Elevator ascending overspeed protection device fault prediction method and system based on big data

Abstract

The invention provides an elevator ascending overspeed protection device fault prediction method and system based on big data, and relates to the technical field of elevator safety monitoring, and the method comprises the steps: setting a thermocouple array on the surface of a brake to collect temperature data, constructing a temperature distribution diagram, and extracting a hot spot region; calculating a hotspot migration track and stress distribution by using a big data analysis platform; fitting a damage development curve based on the fatigue damage degree; and space-time correlation analysis is carried out through a multi-point monitoring network, the fault source position and the expansion trend are determined, and fault early warning is realized. Potential faults of the elevator ascending overspeed protection device can be predicted in advance, and the operation safety of the elevator is improved.

Description

Technical Field

[0001] The present invention relates to elevator safety monitoring technology, and particularly to a method and system for predicting faults of an overspeed protection device for elevator upward travel based on big data. Background Art

[0002] The overspeed protection device for elevator upward travel is a key component to ensure the safe operation of the elevator, and the reliability of its brake directly affects the safe operation of the elevator. At present, the prediction of brake faults mainly relies on regular inspections and manual diagnoses, making it difficult to detect potential fault hazards in a timely manner, and the detection results are greatly affected by human factors.

[0003] With the development of big data and artificial intelligence technologies, by arranging sensors on the surface of the brake to collect real-time operating state data and combining with historical fault data for analysis, the fault trend of the brake can be effectively predicted. However, the existing fault prediction methods mainly focus on single-point temperature and stress data, and cannot accurately reflect the internal stress distribution and fatigue damage development law of the brake, resulting in low prediction accuracy.

[0004] During the actual operation process, the migration characteristics of the hot spot area on the surface of the brake are closely related to the internal stress distribution. However, the existing technologies lack in-depth analysis of the migration law of the hot spot area and cannot effectively identify the location of the fault source and the expansion trend. Therefore, there is an urgent need for a method for predicting faults of an overspeed protection device for elevator upward travel based on big data, which can accurately predict brake faults by analyzing the migration characteristics of the hot spot area and the fatigue damage development law. Summary of the Invention

[0005] The embodiments of the present invention provide a method and system for predicting faults of an overspeed protection device for elevator upward travel based on big data, which can solve the problems in the existing technologies.

[0006] In the first aspect of the embodiments of the present invention, a method for predicting faults of an overspeed protection device for elevator upward travel based on big data is provided, including: Setting a thermocouple array on the surface of the brake of the overspeed protection device for elevator upward travel, collecting the surface temperature data of the brake, performing feature matching on the temperature data with a historical data set, constructing a real-time temperature distribution map of the brake surface, extracting the hot spot area where the temperature in the temperature distribution map exceeds a preset temperature threshold, and calculating the area and temperature gradient of the hot spot area; Using a big data analysis platform, according to the position change of the hot spot area at different times, combining with the migration characteristics in the historical fault database, calculating the migration trajectory and migration rate of the hot spot area, calculating the internal stress distribution of the brake based on the temperature gradient and the migration rate, and obtaining the stress peak value in the stress distribution; When the stress peak continues to increase, call the fault sample library, calculate the fatigue damage degree of the brake material according to the stress peak and the area of the hot spot region, and based on the fatigue damage degree and historical fault sample data, use a machine learning algorithm to fit and obtain the fatigue damage development curve; According to the change trend of the fatigue damage development curve, deploy a multi-point monitoring network on the migration trajectory of the hot spot region, and collect temperature gradient and stress peak data in real time. Conduct spatio-temporal correlation analysis with the fatigue damage development curve to determine the location and expansion trend of the fault source, and output a fault warning message.

[0007] In an alternative embodiment, Set a thermocouple array on the surface of the brake of the elevator overspeed protection device, collect the surface temperature data of the brake, and perform feature matching on the temperature data with the historical data set to construct a real-time temperature distribution map of the brake surface, including: Divide the brake surface into a core monitoring area, a transition area, and an edge monitoring area. Adopt a partition dynamic sampling method to collect the temperature data of the thermocouple array, adjust the sampling frequency according to the temperature change rate of each area, and perform weighted averaging on the temperature data according to the monitoring area position to obtain noise-reduced temperature data; Calculate the temperature gradient and temperature distribution curvature of the noise-reduced temperature data. Identify temperature mutation abnormal points based on the temperature gradient, and identify temperature fluctuation abnormal points based on the temperature distribution curvature; Match the time domain characteristics of the temperature mutation abnormal points and temperature fluctuation abnormal points with the fault characteristics in the historical data set to obtain the confidence level of the abnormal points; Correct the temperature abnormal points according to the confidence level of the abnormal points, perform local weighted fusion on the corrected temperature data and the surrounding normal temperature data to obtain corrected temperature data, and perform cubic spline interpolation on the corrected temperature data according to the spatial position to generate a real-time temperature distribution map of the brake surface.

[0008] In an alternative embodiment, Extract the hot spot regions in the temperature distribution map where the temperature exceeds the preset temperature threshold, and calculate the hot spot region area and temperature gradient, including: Calculate the friction power density distribution based on the instantaneous acceleration and braking torque of the brake. Perform partition weighting on the temperature distribution map according to the friction power density distribution, calculate the temperature change rate of each area, combine the temperature change rate with the brake friction torque to determine the temperature threshold, and mark the regions in the temperature distribution map where the temperature exceeds the temperature threshold as the hot spot regions to be processed; Identify the temperature extreme points in the hotspot area to be processed, classify the temperature extreme points according to the friction power density, and take the temperature extreme points as seed points from high to low level. Calculate the temperature similarity, gradient similarity, and spatial distance similarity between the points to be expanded and the grown area, adjust the weight coefficients of each similarity according to the friction power density distribution for weighting, determine the region growth boundary, and obtain the hotspot area. Perform curve fitting on the boundary points of the hotspot area to obtain a closed curve, calculate the area of the hotspot area according to the closed curve, determine the area threshold according to the friction power density level corresponding to the hotspot area, and identify the hotspot area exceeding the area threshold as the concerned hotspot area. Calculate the temperature gradient of the concerned hotspot area using a kernel function, where the bandwidth of the kernel function is inversely proportional to the friction power density of the corresponding area, and generate the temperature gradient distribution of the concerned hotspot area.

[0009] In an alternative embodiment, Using a big data analysis platform, according to the position changes of the hotspot area at different times, combined with the migration characteristics in the historical fault database, calculate the migration trajectory and migration rate of the hotspot area, calculate the internal stress distribution of the brake based on the temperature gradient and migration rate, and obtain the stress peak in the stress distribution, including: Use the big data analysis platform to collect the temperature distribution data of the hotspot area at different times, calculate the centroid coordinate sequence of the temperature distribution, construct a displacement field according to the centroid coordinate sequence, and at the same time establish the boundary curve of the hotspot area, and convert the deformation amount of the boundary curve at adjacent times into a strain matrix. Calculate the principal strain direction and strain amount according to the strain matrix, determine the principal strain direction as the migration direction, project the displacement field at adjacent times onto the migration direction to obtain the migration distance, construct the migration trajectory of the hotspot area according to the migration distance, and take the derivative of the migration distance with respect to time to obtain the migration rate. Match the migration trajectory with the characteristic patterns in the historical fault database, correct the migration rate according to the matching degree, and perform spatial mapping on the temperature gradient using the corrected migration rate to obtain the temperature field distribution considering the migration effect. Based on the temperature field distribution, divide the stress calculation grid, decompose the migration rate into tangential velocity and normal velocity at the grid nodes, calculate the shear stress using the tangential velocity, calculate the compressive stress using the normal velocity, superimpose the shear stress and compressive stress to obtain the internal stress distribution of the brake, and identify the stress singular points according to the spatial second derivative of the stress distribution, and determine the stress value at the stress singular points as the stress peak.

[0010] In an alternative embodiment, When the stress peak continues to increase, call the fault sample library, calculate the fatigue damage degree of the brake material according to the stress peak and the hot spot area, and based on the fatigue damage degree and historical fault sample data, use the dynamic damage evolution state equation to fit the fatigue damage development curve, including: Collect the stress peak and hot spot area data, call the historical fault data in the fault sample library, segment the historical fault data according to the growth interval of the stress peak, calculate the change rate of the stress peak, the growth rate of the hot spot area and its duration in each segment, and obtain the stress-area evolution characteristic parameters; According to the stress-area evolution characteristic parameters, calculate the fatigue damage increment in each segment, and construct a time series weight matrix with the fatigue damage increment and the duration of the corresponding segment; When the stress peak continues to increase, match the change rate of the current stress peak and hot spot area with the stress-area evolution characteristic parameters, determine the current damage evolution stage according to the matching result, calculate the damage accumulation rate of the current stage based on the time series weight matrix, and calculate the fatigue damage degree according to the damage accumulation rate; Perform weighted combination on the fatigue damage degree and the time series weight matrix to obtain a corrected damage value considering the historical cumulative effect, perform piecewise linear fitting on the corrected damage value, and use the time series weight matrix for weighted processing during the fitting process to obtain a fatigue damage development curve reflecting the damage development trend.

[0011] In an optional embodiment, According to the change trend of the fatigue damage development curve, arrange a multi-point monitoring network on the migration trajectory of the hot spot area, and collect the temperature gradient and stress peak data in real time, perform spatio-temporal correlation analysis with the fatigue damage development curve, determine the fault source location and expansion trend, and output the fault warning information, including: Perform piecewise linearization on the fatigue damage development curve, extract the slope mutation points of the fatigue damage development curve, divide the fatigue damage stage based on the slope mutation points, analyze the migration trajectory and speed change of the hot spot area in each stage, and connect the boundaries of the hot spot area in each stage to construct a dynamic envelope surface; On the dynamic envelope surface, determine the main monitoring point according to the position of the maximum value of the temperature gradient change rate, and based on the spatial distribution law of the stress peak, construct an annular secondary monitoring point array centered on the main monitoring point, establish a multi-objective optimization function including the temperature gradient change rate term, the stress gradient change rate term, and the monitoring point spacing term, and obtain the optimal layout scheme of the monitoring network by iteratively solving the multi-objective optimization function; Adopt an adaptive sliding time window to collect temperature gradient and stress peak data in the monitoring network in real time, project them onto the corresponding stages of the fatigue damage development curve in chronological order, construct a feature matrix including spatial position and time series, and extract the main eigenvector and secondary eigenvector through singular value decomposition; Calculate the time drift rate and spatial drift direction of the main eigenvector and secondary eigenvector. When the time drift rate exceeds the dynamic threshold corresponding to the fatigue damage stage, fit the fault expansion trend according to the spatial drift direction and temperature stress distribution law, combine the historical fault database to determine the fault source location, and output a fault warning message.

[0012] In the second aspect of the embodiments of the present invention, a fault prediction system for an elevator overspeed protection device based on big data is provided, including: The first unit is used to set a thermocouple array on the surface of the brake of the elevator overspeed protection device, collect the surface temperature data of the brake, perform feature matching on the temperature data with the historical data set, construct a real-time temperature distribution map of the brake surface, extract the hot spot area where the temperature exceeds the preset temperature threshold in the temperature distribution map, and calculate the hot spot area and temperature gradient; The second unit is used to utilize the big data analysis platform, calculate the migration trajectory and migration rate of the hot spot area according to the position change of the hot spot area at different times, combine the migration characteristics in the historical fault database, calculate the internal stress distribution of the brake based on the temperature gradient and migration rate, and obtain the stress peak in the stress distribution; The third unit is used to, when the stress peak continues to increase, call the fault sample library, calculate the fatigue damage degree of the brake material according to the stress peak and the hot spot area, and use the machine learning algorithm to fit the fatigue damage development curve based on the fatigue damage degree and historical fault sample data; The fourth unit is used to, according to the change trend of the fatigue damage development curve, deploy a multi-point monitoring network on the migration trajectory of the hot spot area, collect temperature gradient and stress peak data in real time, perform spatio-temporal correlation analysis with the fatigue damage development curve, determine the fault source location and expansion trend, and output a fault warning message.

[0013] In the third aspect of the embodiments of the present invention, an electronic device is provided, including: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0014] In the fourth aspect of the embodiments of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0015] In this embodiment, by arranging a thermocouple array on the surface of the brake of the elevator overspeed protection device, the change of temperature distribution is monitored in real time, and the migration characteristics of the hot spot area are analyzed in combination with the big data analysis platform, so as to accurately identify potential fault sources, predict the possibility of fault occurrence in advance, and effectively avoid the safety hazards caused by the lag of traditional detection methods. The dynamic damage evolution state equation is used to fit the fatigue damage development curve, and key parameters are collected in real time through a multi-point monitoring network for spatio-temporal correlation analysis. It can not only determine the location of the fault source, but also predict the fault expansion trend, improve the accuracy and reliability of fault prediction, and provide a scientific basis for elevator maintenance. By combining big data analysis technology with traditional mechanical fault diagnosis methods, a complete fault prediction system for the elevator overspeed protection device is constructed, realizing the full-process intelligent management from data collection, feature extraction to fault warning, greatly improving the elevator safety operation guarantee ability, and reducing the maintenance cost and safety accident risk. Brief Description of the Drawings

[0016] Figure 1 It is a schematic flow chart of the fault prediction method for the elevator overspeed protection device based on big data according to the embodiment of the present invention; Figure 2 It is a three-dimensional schematic diagram of the relationship between the migration of the hot spot area and the stress distribution; Figure 3 It is a schematic diagram comparing the prediction accuracies of fatigue damage assessment methods in different damage stages. Detailed Embodiments

[0017] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] The technical solutions of the present invention will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0019] Figure 1 It is a schematic flow chart of the fault prediction method for the elevator overspeed protection device based on big data according to the embodiment of the present invention, as Figure 1 shown, the method includes: A thermocouple array is set on the surface of the brake of the elevator overspeed protection device to collect the temperature data of the brake surface. The temperature data is subjected to feature matching with the historical data set to construct a real-time temperature distribution map of the brake surface. The hot spot areas where the temperature exceeds the preset temperature threshold are extracted from the temperature distribution map, and the area and temperature gradient of the hot spot areas are calculated; Using a big data analysis platform, according to the position changes of the hot spot areas at different times, combined with the migration characteristics in the historical fault database, the migration trajectory and migration rate of the hot spot areas are calculated. Based on the temperature gradient and migration rate, the internal stress distribution of the brake is calculated, and the stress peak value in the stress distribution is obtained; When the stress peak value continues to increase, the fault sample library is called. According to the stress peak value and the area of the hot spot area, the fatigue damage degree of the brake material is calculated. Based on the fatigue damage degree and the historical fault sample data, the fatigue damage development curve is fitted by using the dynamic damage evolution state equation; According to the change trend of the fatigue damage development curve, a multi-point monitoring network is arranged on the migration trajectory of the hot spot area, and the temperature gradient and stress peak value data are collected in real time. The spatio-temporal correlation analysis is carried out with the fatigue damage development curve to determine the fault source position and the expansion trend, and the fault warning information is output.

[0020] In an optional implementation manner, setting a thermocouple array on the surface of the brake of the elevator overspeed protection device, collecting the temperature data of the brake surface, and performing feature matching of the temperature data with the historical data set to construct a real-time temperature distribution map of the brake surface includes: The surface of the brake is divided into a core monitoring area, a transition area and an edge monitoring area. The temperature data of the thermocouple array is collected by using a partition dynamic sampling method, the sampling frequency is adjusted according to the temperature change rate of each area, and the temperature data is weighted and averaged according to the monitoring area position to obtain noise-reduced temperature data; Calculate the temperature gradient and temperature distribution curvature of the noise-reduced temperature data, identify the temperature mutation abnormal points based on the temperature gradient, and identify the temperature fluctuation abnormal points based on the temperature distribution curvature; match the time domain characteristics of the temperature mutation abnormal points and temperature fluctuation abnormal points with the fault characteristics in the historical data set to obtain the confidence level of the abnormal points; According to the confidence level of the abnormal points, the temperature abnormal points are corrected, the corrected temperature data is locally weighted and fused with the surrounding normal temperature data to obtain corrected temperature data, and the corrected temperature data is subjected to cubic spline interpolation according to the spatial position to generate a real-time temperature distribution map of the brake surface.

[0021] In the specific implementation process, a thermocouple array is first set on the brake surface, and it is divided into three areas according to the importance of the brake surface: core monitoring area, transition area and edge monitoring area. The core monitoring area refers to the main friction area where the brake contacts the brake wheel. The temperature change in this area is more drastic, and the thermocouple arrangement density is 2 per square centimeter; the transition area refers to the area extending 2-5cm outward from the core area, and the thermocouple arrangement density is 1 per square centimeter; the edge monitoring area refers to the outer area of ​​the brake, and the thermocouple arrangement density is 1 per 5 square centimeters.

[0022] The temperature data of the thermocouple array is collected by partitioned dynamic sampling, that is, the sampling frequency is dynamically adjusted according to the temperature change rate of each area. Specifically, when the temperature change rate in the core monitoring area exceeds 5℃ / s, the sampling frequency is increased to 20Hz; when the change rate is between 2-5℃ / s, the sampling frequency is 10Hz; when the change rate is less than 2℃ / s, the sampling frequency is 5Hz. For the transition zone, when the temperature change rate exceeds 3℃ / s, the sampling frequency is 10Hz; when the change rate is between 1-3℃ / s, the sampling frequency is 5Hz; when the change rate is less than 1℃ / s, the sampling frequency is 2Hz. For the edge monitoring area, the sampling frequency is fixed at 1Hz. This partitioned dynamic sampling method can obtain more accurate temperature change information in key areas while reducing data processing pressure.

[0023] The collected temperature data is weighted averaged according to the location of the monitoring area to reduce the impact of noise. The core monitoring area uses Gaussian weighted averaging. For each thermocouple measurement point, the temperature value of the surrounding 8 adjacent points is weighted averaged. The weight coefficient is inversely proportional to the distance. Specifically, the weight of the center point is 0.5, and the weight of the adjacent point is 0.5 / 8. The transition zone uses simple arithmetic averaging to average the temperature values ​​of each thermocouple measurement point and the adjacent 4 points. Due to the low density of thermocouples in the edge monitoring area, the original measurement value is used directly. After processing in this way, the noise-reduced temperature data is obtained.

[0024] Calculate the temperature gradient and the curvature of temperature distribution based on the noise-reduced temperature data to identify abnormal points. The temperature gradient reflects the rate of temperature change and is obtained by dividing the temperature difference between adjacent measurement points by the distance. For example, at a certain moment, the distance between two adjacent points in the core area is 0.7 cm, and the temperatures are 125 °C and 108 °C respectively. Then the temperature gradient is (125 - 108) / 0.7 = 24.3 °C / cm. When the temperature gradient at a certain point exceeds the threshold of 30 °C / cm, it is marked as an abnormal point of temperature mutation. The curvature of temperature distribution reflects the change of the temperature change trend and is obtained by calculating the change rate of the temperature gradient. When the temperature curvature in a certain area exceeds the threshold of 5 °C / cm², it is marked as an abnormal point of temperature fluctuation. For example, the temperatures of three consecutive points are 105 °C, 115 °C, and 110 °C, and the adjacent distances are all 0.8 cm. Then the temperature gradients are 12.5 °C / cm and -6.25 °C / cm respectively, and the temperature curvature is about 23.4 °C / cm², which exceeds the threshold and is marked as an abnormal point of fluctuation.

[0025] Match the time-domain characteristics of the detected abnormal points of temperature mutation and temperature fluctuation with the fault characteristics in the historical dataset. The time-domain characteristics include the duration of the abnormal point, the amplitude of temperature change, the change rate, and the spatial distribution pattern of the abnormal point, etc. The historical dataset contains the feature templates of known fault types, such as the characteristics of faults like brake overheating, thermal fatigue cracks, and uneven wear. Use the feature distance calculation method to evaluate the similarity between the current abnormal point and the historical fault characteristics to obtain the confidence level of the abnormal point. For example, the temperature of an abnormal point is 145 °C, the duration is 30 s, and the spatial distribution is circular with a radius of about 2 cm. The similarity with the characteristics of the local overheating fault of the brake in the historical data is 85%. Then the confidence level of this abnormal point is set to 0.85.

[0026] Correct the temperature abnormal points according to the confidence level of the abnormal points. For the abnormal points with a confidence level greater than 0.8, they are regarded as real abnormalities and the original temperature data are retained; for the abnormal points with a confidence level between 0.5 and 0.8, partial correction is carried out, and the new temperature value is the weighted average of the original temperature value and the average temperature of the surrounding normal points, with the weights being the confidence level and (1 - confidence level) respectively; for the abnormal points with a confidence level lower than 0.5, they are regarded as noise interference and are replaced by the average temperature of the surrounding normal points. For example, the temperature of an abnormal point is 150 °C, the confidence level is 0.6, and the average temperature of the surrounding normal points is 120 °C. Then the corrected temperature is 150×0.6 + 120×0.4 = 138 °C.

[0027] The corrected temperature data is locally weighted and fused with the surrounding normal temperature data to achieve smooth transition. For each correction point, taking this point as the center, the points within a radius of 5 cm are weighted and fused according to the inverse of the distance to obtain the corrected temperature data. Specifically, for a measurement point at a distance of r cm from the center point, its weight is (5 - r) / 5. In this way, unnatural jumps in the corrected temperature distribution can be avoided.

[0028] Finally, cubic spline interpolation is performed on the corrected temperature data according to the spatial position to generate a high-resolution real-time temperature distribution map of the brake surface. During the interpolation process, taking the actual layout positions of the thermocouples as nodes, 10 calculation points are inserted between every two adjacent nodes to ensure that the resolution of the generated temperature distribution map reaches the millimeter level. The generated temperature distribution map uses different colors to represent different temperature ranges, such as blue for below 80°C, green for 80 - 100°C, yellow for 100 - 120°C, orange for 120 - 140°C, and red for above 140°C, visually showing the temperature distribution on the brake surface and providing a basis for the safety monitoring of the elevator overspeed protection device.

[0029] In this embodiment, by setting a thermocouple array on the brake surface and introducing a partition dynamic sampling mechanism, accurate perception of temperature changes in different regions can be achieved, effectively improving the timeliness and accuracy of temperature acquisition. Through weighted average noise reduction processing of the original temperature data and combined with the joint analysis of temperature gradient and temperature distribution curvature, various abnormal states such as sudden temperature rise and periodic fluctuations can be identified, enhancing the ability to recognize early fault signs. Further introducing a confidence matching and abnormal point correction mechanism can significantly reduce the interference of accidental anomalies or acquisition errors on the results of temperature data analysis, improving the stability and reliability of the data. Finally, the temperature distribution map generated by spatial interpolation reconstruction can truly reflect the thermal evolution process on the brake surface, providing high-precision input data for subsequent stress modeling and fault prediction, thereby enhancing the intelligent perception and predictive control level of the entire elevator overspeed protection system.

[0030] In an alternative embodiment, extracting the hot spot areas in the temperature distribution map where the temperature exceeds a preset temperature threshold and calculating the area and temperature gradient of the hot spot areas includes: Calculating the friction power density distribution based on the instantaneous acceleration and braking torque of the brake, performing partition weighting on the temperature distribution map according to the friction power density distribution, calculating the temperature change rate of each region, combining the temperature change rate with the braking friction torque of the brake to determine the temperature threshold, and marking the regions in the temperature distribution map where the temperature exceeds the temperature threshold as the hot spot areas to be processed; Identify the temperature extreme points in the hot spot area to be processed, classify the temperature extreme points according to the friction power density, and take the temperature extreme points as seed points from high to low level. Calculate the temperature similarity, gradient similarity and spatial distance similarity between the points to be expanded and the grown area, adjust the weight coefficients of each similarity according to the friction power density distribution for weighting, determine the region growth boundary, and obtain the hot spot area; Perform curve fitting on the boundary points of the hot spot area to obtain a closed curve, calculate the area of the hot spot area according to the closed curve, determine the area threshold according to the friction power density level corresponding to the hot spot area, and determine the hot spot area exceeding the area threshold as the concerned hot spot area; Calculate the temperature gradient of the concerned hot spot area by using a kernel function, where the bandwidth of the kernel function is inversely proportional to the friction power density of the corresponding area, and generate the temperature gradient distribution of the concerned hot spot area.

[0031] This embodiment provides a method for predicting faults of an elevator overspeed protection device based on big data. In this method, the specific steps of extracting the hot spot area where the temperature in the temperature distribution map exceeds the preset temperature threshold and calculating the area and temperature gradient of the hot spot area are as follows.

[0032] When calculating the friction power density distribution based on the instantaneous acceleration and braking torque of the brake, the acceleration data and torque data of the elevator brake in different working states can be collected. Obtain the real-time data of multiple measuring points on the surface of the brake through the sensor network, record the operating state of the brake within 30 minutes at a sampling interval of 0.1 s. For a typical traction elevator, the instantaneous acceleration range of the brake is 0.5 to 2.5 m / s², and the braking torque range is 200 to 1200 N·m. Substitute these data into the calculation to obtain the friction power density distribution, and the typical value is between 120 and 580 W / cm 2 ². Perform zonal weighting on the temperature distribution map according to the friction power density distribution. The surface of the brake can be divided into three power density regions: high, medium, and low, and the weight coefficients are set to 1.5, 1.0, and 0.7 respectively. When calculating the temperature change rate of each region, the differential calculation can be performed through the temperature data of 5 consecutive time points to obtain the temperature change trend over time. In the normal working state, the temperature change rate is usually between 0.8 and 3.5 degrees Celsius per minute. Combine the temperature change rate with the brake friction torque to determine the temperature threshold. The reference threshold can be set to 85 degrees Celsius and dynamically adjusted according to the actual value of the friction torque. The adjustment coefficient is 0.05 multiplied by the percentage of the friction torque deviating from the rated value. Mark the areas in the temperature distribution map where the temperature exceeds the temperature threshold as the hot spot areas to be processed, and these areas may have potential fault risks.

[0033] When identifying temperature extreme points in the hotspot area to be processed, the peak detection algorithm is used to find the local highest temperature points in the temperature distribution map. In typical cases, 5 to 12 extreme points may be detected, and the temperatures of these points are usually 2 to 8 degrees Celsius higher than those of the surrounding areas. The temperature extreme points are classified according to the friction power density, and three levels can be set: Level A (greater than 450 W / cm 2 ), Level B (300 to 450 W / cm 2 ), and Level C (less than 300 W / cm 2 ). Starting from Level A, the temperature extreme points are used as seed points for region growing in descending order of level. When calculating the temperature similarity between the points to be expanded and the grown region, a similarity function can be defined. When the temperature difference is less than 3 degrees Celsius, the similarity is high; when the temperature difference is between 3 and 6 degrees Celsius, the similarity is medium; when the temperature difference is greater than 6 degrees Celsius, the similarity is low. The gradient similarity is calculated based on the direction and magnitude of the temperature change. When the gradient direction difference is less than 30 degrees and the magnitude difference is less than 20%, the similarity is high. The spatial distance similarity is calculated based on the Euclidean distance. When the distance is less than 5 millimeters, the similarity is high. The weight coefficients of each similarity are adjusted according to the friction power density distribution for weighting. In the high power density region, the weights of the temperature similarity, gradient similarity, and spatial distance similarity are 0.5, 0.3, and 0.2 respectively; in the medium power density region, the weights are 0.4, 0.4, and 0.2 respectively; in the low power density region, the weights are 0.3, 0.5, and 0.2 respectively. When determining the region growing boundary, the points with a comprehensive similarity greater than 0.75 are included in the region, the points less than 0.4 are excluded, and the points between the two are determined according to their distances from the seed points. In this way, the hotspot areas are obtained, and these areas represent continuous regions with temperature anomalies.

[0034] When fitting a closed curve to the boundary points of the hotspot area, the Bezier curve fitting algorithm can be used. For each hotspot area, 8 to 12 boundary feature points are selected and fitted by a cubic Bezier curve, and the fitting accuracy is controlled within 1 millimeter. The area of the hotspot area can be calculated according to the closed curve by applying Green's formula. According to the friction power density level corresponding to the hotspot area, the area threshold is determined. The area threshold for Level A area is 25 cm 2 , the area threshold for Level B area is 40 cm 2 , and the area threshold for Level C area is 60 cm 2 . The hotspot areas exceeding the area threshold are determined as the hotspot areas of concern, and these areas need to be monitored key points.

[0035] When calculating the temperature gradient of the concerned hot spot area using a kernel function, a Gaussian kernel function can be used. The bandwidth of the kernel function is inversely proportional to the friction power density of the corresponding area, and the calculation formula is bandwidth = reference bandwidth divided by the ratio of the friction power density to the reference power density. The reference bandwidth is set to 15 mm, and the reference power density is 300 W / cm 2 . The temperature distribution is smoothed by the kernel function, and then the spatial derivative is calculated to obtain the temperature gradient. When generating the temperature gradient distribution of the concerned hot spot area, the gradient values can be mapped onto a color map, with red representing high-gradient areas and blue representing low-gradient areas. In practical applications, the temperature gradient is usually between 0.5 and 4.5 degrees Celsius per millimeter. The larger the gradient, the more concentrated the thermal stress and the higher the failure risk.

[0036] In an actual monitoring case of an elevator overspeed protection device, after an elevator brake has been continuously operating for 4 hours, one A-level hot spot area is detected, with an area of 32 cm 2 , the highest temperature reaches 93 °C, and the maximum temperature gradient is 3.8 °C / mm. Through historical data analysis, it is found that when the gradient value exceeds 3.5 °C / mm and the duration exceeds 30 minutes, the probability of the brake failing increases to 85%. Based on this warning mechanism, the system predicts 48 hours in advance that the elevator overspeed protection device may fail, notifies the maintenance personnel for inspection, and avoids the occurrence of safety accidents.

[0037] In this embodiment, by introducing the friction power density distribution to conduct physical correlation analysis on the temperature data, it is possible to more accurately identify the heat accumulation areas generated by the brake under different working conditions and improve the recognition accuracy of abnormal temperature rise parts. Using the hierarchical and region growing algorithm for temperature extreme points, combined with the multi-dimensional similarity weight adjustment mechanism, the adaptive recognition of hot spot areas and the dynamic boundary judgment are realized, ensuring that the hot spot extraction results are highly consistent with the actual heat load distribution. Further, through closed curve fitting and area threshold judgment, it is possible to effectively screen out the key hot areas with potential failure risks and improve the pertinence and reliability of early warning. At the same time, calculating the temperature gradient based on the kernel function adaptively adjusted by the friction power density enhances the expression ability of the internal temperature change trend in the hot spot area, provides high-resolution thermal characteristics input for subsequent stress modeling and fatigue analysis, and thus realizes more refined fault prediction and condition assessment.

[0038] In an alternative implementation manner, using a big data analysis platform, according to the position changes of the hot spot area at different times, combined with the migration characteristics in the historical failure database, calculate the migration trajectory and migration rate of the hot spot area, calculate the internal stress distribution of the brake based on the temperature gradient and the migration rate, and obtain the stress peak value in the stress distribution, including: Use a big data analysis platform to collect temperature distribution data of hot spots at different times, calculate the centroid coordinate sequence of the temperature distribution, construct a displacement field according to the centroid coordinate sequence, and at the same time establish the boundary curve of the hot spot area. Convert the deformation amount of the boundary curve at adjacent times into a strain matrix; Calculate the principal strain direction and strain amount according to the strain matrix, determine the principal strain direction as the migration direction, project the displacement field at adjacent times onto the migration direction to obtain the migration distance, construct the migration trajectory of the hot spot area according to the migration distance, and take the derivative of the migration distance with respect to time to obtain the migration rate; Match the migration trajectory with the characteristic patterns in the historical fault database, correct the migration rate according to the matching degree, and perform spatial mapping on the temperature gradient using the corrected migration rate to obtain the temperature field distribution considering the migration effect; Based on the temperature field distribution, divide the stress calculation grid, decompose the migration rate into tangential velocity and normal velocity at the grid nodes, calculate the shear stress using the tangential velocity, calculate the compressive stress using the normal velocity, superimpose the shear stress and the compressive stress to obtain the internal stress distribution of the brake, and identify the stress singularity points according to the spatial second derivative of the stress distribution. Determine the stress value at the stress singularity point as the stress peak.

[0039] The present invention provides a method for calculating the migration trajectory of the hot spot area of the brake and obtaining the stress peak based on big data analysis. In practical applications, hot spot areas will be generated during the operation of the brake, and the migration characteristics of these hot spot areas are closely related to the fault modes of the brake.

[0040] The big data analysis platform collects the temperature distribution data on the surface of the brake, scans the surface of the brake at a frequency of 10 frames per second through an infrared thermal imager to obtain temperature matrix data. For a typical brake disc, a temperature distribution map of 300×300 pixels can be obtained, and each pixel point corresponds to an area with an actual size of about 0.5mm×0.5mm. After the temperature data collection is completed, the system performs threshold segmentation on the temperature matrix and identifies the area with a temperature higher than 350°C as the hot spot area.

[0041] For the identified hot spot area, calculate the centroid coordinates of its temperature distribution. Taking the center of the brake disc as the coordinate origin, establish a polar coordinate system, and the centroid coordinates of the hot spot area can be expressed as (r, θ). In a continuous time series, such as the centroid coordinates at t1 = 0s are (120mm, 45°), and the centroid coordinates at t2 = 0.1s are (122mm, 47°). By recording these coordinate points, a centroid coordinate sequence is constructed.

[0042] Based on the centroid coordinate sequence, a displacement field is constructed. The displacement field describes the movement of the hot spot area on the two-dimensional plane. For two adjacent time instants t1 and t2, the displacement vector can be determined by the difference between the two centroid coordinates. Meanwhile, the system determines the boundary curve of the hot spot area through the temperature isocontour method, and takes the isothermal line with a temperature of 300 °C as the boundary of the hot spot area.

[0043] By comparing the boundary curves of adjacent time instants, the displacement vectors of the boundary points are calculated to construct the boundary deformation field. The boundary deformation field is transformed into a strain matrix, and the strain matrix describes the deformation amounts of the hot spot area in various directions. For an actual case, the strain amount in the radial direction of the hot spot area is 0.015, and the strain amount in the tangential direction is 0.025.

[0044] Perform eigenvalue decomposition on the strain matrix to obtain the principal strain direction and the principal strain amount. The principal strain direction is the direction of the eigenvector corresponding to the largest eigenvalue, and this direction is determined as the migration direction of the hot spot area. In the actual case, the principal strain direction forms an angle of 30° with the rotation direction of the brake disc, and the principal strain amount is 0.032.

[0045] Project the displacement field onto the principal strain direction and calculate the displacement amount of the hot spot area in this direction, that is, the migration distance. For adjacent time instants t1 and t2, the migration distance is 3.5 mm. By connecting the centroid positions of each time instant and considering the constraint of the principal strain direction, the migration trajectory of the hot spot area is constructed. During a complete braking cycle, the hot spot area may form a spiral or arc-shaped migration trajectory.

[0046] Take the derivative of the migration distance with respect to time to obtain the migration rate. At the initial stage of braking, the migration rate can reach 40 mm / s. As the braking process progresses, the migration rate gradually decreases to 15 mm / s. The system matches the calculated migration trajectory with the characteristic patterns in the historical fault database, and the historical database contains more than 1000 groups of hot spot migration patterns and corresponding fault types under different working conditions.

[0047] By calculating the similarity between the current migration trajectory and the historical pattern, select the historical case with the highest similarity as a reference. If the similarity between the current trajectory and a certain historical case is 85%, then the current migration rate is corrected according to the actual measured value in this historical case, the correction coefficient is 0.92, and the corrected migration rate is 36.8 mm / s.

[0048] Use the corrected migration rate to perform spatial mapping on the temperature gradient, considering the influence of the movement of the hot spot area on the temperature distribution. In the hot spot migration direction, the temperature gradient will change due to the migration effect. Usually, the temperature gradient in front of the migration direction is larger, up to 15 °C / mm, while the temperature gradient behind is smaller, about 8 °C / mm.

[0049] Based on the temperature field distribution considering the migration effect, the system divides the stress calculation grid, and the grid size is 2mm×2mm. The migration rate is decomposed into tangential velocity and normal velocity at the grid nodes. The tangential velocity is parallel to the surface, and the normal velocity is perpendicular to the surface. For the case where the migration rate is 36.8mm / s and the angle between the migration direction and the surface normal is 60°, the tangential velocity is 31.9mm / s and the normal velocity is 18.4mm / s.

[0050] The shear stress is calculated using the tangential velocity, and the shear stress is related to the tangential velocity and the viscoelastic parameters of the material. For the brake disc material, when the tangential velocity is 31.9mm / s, the generated shear stress is 25MPa. The compressive stress is calculated using the normal velocity, and the compressive stress is related to the normal velocity and the elastic modulus of the material. When the normal velocity is 18.4mm / s, the generated compressive stress is 42MPa.

[0051] The shear stress and the compressive stress are vectorially superimposed at each grid node to obtain the stress distribution inside the brake. At the edge of the hot spot area, the stress value is usually high and can reach 85MPa. The system calculates the spatial second derivative of the stress distribution to identify the areas where the stress changes violently, and these areas are usually stress singularities. The stress value at the stress singularity is determined as the stress peak. In actual cases, the stress peak can reach 120MPa, which is much higher than the fatigue limit of the material and is a potential crack source.

[0052] Figure 2 It is a three-dimensional schematic diagram of the relationship between hot spot area migration and stress distribution. Three key feature points are marked in the chart: the blue point represents the normal operating state (15mm / s, 8°C / mm, 50MPa); the orange point represents the warning state (28mm / s, 12°C / mm, 80MPa); the red point highlights the position of the stress peak (30mm / s, 16.5°C / mm, 120MPa). The stress value at this point far exceeds the fatigue limit of the material and is a potential failure source.

[0053] The surface morphology reflects the influence of the non-linear relationship between the migration rate and the temperature gradient on the stress distribution. When the migration rate exceeds 25mm / s and the temperature gradient is greater than 15°C / mm, the stress value rises sharply, forming an obvious high-stress area.

[0054] In the prior art, the monitoring of the hot spot area of elevator brakes mostly uses static temperature threshold judgment, which cannot accurately track the dynamic changes of the hot spot area during operation, resulting in lag and local distortion in the stress evaluation results, and it is difficult to truly reflect the material stress evolution process caused by heat. Therefore, this application introduces a big data analysis platform to finely model the temperature centroid and boundary changes of the hot spot area at different time points, construct a continuous displacement field and strain matrix, and achieve dynamic capture of the migration trajectory and migration rate of the hot spot. The migration path is identified through the principal strain direction, and a historical fault database is introduced for feature matching and speed correction, significantly improving the stability and prediction accuracy of the migration model. Further, the corrected migration rate is jointly analyzed with the temperature gradient to establish a high-resolution stress distribution model, and the shear stress and compressive stress are accurately calculated through grid decomposition, effectively capturing the stress singularity points. The improvement starting point of this application is to improve the modeling accuracy of the thermo-mechanical behavior evolution of the hot spot area, and finally achieve high-sensitivity identification of the internal stress peak of the brake, providing a more forward-looking judgment basis for fault warning.

[0055] In an optional implementation manner, when the stress peak continuously increases, the fault sample library is called, and the fatigue damage degree of the brake material is calculated according to the stress peak and the area of the hot spot area. Based on the fatigue damage degree and the historical fault sample data, the fatigue damage development curve is obtained by fitting using the dynamic damage evolution state equation, including: Collect the stress peak and hot spot area data, call the historical fault data in the fault sample library, segment the historical fault data according to the growth interval of the stress peak, and calculate the change rate of the stress peak, the growth rate of the hot spot area and its duration in each segment to obtain the stress-area evolution characteristic parameters; According to the stress-area evolution characteristic parameters, calculate the fatigue damage increment in each segment, and construct a time series weight matrix with the fatigue damage increment and the duration of the corresponding segment; When the stress peak continuously increases, match the change rate of the current stress peak and the hot spot area with the stress-area evolution characteristic parameters, determine the current damage evolution stage according to the matching result, calculate the damage accumulation rate of the current stage based on the time series weight matrix, and calculate the fatigue damage degree according to the damage accumulation rate; Perform weighted combination of the fatigue damage degree and the time series weight matrix to obtain a corrected damage value considering the historical cumulative effect, and perform piecewise linear fitting on the corrected damage value, and use the time series weight matrix for weighted processing during the fitting process to obtain a fatigue damage development curve reflecting the damage development trend.

[0056] Exemplarily, during the process of collecting stress peak and hot spot area data, the system monitors the stress state of the brake material in real time through a strain sensor array arranged on the elevator overspeed protection device. The sensor array includes 16 high-precision strain gauges, which are evenly distributed at key parts of the brake, and the sampling frequency is set to 100 Hz to ensure capturing the transient stress changes during the braking process. In practical applications, the stress peak of the elevator brake usually fluctuates between 250 MPa and 450 MPa, and the area of the hot spot region is about 20 cm 2 at the initial stage of operation, and can expand to 60 cm 2 or even larger as the usage time increases. When calling the historical fault data in the fault sample library, more than 10,000 groups of elevator overspeed protection device fault cases stored in the cloud database will be read. Each group of cases includes the stress peak change data, hot spot area change data, and the final fault mode and fault time point information in the 30 days before the fault. The historical fault data is segmented according to the growth interval of the stress peak. Usually, the stress peak is divided into five intervals: 250 - 300 MPa, 300 - 350 MPa, 350 - 400 MPa, 400 - 450 MPa, and above 450 MPa. Calculate the change rate of the stress peak within each segment, that is, the increase in the stress peak between two adjacent measurements divided by the time interval, with the unit of MPa / hour. For a normally operating elevator overspeed protection device, the stress peak change rate is usually between 0.5 - 2 MPa / hour; when the change rate exceeds 3 MPa / hour and lasts for more than 12 hours, it indicates that the device may be in an abnormal state. At the same time, calculate the growth rate of the hot spot area, that is, the increase in the hot spot area divided by the time interval, with the unit of cm 2 / h. Under normal circumstances, the growth rate of the hot spot area is between 0.2 - 0.8 cm 2 / h; when the growth rate exceeds 1.5 cm 2 / h and lasts for more than 8 hours, the system will trigger an alarm. Record the duration of the stress peak change rate and the hot spot area growth rate within each segment, and summarize them to form a stress - area evolution characteristic parameter table, which serves as the basic data for subsequent analysis.

[0057] When calculating the fatigue damage increment within each segment according to the stress - area evolution characteristic parameters, the system adopts the modified linear cumulative damage theory. For the brake material of the elevator overspeed protection device, the reference damage value is set to 100, corresponding to the complete failure state. Within each stress interval, calculate the fatigue damage increment according to the current stress peak, stress change rate, hot spot area, and its growth rate. Taking the stress interval of 350 - 400 MPa as an example, if the stress change rate is 2.5 MPa / h and the growth rate of the hot spot area is 1.2 cm 2 / h, and the duration is 10 hours, then the calculated result of the fatigue damage increment within this segment is approximately 8.5. Construct a time-series weight matrix with the fatigue damage increment and the duration of the corresponding segment. The rows of the matrix represent different stress intervals, the columns represent the time series, and the matrix element value is the damage increment at this time point in this stress interval. To reflect the influence of time factors, time weights are assigned to the matrix elements, and the data closer to the current time has a higher weight. A typical time-series weight is 1.0 for the last 7 days, 0.8 for 7 - 14 days, 0.6 for 14 - 21 days, and 0.4 for 21 - 30 days. In this way, a 5×30 time-series weight matrix is constructed to comprehensively reflect the evolution process of fatigue damage over time and stress.

[0058] When the stress peak continues to increase, match the current stress peak and the change rate of the hot spot area with the stress-area evolution characteristic parameters, and determine the best matching point using the criterion of the minimum Euclidean distance. For example, the current stress peak is 385 MPa, the change rate is 2.8 MPa / h, the area of the hot spot is 48 cm 2 , and the growth rate is 1.3 cm 2 / h. The system will search for the case with the most similar characteristics in the historical data and find that the set of parameters with the highest matching degree is in the range of 350 - 400 MPa: stress peak 382 MPa, change rate 2.7 MPa / h, hot spot area 46 cm 2 , and the growth rate is 1.4 cm 2 / h. Determine the current damage evolution stage according to the matching result. The fatigue damage process of the elevator overspeed protection device is usually divided into four stages: initial damage stage (0 - 30), stable development stage (30 - 60), accelerated deterioration stage (60 - 85), and near failure stage (85 - 100). Calculate the damage accumulation rate at the current stage based on the time-series weight matrix, that is, the increase in fatigue damage per unit time. In the accelerated deterioration stage, the damage accumulation rate will increase significantly, up to 2 - 3 times that of the stable development stage. Taking an actual case as an example, the damage value of an elevator increased from 45 to 68 during the monitoring period, taking 15 days, and the average damage accumulation rate was 1.53 per day; after entering the accelerated deterioration stage, the damage value increased from 68 to 85 within 7 days, and the rate increased to 2.43 per day. When calculating the fatigue damage degree based on the damage accumulation rate, compare the current damage accumulation rate with the historical average value to generate a relative damage index. When the relative damage index exceeds 1.5 and lasts for more than 3 days, it indicates that the device is in an abnormal damage state.

[0059] The fatigue damage degree is weighted and combined with the time series weight matrix to obtain a modified damage value considering the historical cumulative effect. The modification process takes into account the non-linear cumulative effect of historical damage, especially the residual effect of stress peaks in the high stress range. In an actual case, the original calculated damage value of an elevator overspeed protection device during upward travel is 72, and the damage value after being corrected by the historical cumulative effect is 76.8, with a correction amplitude of 6.7%. The modified damage value is subjected to piecewise linear fitting, and the time series weight matrix is used for weighting during the fitting process. Key points are selected for linear fitting at each damage evolution stage. For example, at the initial damage stage, three points with damage values of 0, 15, and 30 are selected; at the stable development stage, three points with damage values of 30, 45, and 60 are selected; at the accelerated deterioration stage, three points with damage values of 60, 75, and 85 are selected; and at the stage approaching failure, three points with damage values of 85, 95, and 100 are selected. When fitting, the weights of recent data points are higher than those of far-term data points. For example, the weight of data in the most recent 3 days is 1.0, and the weight of data from 4 to 7 days is 0.8, and so on. Through this piecewise linear fitting method, a fatigue damage development curve reflecting the damage development trend is obtained. This curve can visually display the damage state and development trend of the elevator overspeed protection device, providing a reliable basis for fault prediction.

[0060] Figure 3 It is a schematic diagram for comparing the prediction accuracies of fatigue damage assessment methods at different damage stages. The time series weighting method of the present application has the highest prediction accuracy at all damage stages, especially with a significant advantage in the later stages. At the initial damage stage, the accuracies of the three methods are 82.5%, 85.7%, and 89.2% respectively, with relatively small differences; as the damage enters the accelerated deterioration stage and the stage approaching failure, the accuracies of the traditional methods drop sharply to 65.8% and 54.2%, the linear cumulative method drops to 73.6% and 67.9%, while the method of the present application still remains at a high level of 85.3% and 82.1%.

[0061] In the prior art for fatigue damage assessment of elevator brakes, it often relies on a single stress parameter or life estimation based on a fixed model, ignoring the coupling relationship between stress peaks and the evolution process of hot spot areas, and it is difficult to accurately describe the non-linear evolution characteristics during the fatigue process. The present application effectively characterizes the fatigue increment change law of the brake at different damage stages by introducing a joint analysis mechanism of stress peaks and hot spot area, extracting stress-area evolution characteristic parameters, and constructing a time series weight matrix. Compared with the traditional static assessment method, this solution can dynamically identify the damage stage according to historical fault data and calculate the current cumulative fatigue degree in real time, enhancing the adaptability and real-time performance of the prediction model. By performing weighted piecewise fitting on the modified damage value, a fatigue damage development curve closer to the actual operating state is obtained, thereby significantly improving the identification accuracy of the potential failure trend of the brake and providing a reliable basis for subsequent control strategy optimization.

[0062] In an alternative embodiment, according to the changing trend of the fatigue damage development curve, a multi-point monitoring network is arranged on the migration trajectory of the hot spot area, and temperature gradient and stress peak data are collected in real time. Spatiotemporal correlation analysis is performed with the fatigue damage development curve to determine the location of the fault source and the expansion trend, and the fault warning information output includes: The fatigue damage development curve is subjected to piecewise linearization processing, the slope mutation points of the fatigue damage development curve are extracted, the fatigue damage stages are divided based on the slope mutation points, the migration trajectory and speed change of the hot spot area in each stage are analyzed, and the boundaries of the hot spot area in each stage are connected to construct a dynamic envelope surface; On the dynamic envelope surface, the main monitoring point is determined according to the position of the maximum value of the temperature gradient change rate. Taking the main monitoring point as the center, an annular secondary monitoring point array is constructed based on the spatial distribution law of the stress peak. A multi-objective optimization function including a temperature gradient change rate term, a stress gradient change rate term, and a monitoring point spacing term is established, and the optimal layout scheme of the monitoring network is obtained by iteratively solving the multi-objective optimization function; The temperature gradient and stress peak data in the monitoring network are collected in real time using an adaptive sliding time window and projected onto the corresponding stage of the fatigue damage development curve in chronological order to construct a feature matrix including spatial position and time series. The main eigenvector and the secondary eigenvector are extracted through singular value decomposition; The time drift rate and the spatial drift direction of the main eigenvector and the secondary eigenvector are calculated. When the time drift rate exceeds the dynamic threshold corresponding to the fatigue damage stage, the fault expansion trend is fitted according to the spatial drift direction and the temperature stress distribution law, and the fault source location is determined in combination with the historical fault database, and the fault warning information is output.

[0063] When performing piecewise linearization on the fatigue damage development curve, a piecewise smoothing filtering technique is used to preprocess the original fatigue damage curve data. For a typical elevator overspeed protection device during an inspection cycle of 100 days, fatigue damage values are collected at a frequency of one data point per day, obtaining a fatigue damage development curve containing 100 points. During the processing, a 5-point moving average method is used to eliminate short-term fluctuations and retain the main change trend of the curve. When extracting the slope mutation points of the fatigue damage development curve, the difference method is used to calculate the slope between adjacent points. When the change amplitude of the slope exceeds 30% of the average value of the previous 5 days, this point is marked as a slope mutation point. In practical applications, a typical fatigue damage development curve usually contains 3 - 5 slope mutation points. Based on the slope mutation points, the fatigue damage stage is divided. Generally, the fatigue damage process is divided into four stages: the initial damage stage, the stable development stage, the accelerated deterioration stage, and the near-failure stage. Analyze the migration trajectory and speed change of the hot spot area in each stage. Using infrared thermal imaging technology, the temperature distribution map of the elevator overspeed protection device is collected every 4 hours, and the position change of the center point of the hot spot area is tracked. Record the migration distance and direction of the hot spot area in each stage, and calculate the migration speed. For example, in the stable development stage, the average migration speed of the hot spot area is 0.8 mm / day, mainly migrating along the radial direction of the brake; after entering the accelerated deterioration stage, the migration speed increases to 1.5 mm / day, and a tangential migration component begins to appear. Connect the boundaries of the hot spot area in each stage to construct a dynamic envelope surface, generate a continuous envelope curve through multi-point cubic spline interpolation, and then superimpose the envelope curves at different time points to form a dynamic envelope surface. This envelope surface intuitively shows the expansion trend of the hot spot area over time and provides spatial guidance for the layout of monitoring points.

[0064] On the dynamic envelope surface, when the main monitoring point is determined according to the maximum position of the temperature gradient change rate, the dynamic envelope surface is scanned by a high-precision infrared array sensor to calculate the temperature gradient change rate of each point. The temperature gradient change rate is defined as the amount of change in the temperature gradient per unit time, in units of ℃ / mm·h. The entire envelope surface is divided into 5mm×5mm grids, and the average temperature gradient change rate of each grid is calculated. In an actual case, it was found after calculation that the temperature gradient change rate of the point located 30mm from the outer edge of the brake friction surface was the largest, reaching 0.42℃ / mm·h, and this point was determined as the main monitoring point. With the main monitoring point as the center, when constructing an annular secondary monitoring point array based on the spatial distribution law of the stress peak, the stress distribution on the dynamic envelope surface is measured by the strain sensor array, and the spatial distribution law of the stress peak is analyzed. Under normal circumstances, the stress peak shows a radial attenuation trend centered on the main monitoring point. Based on this law, a two-layer annular monitoring point array is designed: the inner ring radius is 20mm, and 8 monitoring points are evenly arranged; the outer ring radius is 40mm, and 12 monitoring points are evenly arranged. A multi-objective optimization function including the temperature gradient change rate term, the stress gradient change rate term, and the monitoring point spacing term is established. The weight of the temperature gradient change rate term is set to 0.4, the weight of the stress gradient change rate term is set to 0.4, and the weight of the monitoring point spacing term is set to 0.2. The temperature gradient change rate term is used to ensure that the monitoring point is located at the position where the temperature changes most significantly; the stress gradient change rate term ensures that the monitoring point covers the area where the stress changes most dramatically; and the monitoring point spacing term ensures the uniformity of the spatial coverage of the monitoring network. When the optimal layout plan of the monitoring network is obtained by iteratively solving the multi-objective optimization function, the simulated annealing algorithm is used for solution, and the initial temperature is set to 100, the termination temperature is set to 0.1, and the cooling coefficient is set to 0.95. After 500 iterations, the optimized position coordinates of 21 monitoring points are obtained to form a complete monitoring network.

[0065] When adaptively sliding time windows are used to collect temperature gradient and stress peak data in the monitoring network in real time, the collection frequency is dynamically adjusted according to the fatigue damage development stage. In the initial damage stage, the collection frequency is set to once every 4 hours; in the stable development stage, it is once every 2 hours; in the accelerated deterioration stage, it is once every 1 hour; and in the near failure stage, it is once every 30 minutes. The length of the time window also changes with the stage: 48 hours in the initial stage, 24 hours in the stable stage, 12 hours in the accelerated stage, and 6 hours in the near failure stage. Within each time window, the temperature gradient and stress peak data of all monitoring points are recorded to form a two-dimensional data matrix. And it is projected onto the corresponding stage of the fatigue damage development curve in chronological order, associating the monitoring data at each time point with the fatigue damage value at the corresponding time point. A feature matrix containing spatial location and time series is constructed. The rows of the matrix represent different monitoring points, the columns represent different time points, and the matrix element values are the temperature gradient or stress peak data of the corresponding point at the corresponding time. For example, for 21 monitoring points, data is collected once an hour within a 24-hour window, resulting in a 21×24 feature matrix. When extracting the principal eigenvector and the secondary eigenvector through singular value decomposition, the feature matrix is decomposed into a spatial eigenvector and a time eigenvector. Usually, the eigenvectors corresponding to the first 3 largest singular values are taken as the principal eigenvector, and the eigenvectors corresponding to the 4th to 6th singular values are taken as the secondary eigenvector. The principal eigenvector reflects the main change pattern of the system, while the secondary eigenvector captures the secondary change trend of the system.

[0066] When calculating the time drift rate and spatial drift direction of the main eigenvector and secondary eigenvector, the eigenvectors of adjacent time windows are compared, and the change rate of their Euclidean distance is calculated as the time drift rate. The spatial drift direction is determined by calculating the gradient field of the eigenvector in the spatial dimension. In an actual case, the time drift rate of the main eigenvector of an elevator overspeed protection device in the accelerating deterioration stage is 0.068 / hour, and the spatial drift direction mainly points to the radial outside of the brake friction surface. When the time drift rate exceeds the dynamic threshold corresponding to the fatigue damage stage, the system triggers a fault trend analysis. The dynamic threshold is determined based on historical data statistics: 0.02 / hour for the initial damage stage, 0.04 / hour for the stable development stage, 0.06 / hour for the accelerating deterioration stage, and 0.1 / hour for the near failure stage. According to the spatial drift direction and the temperature stress distribution law, the fault propagation trend is fitted, and a two-dimensional Gaussian process regression model is used with the spatial drift direction as the main axis to predict the expansion range of the hot spot area within the next 48 hours. When determining the fault source location by combining with the historical fault database, the currently monitored eigenvector is matched with the samples in the historical fault database to find the most similar historical case and refer to its fault source location. In a specific case, the system analysis finds that the fault source is located at the outer edge of the contact surface between the brake friction plate and the steel backplate, and the fault propagation trend points to the center of the brake. It is expected that the expansion area will increase by 40% within 36 hours. When outputting the fault warning information, a warning report including the fault source location coordinates, fault type, severity level, expected development time, and recommended maintenance measures is generated. The severity level is divided into four levels: reminder, attention, warning, and emergency. This case is judged to be at the "warning" level, and it is recommended to carry out maintenance and repair within 24 hours.

[0067] In this embodiment, through the trend change of the fatigue damage development curve and combining with the migration trajectory of the hot spot area, the optimal layout of the multi-point monitoring network is realized, which can effectively improve the perception coverage and monitoring accuracy of the temperature gradient and stress peak change. By using piecewise linearization and slope mutation point analysis, the system can identify the key features of different damage stages and realize the dynamic tracking of the damage evolution process. By constructing a spatio-temporal feature matrix and extracting the main and secondary eigenvectors, the prediction of the fault propagation trend and the accurate positioning of the fault source can be realized. Compared with the existing scheme based on single-point monitoring and static threshold judgment, this scheme introduces a dynamic envelope surface, multi-objective optimization, and spatio-temporal fusion analysis, enhancing the ability to identify early fault signs and improving the timeliness and accuracy of early warning.

[0068] In the second aspect of the embodiment of the present invention, a fault prediction system for an elevator overspeed protection device based on big data is provided, and the system includes: The first unit is used to set a thermocouple array on the surface of the brake of the elevator overspeed protection device, collect the temperature data of the brake surface, perform feature matching on the temperature data with the historical data set, construct a real-time temperature distribution map of the brake surface, extract the hot spot areas where the temperature exceeds the preset temperature threshold in the temperature distribution map, and calculate the area and temperature gradient of the hot spot areas; The second unit is used to utilize the big data analysis platform to calculate the migration trajectory and migration rate of the hot spot areas according to the position changes of the hot spot areas at different times and in combination with the migration characteristics in the historical fault database, calculate the internal stress distribution of the brake based on the temperature gradient and migration rate, and obtain the stress peak value in the stress distribution; The third unit is used to, when the stress peak value continues to increase, call the fault sample library, calculate the fatigue damage degree of the brake material according to the stress peak value and the area of the hot spot area, and based on the fatigue damage degree and the historical fault sample data, fit the fatigue damage development curve by using the dynamic damage evolution state equation; The fourth unit is used to, according to the change trend of the fatigue damage development curve, arrange a multi-point monitoring network on the migration trajectory of the hot spot area, collect the temperature gradient and stress peak value data in real time, perform spatio-temporal correlation analysis with the fatigue damage development curve, determine the fault source location and the expansion trend, and output a fault warning message.

[0069] In a third aspect of the embodiments of the present invention, an electronic device is provided, including: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to call the instructions stored in the memory to execute the method described above.

[0070] In a fourth aspect of the embodiments of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method described above is implemented.

[0071] The present invention may be a method, device, system, and / or computer program product. The computer program product may include a computer-readable storage medium, on which computer-readable program instructions for executing various aspects of the present invention are uploaded.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting faults of an elevator overspeed protection device based on big data, characterized in that, Including: A thermocouple array is arranged on the surface of the brake of the elevator overspeed protection device to collect the temperature data of the brake surface, perform feature matching on the temperature data with the historical data set, construct a real-time temperature distribution map of the brake surface, extract the hot spot areas where the temperature exceeds the preset temperature threshold in the temperature distribution map, and calculate the area and temperature gradient of the hot spot areas; Using a big data analysis platform, according to the position changes of the hot spot areas at different times, combined with the migration characteristics in the historical fault database, calculate the migration trajectory and migration rate of the hot spot areas, calculate the internal stress distribution of the brake based on the temperature gradient and migration rate, and obtain the stress peak value in the stress distribution; When the stress peak value continues to increase, call the fault sample library, calculate the fatigue damage degree of the brake material according to the stress peak value and the area of the hot spot area, and based on the fatigue damage degree and the historical fault sample data, use the dynamic damage evolution state equation to fit and obtain the fatigue damage development curve; According to the change trend of the fatigue damage development curve, arrange a multi-point monitoring network on the migration trajectory of the hot spot area, and collect the temperature gradient and stress peak value data in real time, perform spatio-temporal correlation analysis with the fatigue damage development curve, determine the fault source location and expansion trend, and output a fault warning message.

2. The method according to claim 1, characterized in that A thermocouple array is arranged on the surface of the brake of the elevator overspeed protection device to collect the temperature data of the brake surface, perform feature matching on the temperature data with the historical data set, and constructing a real-time temperature distribution map of the brake surface includes: The brake surface is divided into a core monitoring area, a transition area and an edge monitoring area, and the temperature data of the thermocouple array is collected by a partition dynamic sampling method, the sampling frequency is adjusted according to the temperature change rate of each area, and the temperature data is weighted and averaged according to the monitoring area position to obtain noise-reduced temperature data; Calculate the temperature gradient and temperature distribution curvature of the noise-reduced temperature data, identify the temperature mutation abnormal points based on the temperature gradient, and identify the temperature fluctuation abnormal points based on the temperature distribution curvature; match the time domain characteristics of the temperature mutation abnormal points and temperature fluctuation abnormal points with the fault characteristics in the historical data set to obtain the confidence level of the abnormal points; Correct the temperature abnormal points according to the confidence level of the abnormal points, perform local weighted fusion on the corrected temperature data and the surrounding normal temperature data to obtain corrected temperature data, and perform cubic spline interpolation on the corrected temperature data according to the spatial position to generate a real-time temperature distribution map of the brake surface.

3. The method according to claim 1, characterized in that, Extracting the hot spot areas where the temperature exceeds the preset temperature threshold in the temperature distribution map and calculating the area and temperature gradient of the hot spot areas includes: Calculate the friction power density distribution based on the instantaneous acceleration and braking torque of the brake, perform partition weighting on the temperature distribution map according to the friction power density distribution, calculate the temperature change rate of each area, combine the temperature change rate with the brake friction torque to determine the temperature threshold, and mark the areas where the temperature in the temperature distribution map exceeds the temperature threshold as the hot spot areas to be processed; Identify the temperature extreme points in the hot spot area to be processed, classify the temperature extreme points according to the friction power density, and use the temperature extreme points from high to low level as seed points. Calculate the temperature similarity, gradient similarity, and spatial distance similarity between the points to be expanded and the grown area, adjust the weight coefficients of each similarity according to the friction power density distribution for weighting, determine the region growth boundary, and obtain the hot spot area; Perform curve fitting on the boundary points of the hot spot area to obtain a closed curve, calculate the area of the hot spot area according to the closed curve, determine the area threshold according to the friction power density level corresponding to the hot spot area, and identify the hot spot area exceeding the area threshold as the concerned hot spot area; Calculate the temperature gradient of the concerned hot spot area using a kernel function, where the bandwidth of the kernel function is inversely proportional to the friction power density of the corresponding area, and generate the temperature gradient distribution of the concerned hot spot area.

4. The method according to claim 1, wherein Using a big data analysis platform, according to the position changes of the hot spot area at different times, combined with the migration characteristics in the historical fault database, calculate the migration trajectory and migration rate of the hot spot area, and calculate the internal stress distribution of the brake based on the temperature gradient and migration rate. The stress peaks obtained from the stress distribution include: Use the big data analysis platform to collect the temperature distribution data of the hot spot area at different times, calculate the centroid coordinate sequence of the temperature distribution, construct a displacement field according to the centroid coordinate sequence, and establish the boundary curve of the hot spot area at the same time. Convert the deformation amount of the boundary curve at adjacent times into a strain matrix; Calculate the principal strain direction and strain amount according to the strain matrix, determine the migration direction as the principal strain direction, project the displacement field at adjacent times onto the migration direction to obtain the migration distance, construct the migration trajectory of the hot spot area according to the migration distance, and take the derivative of the migration distance with respect to time to obtain the migration rate; Match the migration trajectory with the characteristic patterns in the historical fault database, correct the migration rate according to the matching degree, and perform spatial mapping on the temperature gradient using the corrected migration rate to obtain the temperature field distribution considering the migration effect; Based on the temperature field distribution, divide the stress calculation grid, decompose the migration rate into the tangential velocity and normal velocity at the grid nodes, calculate the shear stress using the tangential velocity, calculate the compressive stress using the normal velocity, superimpose the shear stress and compressive stress to obtain the internal stress distribution of the brake, and identify the stress singular points according to the spatial second derivative of the stress distribution. Determine the stress value at the stress singular point as the stress peak.

5. The method according to claim 1, characterized in that, When the stress peak continues to increase, call the fault sample library, calculate the fatigue damage degree of the brake material according to the stress peak and the area of the hot spot area, and use the dynamic damage evolution state equation to fit the fatigue damage development curve based on the fatigue damage degree and historical fault sample data, including: Collect the stress peak and hot spot area data, call the historical fault data in the fault sample library, segment the historical fault data according to the growth interval of the stress peak, calculate the change rate of the stress peak, the growth rate of the hot spot area, and its duration in each segment to obtain the stress-area evolution characteristic parameters; Calculate the fatigue damage increment within each segment according to the stress - area evolution characteristic parameters, and construct a time - series weight matrix by combining the fatigue damage increment with the duration of the corresponding segment. When the stress peak continues to increase, match the current stress peak and the change rate of the hot - spot area with the stress - area evolution characteristic parameters, determine the current damage evolution stage according to the matching result, calculate the damage accumulation rate at the current stage based on the time - series weight matrix, and calculate the fatigue damage degree according to the damage accumulation rate. Perform weighted combination on the fatigue damage degree and the time - series weight matrix to obtain a corrected damage value considering historical cumulative effects, and perform piece - wise linear fitting on the corrected damage value. During the fitting process, use the time - series weight matrix for weighted processing to obtain a fatigue damage development curve reflecting the damage development trend.

6. The method according to claim 1, characterized in that, According to the change trend of the fatigue damage development curve, deploy a multi - point monitoring network on the migration trajectory of the hot - spot area, and collect temperature gradient and stress peak data in real - time. Conduct spatio - temporal correlation analysis with the fatigue damage development curve to determine the location and expansion trend of the fault source, and output fault warning information including: Perform piece - wise linearization on the fatigue damage development curve, extract the slope mutation points of the fatigue damage development curve, divide the fatigue damage stages based on the slope mutation points, analyze the migration trajectory and speed change of the hot - spot area in each stage, and connect the boundaries of the hot - spot area in each stage to construct a dynamic envelope surface. On the dynamic envelope surface, determine the main monitoring point according to the position of the maximum value of the temperature gradient change rate. Taking the main monitoring point as the center, construct an annular secondary monitoring point array based on the spatial distribution law of the stress peak, establish a multi - objective optimization function including temperature gradient change rate term, stress gradient change rate term, and monitoring point spacing term, and obtain the optimal deployment plan of the monitoring network by iteratively solving the multi - objective optimization function. Use an adaptive sliding time window to collect temperature gradient and stress peak data in the monitoring network in real - time, project them onto the corresponding stages of the fatigue damage development curve in chronological order, construct a feature matrix including spatial position and time series, and extract the main eigenvector and secondary eigenvector through singular value decomposition. Calculate the time drift rate and spatial drift direction of the main eigenvector and secondary eigenvector. When the time drift rate exceeds the dynamic threshold corresponding to the fatigue damage stage, fit the fault expansion trend according to the spatial drift direction and temperature - stress distribution law, determine the fault source location in combination with the historical fault database, and output fault warning information.

7. A fault prediction system for an elevator overspeed protection device based on big data, which is used to implement the method described in any one of the foregoing claims 1-6, characterized in that, Including: The first unit is used to set a thermocouple array on the surface of the brake of the elevator overspeed protection device, collect the temperature data on the brake surface, perform feature matching on the temperature data with the historical data set, construct a real - time temperature distribution map of the brake surface, extract the hot - spot area where the temperature exceeds the preset temperature threshold in the temperature distribution map, and calculate the hot - spot area and temperature gradient. A second unit, configured to utilize a big data analysis platform to calculate the migration trajectory and migration rate of a hot spot area based on the position change of the hot spot area at different times and in combination with the migration characteristics in the historical fault database, calculate the internal stress distribution of the brake based on the temperature gradient and the migration rate, and obtain the stress peak in the stress distribution; A third unit, configured to call a fault sample library when the stress peak continues to increase, calculate the fatigue damage degree of the brake material based on the stress peak and the area of the hot spot area, and use a machine learning algorithm to fit a fatigue damage development curve based on the fatigue damage degree and historical fault sample data; A fourth unit, configured to deploy a multi-point monitoring network on the migration trajectory of the hot spot area according to the change trend of the fatigue damage development curve, collect temperature gradient and stress peak data in real time, perform spatio-temporal correlation analysis with the fatigue damage development curve, determine the fault source location and expansion trend, and output a fault warning message.

8. An electronic device, characterized in that, Comprising: A processor; A memory for storing instructions executable by the processor; Wherein, the processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, the method according to any one of claims 1 to 6 is implemented.

Images