromechanical brakes are collected using current sensors and real-time braking force data is calculated based on existing current-electromagnetic force conversion models; and pressure values ​​of the contact surface between the brake shoe and the brake drum of drum brakes are collected using pressure sensors and used as braking force data. Using the moment of acceleration anomaly as the reference time point, vibration data, temperature data, and braking force data are aligned according to the timestamp difference to form a six-dimensional associated dataset of time, acceleration, trajectory image, vibration, temperature, and braking force, which is then stored in a structured manner.

5. The elevator brake anomaly diagnosis method based on multimodal perception according to claim 1, characterized in that, The process involves extracting motion features from the trajectory image and matching fault types, including: The trajectory image is preprocessed, including noise suppression, edge enhancement, and dynamic ROI locking; For drum brakes, identify stagnation and abrupt movement features where the duration of stagnation on one side of the brake shoe exceeds a preset stagnation threshold and the abrupt movement displacement exceeds a preset abrupt movement threshold, and extract the time difference between the contact of the brake shoes on both sides; for disc brakes, extract the piston-brake disc clearance and piston speed fluctuation patterns; for electromechanical brakes, extract the armature trajectory offset from the preset axis and the armature response delay. Analyzing the fault types of drum brakes, when one side of the brake shoe exhibits stagnation and sudden movement characteristics, and the impact amplitude in the vibration data at the corresponding moment exceeds the preset vibration impact threshold, while the temperature rise rate of the temperature data is less than the preset normal temperature rise rate, it is determined to be a local jamming of the brake shoe; when the time difference between the contact of the two brake shoes exceeds the preset contact synchronization allowable value and the lateral acceleration fluctuation value exceeds the preset lateral fluctuation threshold, it is determined to be an abnormal adjustment of the brake shoe gap. Analyzing the fault types of disc brakes, when the piston movement speed fluctuates beyond the preset stable speed range and the pressure data fluctuation amplitude exceeds the preset pressure fluctuation threshold, it is determined that the piston seal ring is aged and stuck; when the gap between the piston and the brake disc is greater than the preset upper limit of normal gap and the braking force data is less than the preset lower limit of rated braking force, it is determined that the hydraulic system pressure is abnormal. Analyzing the fault types of electromechanical brakes, when the armature response delay exceeds the preset delay threshold and the current data is within the preset rated working range of the electromagnet, it is determined to be an electromagnet residual magnetism fault; when the armature trajectory deviates from the preset axis by more than the preset deviation threshold and the vibration data shows noise consistent with the trajectory deviation direction, it is determined to be an armature guide shaft wear. The extracted motion features, matched fault types, and associated logic are stored in the format of brake type, abnormal timestamp, motion feature, fault type, and associated logic.

6. The elevator brake anomaly diagnosis method based on multimodal perception according to claim 1, characterized in that, The step of eliminating false positives through cross-validation using auxiliary modal data includes: When it is initially determined that the brake shoe is partially stuck, verify whether the time difference between the time node of the brake shoe stagnation and sudden movement on one side and the impact signal timestamp of the vibration data is less than the preset time synchronization threshold; when it is initially determined that the brake shoe gap adjustment is abnormal, verify whether the correlation coefficient between the time difference of the brake shoe contact on both sides and the lateral acceleration fluctuation value is greater than the preset correlation threshold. When the initial diagnosis is that the piston seal ring is aging and stuck, verify whether the synchronization deviation of the period and peak time of the piston speed fluctuation and pressure fluctuation is less than the preset period synchronization threshold; when the initial diagnosis is that the hydraulic system pressure is abnormal, verify whether the negative correlation coefficient between the piston and brake disc clearance and the braking force data is less than the preset negative correlation threshold. When it is initially determined that the electromagnet is remanent, verify whether the current data is within the preset rated current range during the armature response delay period; when it is initially determined that the armature guide shaft is worn, verify whether the direction of the armature trajectory offset and the direction of vibration noise meet the preset direction matching threshold. When the core correlation is verified to meet the threshold requirements, the verification is marked as passed. When the core correlation is not verified, the failed item is recorded, trajectory images and auxiliary data are re-acquired, and fault matching is performed.

7. The elevator brake anomaly diagnosis method based on multimodal perception according to claim 1, characterized in that, The process involves adjusting parameters based on the fault type, monitoring acceleration and trajectory images in real time, and determining successful adjustment when the abnormality disappears after parameter adjustment. If adjustment to the mechanical limit is still ineffective, a dual-brake relay control is initiated, including: To address localized brake shoe jamming, adjust the brake shoe clearance compensation parameters; to address abnormal brake shoe clearance adjustment, initiate a synchronized calibration program for both brake shoes, dynamically adjusting the stroke limit parameters of a single brake shoe to achieve the target of synchronized fit; to address piston seal aging and jamming, adjust the hydraulic system pressure compensation coefficient and correct the allowable threshold for piston movement speed fluctuations; to address abnormal hydraulic system pressure, adjust the hydraulic pump output pressure setting; to address electromagnet residual magnetism, introduce reverse current pulse parameters to shorten the armature engagement and release interval; to address armature guide shaft wear, adjust the guide shaft lubrication cycle parameters and correct the allowable offset threshold for the armature trajectory. During the adjustment process, acceleration data and trajectory images are collected for each sub-stage. After each adjustment, the acceleration recovery rate and trajectory feature recovery rate are calculated. When both reach the preset recovery threshold, the adjustment is considered successful, and the final adjustment parameters and recovery process data are recorded. When the parameter adjustment reaches the mechanical limit and both the acceleration recovery rate and trajectory feature recovery rate are less than the preset recovery threshold, the adjustment is considered invalid, triggering dual-brake relay control, and recording the adjustment limit value and unrecovered abnormal features.

8. The elevator brake anomaly diagnosis method based on multimodal perception according to claim 1, characterized in that, The process involves selecting a relay brake with complementary capabilities based on the type of fault brake and elevator configuration, calculating the required capacity, optimal handover time, and distance of the relay brake, and executing relay control according to the calculated parameters, including: When the faulty brake is a drum brake, a disc brake should be selected as the replacement brake; when the faulty brake is a disc brake, a drum brake should be selected as the replacement brake; when the faulty brake is an electromechanical brake, a disc brake should be selected as the replacement brake. Calculate the current remaining braking force of the faulty brake based on real-time braking force data. The formula is: ； in, Dedicated power to the faulty brake. For a faulty brake, the coefficient of performance degradation is: for a drum brake, it is the ratio of the actual travel of the brake shoe to the rated travel of the brake shoe; for a disc brake, it is the ratio of the actual piston pressure to the rated piston pressure; and for an electromechanical brake, it is the ratio of the actual current to the rated current of the electromagnet. Calculate the total braking force required for the elevator based on the current operating conditions. The calculation formula for the uniform motion sub-stage is: ； The calculation formulas for the acceleration and deceleration sub-stages near the target layer are as follows: ； in, Given the current elevator load, To accelerate or decelerate the current sub-stage objective. It is the acceleration due to gravity. The coefficient of friction between the car and the guide rail; Calculate the required capacity of the relay brake The formula is: ； To calculate the optimal handover period, based on the acceleration change curve, candidate periods with acceleration fluctuation values ​​less than the acceleration fluctuation threshold within the current sub-stage are selected, and the period with the longest duration among the candidate periods is chosen as the optimal handover period. When the deceleration sub-stage is near the target floor, the distance between the elevator position and the target floor corresponding to the candidate time period must be greater than the safety distance. The formula is: ； in, For real-time running speed; Calculate the optimal junction distance The formula is: ； in, The known response time of the relay brake; according to Send control command, disc brake adjusts pressure to The calculation formula is: ； in, The piston's surface area under stress; Electromechanical brake adjustment current to The calculation formula is: ； in, The conversion coefficient between electric current and electromagnetic force; Drum brake adjustment: The pressure between the brake shoe and the brake drum contact surface is adjusted to... = Send a discharge command to the faulty brake, discharge rate The calculation formula is: ； During the control process, acceleration data, trajectory images, and vibration data are monitored in real time. When all of them meet the requirements, the handover is deemed complete, the control parameters and monitoring data are recorded, and a relay control log is generated.

9. A multimodal sensing-based elevator brake anomaly diagnosis system, using the multimodal sensing-based elevator brake anomaly diagnosis method according to any one of claims 1-8, characterized in that, include: Multimodal data acquisition module: Among them, the acceleration acquisition unit collects acceleration data of the elevator in all operating conditions, divides it into sub-stages, aligns the timestamps, generates curves and records the normal baseline; the visual trajectory acquisition unit sets the ROI according to the associated brake, collects trajectory images before and after abnormal moments, aligns them and then crops and stores them; the auxiliary physical quantity acquisition unit collects brake vibration, temperature and braking force data, and forms a six-dimensional structured dataset after alignment. Anomaly detection and fault matching module: The anomaly detection unit sets an acceleration threshold, divides the load conditions, compares data to determine sub-stage anomalies and associates them with the corresponding brakes; the trajectory processing and feature extraction unit preprocesses the trajectory image and extracts the core motion features of the brakes; the fault matching unit combines motion features and auxiliary data to determine the specific fault type and stores the information. Multimodal cross-validation module: The core correlation validation unit verifies whether the key correlation meets the threshold requirements for the initial fault; the validation result processing unit confirms the fault when the standard is met, and records the problem if the standard is not met, and re-collects data and matches the fault a second time. Fault handling and relay control module: The parameter adjustment unit formulates adjustment plans according to the fault type and monitors the adjustment process in real time; the effect evaluation unit calculates the acceleration and trajectory feature recovery degree, and if the standard is met, the adjustment is judged to be successful; if it is still ineffective after reaching the limit, the relay is triggered; the relay control unit selects the complementary relay brake, calculates the force value and handover parameters, generates control commands and monitors the handover process, and switches to individual control after completion. System Log and Information Management Unit: The information recording unit records structured information throughout the entire process; the log generation unit integrates the recorded data and generates a full-process log.

Images