An elevator wire rope testing device and its testing method

Abstract

This invention discloses an elevator wire rope detection device and its detection method. The device comprises the following steps: first, a pre-scan of the initial magnetic field; second, magnetization of the wire rope using a multi-stage adjustable electromagnet, with the magnetization intensity controlled by current; then, real-time monitoring and feedback control of the magnetized magnetic field, and synchronous acquisition via multi-modal sensors; finally, extraction of multi-domain features; the model receiving the feature vector set and performing dynamic weighted calculations; and finally, output of the task. This method, by combining the initial magnetic field pre-scanning with multi-stage adjustable electromagnet magnetization, can effectively eliminate the problem of inconsistent initial magnetic fields of the wire rope. Furthermore, by synchronously acquiring data through multi-modal sensors, it achieves effective differentiation and quantitative evaluation of different types of defects such as broken wires, wear, corrosion, and fatigue damage, thus overcoming the shortcomings of existing detection methods in defect subdivision and quantification.

Description

Technical Field

[0001] This invention relates to the field of elevator safety inspection, and in particular to an elevator wire rope inspection device and its inspection method. Background Technology

[0002] Currently, various elevator wire rope testing devices have appeared on the market. These devices typically use weak magnetic detection technology as their core. They collect magnetic field signals of the wire rope after it has been magnetized by setting up magnetic sensors, and then analyze the signals to determine whether the wire rope has defects such as broken wires or wear.

[0003] However, in actual testing, elevator wire ropes exhibit inconsistent initial magnetic fields due to variations in material, model, and service life. This difference in initial magnetic field causes drift in the reference magnetic field during subsequent testing, making it difficult for the detection system to accurately distinguish between normal and defective magnetic field signals. This leads to misjudgments or missed detections of wire rope defects. Furthermore, in actual elevator wire ropes, defects such as broken wires, wear, corrosion, and fatigue are often not isolated but intertwined. The weak magnetic field characteristic signals generated by these different types of defects are spatially very similar, sometimes even overlapping. Existing detection methods can only make an overall judgment on the presence of defects in the wire rope, unable to quantify the severity of each defect or effectively distinguish between different types of defects, thus exhibiting significant limitations in the accuracy and detail of defect detection.

[0004] The reason for this problem is that existing detection devices do not specifically treat the initial magnetic field of the wire rope before detection. They directly magnetize the wire rope with a fixed intensity, resulting in wire ropes with large differences in initial magnetic field that cannot form a uniform and stable reference magnetic field after magnetization, which has limitations. In addition, the sensors used in existing detection devices are mostly single-type magnetic sensors, which can only collect magnetic field signals in a single dimension. It is difficult to fully capture the unique magnetic field characteristics of different defects. Furthermore, they can only perform basic filtering and amplification on the collected raw signals, lacking multi-dimensional in-depth analysis of the signals in the time domain, frequency domain, and spatial domain. They cannot extract the feature information corresponding to different defects from overlapping signals, ultimately making it impossible to effectively distinguish and quantitatively evaluate different types of defects. Summary of the Invention

[0005] The purpose of this invention is to provide an elevator wire rope detection device and its detection method to solve the problems of misjudgment and missed detection caused by the drift of the reference magnetic field in existing detection methods, as well as the inability to distinguish between different types of defects.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an elevator wire rope detection device and its detection method, comprising the following steps:

[0007] S1. Initial magnetic field pre-scan: Before the wire rope enters the detection area, the distribution of the residual magnetic field intensity inside the wire rope is quickly scanned by the initial magnetic field sensor array.

[0008] S2. Multi-stage adjustable electromagnet magnetization: The steel wire rope is magnetized using a multi-stage adjustable electromagnet, and the magnetization intensity is controlled by the current.

[0009] S3. Real-time monitoring and feedback control of the magnetic field after magnetization: The magnetic field after actual magnetization is measured by a reference magnetic field sensor array, compared with the preset target magnetic field value, and the current intensity is adjusted according to the difference.

[0010] S4. Multimodal sensor synchronous acquisition: During the weak magnetic field detection stage, multiple multimodal sensors are deployed to detect the contour, surface and near-surface defects and overall damage of the wire rope defects, and transmit the data to the processor.

[0011] S5. Multi-domain feature extraction: The processor performs time-domain, frequency-domain, and spatial-domain analysis on the raw signals acquired by the multimodal sensor, extracts physically meaningful feature values, and constructs a set of feature vectors.

[0012] S6, Feature Weighted Focus: The multi-task learning model receives the set of feature vectors generated by S5 and performs dynamic weighted operations on the input features through the attention weight matrix embedded within it.

[0013] S7. Multi-task output: The weighted feature vector is fed into the output layer of the multi-task learning model.

[0014] S8. Final Result Output: The output layer of the multi-task learning model compares the feature vectors with the database to determine the specific defects of the wire rope and outputs the results.

[0015] Preferably, the calculation formula in S1 is as follows:

[0016]

[0017] This represents the initial residual magnetic field; These represent the axial, circumferential, and normal magnetic field components, respectively.

[0018] Preferably, the calculation formula in S2 is as follows:

[0019]

[0020] The magnetization intensity; Current intensity; denoted as the magnetization coefficient of the electromagnet.

[0021] Preferably, the magnetization control loop formula in S3 is as follows:

[0022]

[0023] for The excitation current value that needs to be supplied to the multi-stage adjustable electromagnet is calculated at all times;

[0024] In order to be in The output excitation current value;

[0025] exist The control deviation value measured at each moment is calculated using the following formula;

[0026]

[0027] The target magnetic field value preset for the processor; for At any given moment, the actual magnetic field value on the surface of the wire rope is measured in real time by a reference magnetic field sensor array;

[0028] This is the proportional gain coefficient; This is the integral gain coefficient; The differential gain coefficient;

[0029] For deviation From time 0 to the present The integral at any given moment represents the cumulative sum of all historical deviations; For deviation At the present moment The derivative represents the instantaneous rate of change of the deviation.

[0030] Preferably, the multimodal acquisition sensor used in S4 and the formula are as follows:

[0031] Multimodal acquisition sensors include eddy current sensors, fluxgate sensors, and three-dimensional leakage field sensors;

[0032] Eddy current sensors are used to detect surface and near-surface defects, and their formula is as follows;

[0033]

[0034] The skin depth is used to determine the effective depth of eddy current detection; The frequency of the excitation current;

[0035] The magnetic permeability of the material; The electrical conductivity of the material;

[0036] The output of an eddy current sensor is typically reflected as a change in complex impedance;

[0037]

[0038] It is a complex impedance; For resistance components; For the inductive component;

[0039] Fluxgate sensors are used to measure slow, low-frequency changes in the background magnetic field, aiding in the assessment of overall damage such as uniform wear. The object being measured is... , This represents the low-frequency, slowly varying component of the background magnetic field.

[0040] A three-dimensional leakage magnetic field sensor measures the magnetic field strength and gradient changes in the normal Z, axial X, and circumferential Y directions on the surface of a steel wire rope. The measurement object is as follows;

[0041]

[0042] The magnetic field strength vector at a point in space; The axial component represents the magnetic field strength along the length of the wire rope. The magnetic field strength along the circumferential direction of the wire rope is represented by the circumferential component. The normal component is the magnetic field strength perpendicular to the surface of the wire rope.

[0043] Preferably, the formula in S5 is as follows:

[0044] Time-domain peak formula

[0045] The representative signal during the observation period The maximum absolute value range that can be achieved within; Maximum value function; It is an absolute value function; It is a time-domain signal; For time;

[0046] Root mean square formula

[0047] It is the root mean square value; To find the coefficient of the average, where It is the total time taken for the calculation; The symbol for definite integral; For the time domain signal at time... The square of the value; For integration variables;

[0048] Waveform factor formula

[0049] Waveform factor; Peak value; It is the root mean square value;

[0050] Frequency domain analysis - power spectral density;

[0051]

[0052] For the signal at frequency Power spectral density at; For time The changing original time-domain signal; For frequency; The imaginary unit; It is a complex exponential function used to decompose a time-domain signal into a sum of complex exponential components of different frequencies; For the integral of time from negative infinity to positive infinity;

[0053] Spatial Domain Analysis - Magnetic Field Gradient;

[0054]

[0055] The magnetic field strength vector The gradient; is the rate of change of the magnetic field strength along the axial direction;

[0056] This represents the rate of change of the magnetic field strength in the circumferential direction; This represents the rate of change of the magnetic field strength along the normal direction;

[0057] Spatial Domain Analysis - Laplace Operator;

[0058]

[0059] For the Laplace operator; The magnetic field strength; magnetic field strength exist Second-order partial derivatives along the axial direction; magnetic field strength exist Second-order partial derivatives along the axial direction; magnetic field strength exist Second-order partial derivatives in the axial direction.

[0060] Preferably, the formula in S6 is as follows:

[0061] ;

[0062] This is the original feature vector; This is the weight matrix; This is the weighted eigenvector.

[0063] Preferably, the formula for the metal cross-sectional area loss rate in S7 is as follows:

[0064]

[0065] This refers to the metal cross-sectional area loss rate. This represents the measured magnetic flux value of the current detection section; The reference magnetic flux value is the one without defects; This is the calibration coefficient.

[0066] Preferably, the probability distribution formula for the defect type in S7 is as follows:

[0067]

[0068] The probability that the current defect belongs to the i-th class is determined by the model; This represents the raw score corresponding to the i-th category in the model's output layer. This indicates that the original scores for all K categories are iterated over. It is a natural exponential function; This represents the total number of defect categories.

[0069] Preferably, the device includes a frame, with a magnetic memory planning component on one side of the frame and a weak magnetic detection component on the other side of the frame. A connecting block is provided between the magnetic memory planning component, the weak magnetic detection component and the frame.

[0070] The technical effects and advantages of this invention are as follows:

[0071] This method combines initial magnetic field pre-scanning with multi-level adjustable electromagnet magnetization, along with real-time monitoring and feedback control of the magnetic field after magnetization. This effectively eliminates the problem of inconsistent initial magnetic fields in wire ropes, avoids drift of the reference magnetic field, reduces the probability of defect misjudgment and missed detection, and improves the reliability of detection results.

[0072] This method uses multi-modal sensors to collect data synchronously, which can comprehensively capture the feature information of different types of defects in steel wire ropes. Then, through multi-domain feature extraction and feature weighted focusing, it can effectively distinguish and quantify different types of defects such as broken wires, wear, corrosion, and fatigue damage. This makes up for the shortcomings of existing detection methods in defect subdivision and quantification, and provides accurate data support for the maintenance and replacement of elevator steel wire ropes.

[0073] The magnetic memory planning component and the weak magnetic detection component in the testing device are stably connected to the frame through connecting blocks. The frame is equipped with adjustable feet to adapt to different installation environments, ensuring the stability of the testing process. The modular design facilitates manufacturing and subsequent maintenance, making it practical. Attached Figure Description

[0074] Figure 1 This is a schematic diagram of the overall structure of the present invention;

[0075] Figure 2 This is a schematic diagram of the connecting block of the present invention;

[0076] Figure 3 This is a schematic diagram of the process of the present invention;

[0077] Figure 4 This is a schematic diagram of the magnetization control closed loop of the present invention;

[0078] Figure 5 This is a schematic diagram of synchronous data acquisition using the multimodal sensor of the present invention;

[0079] Figure 6 This is a schematic diagram of multi-domain feature extraction according to the present invention;

[0080] Figure 7 This is a schematic diagram of the intelligent diagnostic decision-making process of the present invention.

[0081] In the diagram: 1. Frame; 2. Magnetic memory planning component; 3. Weak magnetic field detection component; 4. Connecting block; Detailed Implementation

[0082] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0083] Example 1: The present invention provides as follows Figures 1 to 3 The elevator wire rope testing device and its testing method shown herein include the following steps:

[0084] S1. Initial magnetic field pre-scan: Before the wire rope enters the detection area, the distribution of the residual magnetic field intensity inside the wire rope is quickly scanned by the initial magnetic field sensor array.

[0085] S2. Multi-stage adjustable electromagnet magnetization: The steel wire rope is magnetized using a multi-stage adjustable electromagnet, and the magnetization intensity is controlled by the current.

[0086] S3. Real-time monitoring and feedback control of the magnetic field after magnetization: The magnetic field after actual magnetization is measured by a reference magnetic field sensor array, compared with the preset target magnetic field value, and the current intensity is adjusted according to the difference.

[0087] S4. Multimodal sensor synchronous acquisition: During the weak magnetic field detection stage, multiple multimodal sensors are deployed to detect the contour, surface and near-surface defects and overall damage of the wire rope defects, and transmit the data to the processor.

[0088] S5. Multi-domain feature extraction: The processor performs time-domain, frequency-domain, and spatial-domain analysis on the raw signals acquired by the multimodal sensor, extracts physically meaningful feature values, and constructs a set of feature vectors.

[0089] S6, Feature Weighted Focus: The multi-task learning model receives the set of feature vectors generated by S5 and performs dynamic weighted operations on the input features through the attention weight matrix embedded within it.

[0090] S7. Multi-task output: The weighted feature vector is fed into the output layer of the multi-task learning model.

[0091] S8. Final Result Output: The output layer of the multi-task learning model compares the feature vectors with the database to determine the specific defects of the wire rope and outputs the results.

[0092] The device includes a frame 1, a magnetic memory planning component 2 on one side of the frame 1, a weak magnetic detection component 3 on the other side of the frame 1, and a connecting block 4 between the magnetic memory planning component 2, the weak magnetic detection component 3 and the frame 1.

[0093] In this embodiment, the initial magnetic field distribution data is first obtained by pre-scanning the steel wire rope using an initial magnetic field sensor array. Then, based on this data, the steel wire rope is differentially magnetized using multi-stage adjustable electromagnets to eliminate the initial magnetic field differences. Next, the magnetic field after magnetization is monitored by a reference magnetic field sensor array, compared with the target value, and the current is adjusted to ensure the stability of the reference magnetic field. Afterward, defect signals are simultaneously acquired using eddy current, fluxgate, and three-dimensional leakage magnetic field sensors. The processor extracts features from the signals in the time domain, frequency domain, and spatial domain to form a set of feature vectors. The multi-task learning model weights the features through an attention weight matrix to highlight key features. Finally, through the various branches of the output layer, defects such as wire breakage, corrosion, and fatigue damage are quantitatively evaluated to achieve accurate detection.

[0094] Example 2, as Figures 3 to 4 The elevator wire rope testing device and its testing method shown herein include the following steps:

[0095] S1. Initial magnetic field pre-scan: Before the wire rope enters the detection area, the distribution of the residual magnetic field intensity inside the wire rope is quickly scanned by the initial magnetic field sensor array.

[0096] S2. Multi-stage adjustable electromagnet magnetization: The steel wire rope is magnetized using a multi-stage adjustable electromagnet, and the magnetization intensity is controlled by the current.

[0097] S3. Real-time monitoring and feedback control of the magnetic field after magnetization: The magnetic field after actual magnetization is measured by a reference magnetic field sensor array, compared with the preset target magnetic field value, and the current intensity is adjusted according to the difference.

[0098] S4. Multimodal sensor synchronous acquisition: During the weak magnetic field detection stage, multiple multimodal sensors are deployed to detect the contour, surface and near-surface defects and overall damage of the wire rope defects, and transmit the data to the processor.

[0099] S5. Multi-domain feature extraction: The processor performs time-domain, frequency-domain, and spatial-domain analysis on the raw signals acquired by the multimodal sensor, extracts physically meaningful feature values, and constructs a set of feature vectors.

[0100] S6, Feature Weighted Focus: The multi-task learning model receives the set of feature vectors generated by S5 and performs dynamic weighted operations on the input features through the attention weight matrix embedded within it.

[0101] S7. Multi-task output: The weighted feature vector is fed into the output layer of the multi-task learning model.

[0102] S8. Final Result Output: The output layer of the multi-task learning model compares the feature vectors with the database to determine the specific defects of the wire rope and outputs the results.

[0103] The calculation formula in S1 is as follows:

[0104]

[0105] This represents the initial residual magnetic field; These represent the axial, circumferential, and normal magnetic field components, respectively.

[0106] The calculation formula for S2 is as follows:

[0107]

[0108] The magnetization intensity; Current intensity; denoted as the magnetization coefficient of the electromagnet.

[0109] The formula for the magnetization control loop in S3 is as follows:

[0110]

[0111] for The excitation current value that needs to be supplied to the multi-stage adjustable electromagnet is calculated at all times;

[0112] In order to be in The output excitation current value;

[0113] exist The control deviation value measured at each moment is calculated using the following formula;

[0114]

[0115] The target magnetic field value preset for the processor; for At any given moment, the actual magnetic field value on the surface of the wire rope is measured in real time by a reference magnetic field sensor array;

[0116] This is the proportional gain coefficient; This is the integral gain coefficient; The differential gain coefficient;

[0117] For deviation From time 0 to the present The integral at any given moment represents the cumulative sum of all historical deviations; For deviation At the present moment The derivative represents the instantaneous rate of change of the deviation.

[0118] In this embodiment, the magnetic memory planning component 2 contains, from left to right, an initial magnetic field sensor, a multi-level adjustable electromagnet, and a reference magnetic field sensor array, while the weak magnetic field detection component 3 contains a processor, an eddy current sensor, a fluxgate sensor, and a three-dimensional leakage magnetic field sensor.

[0119] When the wire rope enters from the left entrance of the magnetic memory planning component 2, it first passes through the initial magnetic field sensor. This sensor uses high-sensitivity Hall elements, with six sensors evenly arranged along the circumference of the wire rope at 60° intervals. It collects the residual magnetic field strength data on the surface of the wire rope in real time and transmits the data to the processor of the weak magnetic field detection component 3. The processor generates an initial magnetic field distribution map based on the data and analyzes the magnetic field differences at different locations on the wire rope. Subsequently, the wire rope enters the multi-stage adjustable electromagnet region. This component contains four independent electromagnets, each with a core made of high-permeability silicon steel sheets stacked together, and the coil has 800 turns. Based on the initial magnetic field difference, the processor sends instructions to the current controller of each electromagnet to adjust the input current, gradually offsetting the influence of the initial magnetic field unevenness through differentiated magnetization. Finally, the magnetized wire rope passes through a reference magnetic field sensor array, which has the same structure as the initial magnetic field sensor. This array collects the magnetic field data after magnetization in real time and feeds it back to the processor for comparison with the target magnetic field value. If the difference exceeds the allowable error range, the processor immediately adjusts the current of the multi-stage adjustable electromagnets to form a closed-loop control. Ultimately, when the wire rope leaves the magnetic memory planning component 2, the surface magnetic field uniformity reaches more than 98%, providing a stable benchmark for subsequent weak magnetic field detection.

[0120] When the wire rope, after being regulated by the magnetic memory planning component 2, enters the weak magnetic field detection component 3, the eddy current sensor is first triggered. This sensor emits a 1MHz high-frequency sinusoidal excitation signal, inducing eddy currents on the surface of the wire rope. If there are defects such as rust or microcracks, the eddy current path will be distorted, causing the phase shift of the eddy current reflection signal received by the sensor. The processor calculates the rust depth or crack length by analyzing the phase shift. Next, the fluxgate sensor detects the overall magnetic field change of the wire rope. When the wire rope has broken wires or fatigue damage, the internal magnetic domains are disordered, causing abnormal fluctuations in the magnetic field strength. The processor captures the fluctuation signal and outputs the data. Finally, the three-dimensional leakage magnetic field sensor collects the leakage magnetic field signals in the X, Y, and Z directions at the defect location at a sampling rate of 2kHz. The three-dimensional coordinates and morphological characteristics of the defect are determined through spatial domain analysis. The processor performs multi-domain feature extraction and weighted focusing calculation on the signals collected by the three sensors and transmits them to the external terminal device through the data module.

[0121] The processor is an industrial-grade microcontroller chip. Combined with the data transmission module, the processor forms a complete data processing and transmission link: First, the processor receives raw data collected by various sensors and performs multi-domain feature extraction, feature weighting and focusing, and other algorithmic processing. Then, the processed feature data is uploaded to the cloud server through the data transmission module. The cloud uses more powerful computing resources to run a deep-trained multi-task model to complete defect type judgment and quantitative evaluation. Finally, the cloud sends the judgment results back to the processor through the data transmission module. The processor integrates the results and generates a specific defect detection report, which is then transmitted to the user terminal.

[0122] Example 3, as Figures 5 to 7 The elevator wire rope detection device and its detection method are shown, including S1, initial magnetic field pre-scanning, which involves rapidly scanning the distribution of residual magnetic field intensity inside the wire rope through an initial magnetic field sensor array before the wire rope enters the detection area.

[0123] S2. Multi-stage adjustable electromagnet magnetization: The steel wire rope is magnetized using a multi-stage adjustable electromagnet, and the magnetization intensity is controlled by the current.

[0124] S3. Real-time monitoring and feedback control of the magnetic field after magnetization: The magnetic field after actual magnetization is measured by a reference magnetic field sensor array, compared with the preset target magnetic field value, and the current intensity is adjusted according to the difference.

[0125] S4. Multimodal sensor synchronous acquisition: During the weak magnetic field detection stage, multiple multimodal sensors are deployed to detect the contour, surface and near-surface defects and overall damage of the wire rope defects, and transmit the data to the processor.

[0126] S5. Multi-domain feature extraction: The processor performs time-domain, frequency-domain, and spatial-domain analysis on the raw signals acquired by the multimodal sensor, extracts physically meaningful feature values, and constructs a set of feature vectors.

[0127] S6, Feature Weighted Focus: The multi-task learning model receives the set of feature vectors generated by S5 and performs dynamic weighted operations on the input features through the attention weight matrix embedded within it.

[0128] S7. Multi-task output: The weighted feature vector is fed into the output layer of the multi-task learning model.

[0129] S8. Final Result Output: The output layer of the multi-task learning model compares the feature vectors with the database to determine the specific defects of the wire rope and outputs the results.

[0130] The multimodal acquisition sensor and formula used in S4 are as follows:

[0131] Multimodal acquisition sensors include eddy current sensors, fluxgate sensors, and three-dimensional leakage field sensors;

[0132] Eddy current sensors are used to detect surface and near-surface defects, and their formula is as follows;

[0133]

[0134] The skin depth is used to determine the effective depth of eddy current detection; The frequency of the excitation current;

[0135] The magnetic permeability of the material; The electrical conductivity of the material;

[0136] The output of an eddy current sensor is typically reflected as a change in complex impedance;

[0137]

[0138] It is a complex impedance; For resistance components; For the inductive component;

[0139] Fluxgate sensors are used to measure slow, low-frequency changes in the background magnetic field, aiding in the assessment of overall damage such as uniform wear. The object being measured is... , This represents the low-frequency, slowly varying component of the background magnetic field.

[0140] A three-dimensional leakage magnetic field sensor measures the magnetic field strength and gradient changes in the normal Z, axial X, and circumferential Y directions on the surface of a steel wire rope. The measurement object is as follows;

[0141]

[0142] The magnetic field strength vector at a point in space; The axial component represents the magnetic field strength along the length of the wire rope. The magnetic field strength along the circumferential direction of the wire rope is represented by the circumferential component. The normal component is the magnetic field strength perpendicular to the surface of the wire rope.

[0143] The formula in S5 is as follows:

[0144] Time-domain peak formula

[0145] The representative signal during the observation period The maximum absolute value range that can be achieved within; Maximum value function; It is an absolute value function; It is a time-domain signal; For time;

[0146] Root mean square formula

[0147] It is the root mean square value; To find the coefficient of the average, where It is the total time taken for the calculation; The symbol for definite integral; For the time domain signal at time... The square of the value; For integration variables;

[0148] Waveform factor formula

[0149] Waveform factor; Peak value; It is the root mean square value;

[0150] Frequency domain analysis - power spectral density;

[0151]

[0152] For the signal at frequency Power spectral density at; For time The changing original time-domain signal; For frequency; The imaginary unit; It is a complex exponential function used to decompose a time-domain signal into a sum of complex exponential components of different frequencies; For the integral of time from negative infinity to positive infinity;

[0153] Spatial Domain Analysis - Magnetic Field Gradient;

[0154]

[0155] The magnetic field strength vector The gradient; is the rate of change of the magnetic field strength along the axial direction;

[0156] This represents the rate of change of the magnetic field strength in the circumferential direction; This represents the rate of change of the magnetic field strength along the normal direction;

[0157] Spatial Domain Analysis - Laplace Operator;

[0158]

[0159] For the Laplace operator; The magnetic field strength; magnetic field strength exist Second-order partial derivatives along the axial direction; magnetic field strength exist Second-order partial derivatives along the axial direction; magnetic field strength exist Second-order partial derivatives in the axial direction.

[0160] The formula in S6 is as follows:

[0161] ;

[0162] This is the original feature vector; This is the weight matrix; This is the weighted eigenvector.

[0163] The formula for the metal cross-sectional area loss rate in S7 is as follows:

[0164]

[0165] This refers to the metal cross-sectional area loss rate. This represents the measured magnetic flux value of the current detection section; The reference magnetic flux value is the one without defects; This is the calibration coefficient.

[0166] The formula for the probability distribution of defect types in S7 is as follows:

[0167]

[0168] The probability that the current defect belongs to the i-th class is determined by the model; This represents the raw score corresponding to the i-th category in the model's output layer. This indicates that the original scores for all K categories are iterated over. It is a natural exponential function; This represents the total number of defect categories.

[0169] Finally, it should be noted that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An elevator steel wire rope inspection method characterized by, The method comprises the following steps: S1, initial magnetic field pre-scanning, before the steel wire rope enters the detection area, the distribution of the residual magnetic field strength inside the steel wire rope is quickly scanned by the initial magnetic field sensor array; S2, multi-stage adjustable electromagnet magnetization, the steel wire rope is magnetized by using a multi-stage adjustable electromagnet, the magnetization strength is controlled by the current, the processor sends instructions to the current controller of each stage of electromagnet according to the initial magnetic field difference, adjusts the input current, and gradually offsets the initial magnetic field difference by differential magnetization; S3, real-time monitoring and feedback control of the magnetized magnetic field, the actual magnetized magnetic field is measured by the reference magnetic field sensor array, compared with the pre-set target magnetic field value, and the current strength is adjusted according to the difference; S4, multi-modal sensor synchronous acquisition, a plurality of multi-modal sensors are arranged in the weak magnetic detection stage, which respectively detect the profile, surface and near-surface defects and overall damage of the steel wire rope, and transmit the data to the processor; S5, multi-domain feature extraction, the processor analyzes the original signal collected by the multi-modal sensor in time domain, frequency domain and space domain, extracts the characteristic value with physical meaning, and forms a feature vector set; The original signal collected by the multi-modal sensor includes three-dimensional magnetic leakage field signal, eddy current signal and background magnetic field signal; S6, feature weighting focusing, the multi-task learning model receives the feature vector set generated by S5, and performs dynamic weighting operation on the input features through the attention weight matrix embedded in it; S7, multi-task output, the feature vector after weighting is transmitted to the output layer of the multi-task learning model; S8, final result output, the output layer of the multi-task learning model compares the feature vector with the database, and then judges the specific defects of the steel wire rope and outputs.

2. The elevator steel wire rope inspection method according to claim 1, characterized by In step S3, a high magnetic energy product permanent magnet is used as the main magnetization source, which is used to form a high strength steady magnetic field in the detection area, so that the measured steel wire rope reaches a near-saturation magnetization state.

3. The method of claim 1, wherein the method further comprises: The multi-modal acquisition sensor used in S4 is as follows: The multi-modal acquisition sensor includes an eddy current sensor, a flux gate sensor and a three-dimensional magnetic leakage field sensor; The eddy current sensor is used to detect surface and near-surface defects; The flux gate sensor is used to measure the low-frequency slow change of the background magnetic field, and assists in judging the overall damage; The three-dimensional magnetic leakage field sensor measures the magnetic field strength and gradient change of the steel wire rope in the normal direction, axial direction and circumferential direction.

4. The method of claim 1, wherein, In step S5, the original signal data is analyzed to extract characteristic values with physical meaning; the characteristic values are used to construct a heterogeneous feature vector set, which is used to represent the type, geometric shape and severity of the defects.

5. An elevator rope detection apparatus for performing the elevator rope detection method according to any one of claims 1 to 4, characterized by The frame body (1) is provided with a magnetic memory planning part (2) on one side, and a weak magnetic detection part (3) on the other side, and a connecting block (4) is arranged between the magnetic memory planning part (2), the weak magnetic detection part (3) and the frame body (1).

Images