Steel wire rope broken wire identification method and device based on magnetic memory effect and storage medium

Abstract

The application discloses a steel wire rope broken wire identification method and device based on a magnetic memory effect and a storage medium. A magnetic memory detection signal of a to-be-tested steel wire rope is acquired. Noise reduction processing is performed on the magnetic memory detection signal to obtain a noise-reduced signal. A tangential component zero-crossing point position and a normal component peak value position are determined according to the noise-reduced signal. A spatial phase delay amount is determined according to the tangential component zero-crossing point position and the normal component peak value position. In the case that the spatial phase delay amount is greater than or equal to a preset threshold, the noise-reduced signal is subjected to deconvolution recovery processing to obtain a recovered magnetic memory detection signal, so as to determine a multi-dimensional feature vector based on the recovered magnetic memory detection signal. Otherwise, the multi-dimensional feature vector is determined based on the noise-reduced signal. The number of broken wires of the to-be-tested steel wire rope is determined through a trained neural network model and the multi-dimensional feature vector.

Description

Technical Field

[0001] This application relates to the field of wire rope damage detection technology, specifically to a method, device, and storage medium for identifying broken wires in wire ropes based on magnetic memory effect. Background Technology

[0002] As a crucial load-bearing component, wire rope is widely used in critical applications such as mine hoisting, elevator traction, and cable-stayed bridges. During long-term service, wire rope inevitably suffers from broken wires due to wear, fatigue, and corrosion. The number of broken wires directly affects the remaining strength and service life of the wire rope; failure to detect these breaks in time can lead to serious safety accidents.

[0003] Currently, non-destructive testing methods for steel wire ropes mainly include traditional magnetic flux leakage detection and ultrasonic guided wave detection. However, existing methods suffer from low signal-to-noise ratios and difficulties in feature extraction when dealing with weak signals in noisy environments. Particularly for internal wire breaks, the resulting magnetic flux leakage needs to penetrate the undamaged wire layer, leading to signal amplitude attenuation, waveform broadening, and phase delay. Conventional feature extraction methods struggle to accurately quantify the number of broken wires. Furthermore, traditional back propagation (BP) neural network models are prone to getting trapped in local minima when quantitatively identifying the number of broken wires, exhibiting slow convergence and poor generalization ability, resulting in low overall quantitative accuracy for identifying the number of broken wires. Simultaneously, existing methods typically treat distinguishing between internal and external broken wires and quantifying the number of broken wires as two independent tasks, failing to utilize the inherent correlation between phase delay and wire depth to guide preprocessing strategy selection. Additionally, the traditional genetic algorithm (GA)-BP model treats all training samples equally, causing the model to tend to learn easily identifiable external broken wire features while ignoring the weak signals of internal broken wires, further reducing the quantification accuracy of internal broken wires. Summary of the Invention

[0004] This application provides a method, device, and storage medium for identifying broken wires in steel wire ropes based on the magnetic memory effect. It aims to solve the problems of existing steel wire rope broken wire detection technologies, such as severe distortion of internal broken wire signals due to magnetic shielding, making it difficult to accurately quantify the number of broken wires, failure to utilize the physical prior of phase delay to guide the selection of preprocessing strategies, and overfitting of related BP neural networks and general optimization algorithms to strong signal samples and underfitting to weak signal samples during training, resulting in the overall quantification accuracy of the number of broken wires failing to meet the requirements of industrial sites.

[0005] To achieve the above objectives, this application provides the following technical solution: In a first aspect, embodiments of this application provide a method for identifying broken wires in a steel wire rope based on the magnetic memory effect, including: Acquire the magnetic memory detection signal of the steel wire rope under test; The magnetic memory detection signal is denoised to obtain the denoised signal. The zero-crossing position of the tangential component and the peak position of the normal component are determined based on the denoised signal. The zero-crossing position of the tangential component represents the spatial coordinates of the tangential component signal crossing the zero point in the magnetic memory detection signal, and the peak position of the normal component represents the spatial coordinates of the normal component signal reaching its maximum value in the magnetic memory detection signal. The spatial phase delay is determined based on the zero-crossing position of the tangential component and the peak position of the normal component. If the spatial phase delay is greater than or equal to a preset threshold, the denoised signal is deconvolutionally recovered to obtain the recovered magnetic memory detection signal, and the multidimensional feature vector is determined based on the recovered magnetic memory detection signal; otherwise, the multidimensional feature vector is determined based on the denoised signal. The number of broken wires in the steel wire rope under test is determined by using a trained neural network model and multidimensional feature vectors.

[0006] In some embodiments of this application, determining the spatial phase delay based on the zero-crossing position of the tangential component and the peak position of the normal component includes: The absolute value of the spatial distance between the zero-crossing position of the tangential component and the peak position of the normal component is determined as the spatial phase delay.

[0007] In some embodiments of this application, the denoised signal is deconvolutionally recovered to obtain the recovered magnetic memory detection signal, including: The first pulse response signal of the steel wire rope under test is obtained and used as a kernel function; wherein, the first pulse response signal characterizes the magnetic pulse response signal of the steel wire rope under test in the non-destructive region; The denoised signal and kernel function are deconvolved to obtain the deconvolution result; The deconvolution result is subjected to frequency domain Wiener filtering to obtain the recovered magnetic memory detection signal.

[0008] In some embodiments of this application, determining a multidimensional feature vector based on the recovered magnetic memory detection signal includes: Extract the first peak value, first peak-to-peak value, first waveform energy, first zero-crossing rate, and first waveform broadening coefficient of the recovered magnetic memory detection signal; The multidimensional feature vector is determined based on the first peak value, the first peak-to-peak value, the first waveform energy, the first zero-crossing rate, the first waveform broadening coefficient, and the spatial phase delay. Accordingly, a multidimensional feature vector is determined based on the denoised signal, including: Extract the second peak value, second peak-to-peak value, second waveform energy, second zero-crossing rate, and second waveform broadening factor of the denoised signal; The multidimensional feature vector is determined based on the second peak value, the second peak-to-peak value, the second waveform energy, the second zero-crossing rate, the second waveform broadening coefficient, and the spatial phase delay.

[0009] In some embodiments of this application, the method further includes: Construct an initial neural network model; the initial neural network model includes an input layer, a hidden layer, and an output layer. The number of nodes in the input layer is equal to the dimension of the multidimensional feature vector, and the number of nodes in the output layer is 1. The initial weights and thresholds of the initial neural network model are optimized using a genetic algorithm to obtain the optimized weights and thresholds. The first neural network model is determined based on the optimized weights, optimized thresholds, and the initial neural network model. The first neural network model is trained using the training sample set to obtain the trained neural network model; the training sample set includes multiple samples, each containing a feature vector sample and a corresponding label for the number of broken wires.

[0010] In some embodiments of this application, the initial weights and initial thresholds of the initial neural network model are optimized using a genetic algorithm to obtain optimized weights and optimized thresholds, including: The initial weights and initial thresholds are encoded to generate an initial population; each individual in the initial population contains a set of weight vectors and threshold vectors. The weight vector and threshold vector of each individual in the initial population are assigned to the initial neural network model, and the predicted value of each sample in the training sample set is determined by the assigned initial neural network model. Determine the weight of each sample in the initial population; Based on the predicted value of each sample, the number of broken wires corresponding to each sample, and the weight of each sample, the weighted mean square error is determined. The reciprocal of the weighted mean square error is used as the fitness function to calculate the fitness value of each individual in the initial population. Based on the fitness values, selection, crossover, and mutation operations are performed on the initial population to generate a new generation of population. Repeat the operation of generating a new generation of population based on fitness values ​​until a target population that meets the convergence condition is obtained. The weight vector and threshold vector corresponding to the best individual in the target population are determined as the optimized weight and threshold.

[0011] In some embodiments of this application, determining the weight of each sample in the initial population includes: In the initial population, the weight of samples with the internal broken wire type is set as the first weight value, and the weight of samples with the external broken wire type is set as the second weight value. Among them, the first weight value is greater than the second weight value.

[0012] In some embodiments of this application, the magnetic memory detection signal is denoised to obtain a denoised signal, including: Multi-level wavelet decomposition of the tangential and normal components of the magnetic memory detection signal is performed using wavelet basis functions to obtain the high-frequency coefficients of each level. The high-frequency coefficients of each layer are quantized by a soft thresholding function to obtain the quantized high-frequency coefficients. The soft thresholding function is a function that sets coefficients with absolute values ​​less than the threshold to zero and shrinks coefficients with absolute values ​​greater than or equal to the threshold to zero. The quantized high-frequency coefficients are reconstructed to obtain the denoised signal.

[0013] Secondly, embodiments of this application provide a wire rope broken wire identification device based on magnetic memory effect, including a processor and a memory storing processor-executable instructions; when the instructions are executed by the processor, the above-mentioned wire rope broken wire identification method based on magnetic memory effect is implemented.

[0014] Thirdly, embodiments of this application provide a computer-readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements the above-mentioned method for identifying broken wires in steel wire ropes based on magnetic memory effect. Attached Figure Description

[0015] To more intuitively illustrate the prior art and this application, several exemplary figures are provided below. It should be understood that the specific shapes and structures shown in the figures should not generally be regarded as limiting conditions for implementing this application; for example, based on the technical concept disclosed in this application and the exemplary figures, those skilled in the art are able to easily make conventional adjustments or further optimizations to the addition / reduction / classification, specific shapes, positional relationships, connection methods, size ratios, etc. of certain units (components).

[0016] Figure 1 A schematic diagram illustrating the implementation process of the wire rope broken wire identification method based on magnetic memory effect provided in this application embodiment; Figure 2 A waveform diagram of a noisy magnetic memory signal provided for an embodiment of this application; Figure 3 A waveform diagram of the noise-reduced signal provided in an embodiment of this application; Figure 4 The image shows the result of identifying the number of broken wires in the GA-BP model provided in the embodiments of this application; Figure 5 A schematic diagram of the structural composition of the wire rope broken wire identification device based on magnetic memory effect provided in the embodiments of this application.

[0017] Figure label: A wire rope broken wire identification device based on magnetic memory effect includes: 1. Processor 11. Memory 12. Communication interface 13. Bus 14. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. Any combination of different embodiments is possible.

[0019] In the description of this application: unless otherwise stated, "a plurality of" means two or more. The terms "first," "second," "third," etc., in this application are intended to distinguish the objects referred to and do not have any special meaning in terms of technical connotation (e.g., they should not be construed as an emphasis on importance or order). Expressions such as "including," "comprising," and "having" also mean "not limited to" (certain units, components, materials, steps, etc.).

[0020] This application provides a method for identifying broken wires in steel wire ropes based on the magnetic memory effect, such as... Figure 1 As shown, the method for identifying broken wires in a steel wire rope based on the magnetic memory effect may include the following steps: Step 101: Obtain the magnetic memory detection signal of the steel wire rope to be tested.

[0021] In the embodiments of this application, the wire rope broken wire identification device based on magnetic memory effect can first acquire the magnetic memory detection signal of the wire rope to be tested.

[0022] In the embodiments of this application, the magnetic memory detection signal refers to the leakage magnetic field signal of the steel wire rope acquired using the principle of magnetic memory effect. When the steel wire rope is under load, the stress concentration at the broken wire causes irreversible reorientation of the magnetic domain structure of the ferromagnetic material, generating a leakage magnetic field. This signal includes two components: a tangential component (parallel to the axis of the steel wire rope) and a normal component (perpendicular to the axis of the steel wire rope).

[0023] In some embodiments of this application, a high-precision triaxial magnetic memory sensor can be used to scan along the axial direction of the wire rope, simultaneously acquiring the tangential component Bx (parallel to the wire rope axis) and the normal component By. The sampling frequency is set to 2kHz, and the scanning speed is 0.2m / s. The original noisy signal is as follows: Figure 2 As shown, the acquired magnetic memory detection signal contains data from two channels: the tangential component signal and the normal component signal.

[0024] Step 102: Perform noise reduction processing on the magnetic memory detection signal to obtain the noise-reduced signal.

[0025] In the embodiments of this application, the wire rope broken wire identification device based on magnetic memory effect can perform noise reduction processing on the magnetic memory detection signal after acquiring the magnetic memory detection signal of the wire rope to be tested, and obtain the noise-reduced signal.

[0026] In some embodiments of this application, when the wire rope breakage identification device based on magnetic memory effect performs noise reduction processing on the magnetic memory detection signal to obtain the noise-reduced signal, it can perform multi-layer wavelet decomposition on the tangential and normal component signals in the magnetic memory detection signal using wavelet basis functions to obtain high-frequency coefficients at each layer; it can then perform quantization processing on the high-frequency coefficients at each layer using a soft threshold function to obtain quantized high-frequency coefficients; wherein, the soft threshold function is a function that sets coefficients with absolute values ​​less than a threshold to zero and shrinks coefficients with absolute values ​​greater than or equal to the threshold towards zero; and finally, it reconstructs the quantized high-frequency coefficients to obtain the noise-reduced signal.

[0027] In the embodiments of this application, the db4 wavelet basis function is selected, and the tangential component signal and the normal component signal in the magnetic memory detection signal are decomposed into three layers of wavelet decomposition. The wavelet decomposition decomposes the signal into low-frequency approximation coefficients and high-frequency detail coefficients of each layer; wherein, the high-frequency coefficients mainly correspond to the noise components in the signal.

[0028] For example, for each high-frequency coefficient c, if < If, then set to 0; if ≥ Then set to .

[0029] In some embodiments of this application, the quantized high-frequency coefficients of each layer are reconstructed using inverse wavelet transform with the unprocessed low-frequency approximation coefficients to obtain a denoised signal with a high signal-to-noise ratio. The reconstructed signal retains the main characteristics of the broken wire signal while effectively suppressing on-site electromagnetic interference and noise.

[0030] Step 103: Determine the zero-crossing position of the tangential component and the peak position of the normal component based on the denoised signal; wherein, the zero-crossing position of the tangential component represents the spatial coordinates of the tangential component signal crossing the zero point in the magnetic memory detection signal, and the peak position of the normal component represents the spatial coordinates of the normal component signal reaching its maximum value in the magnetic memory detection signal.

[0031] In the embodiments of this application, the wire rope breakage identification device based on magnetic memory effect can, after denoising the magnetic memory detection signal to obtain the denoised signal, determine the zero-crossing position of the tangential component and the peak position of the normal component based on the denoised signal; wherein, the zero-crossing position of the tangential component represents the spatial coordinates of the tangential component signal crossing the zero point in the magnetic memory detection signal, and the peak position of the normal component represents the spatial coordinates of the normal component signal reaching its maximum value in the magnetic memory detection signal.

[0032] In the embodiments of this application, the tangential component signal refers to the magnetic field component signal parallel to the axis of the wire rope in the magnetic memory detection signal. At the center of the broken wire, the tangential component exhibits a zero-crossing characteristic.

[0033] In the embodiments of this application, the normal component signal refers to the magnetic field component signal perpendicular to the axis of the wire rope in the magnetic memory detection signal. At the center of the broken wire, the normal component exhibits a peak characteristic.

[0034] In the embodiments of this application, the zero-crossing position of the tangential component and the peak position of the normal component can jointly indicate the approximate position of the broken wire center in the axial direction of the wire rope.

[0035] Step 104: Determine the spatial phase delay based on the zero-crossing position of the tangential component and the peak position of the normal component.

[0036] In the embodiments of this application, the wire rope breakage identification device based on magnetic memory effect can determine the spatial phase delay based on the zero-crossing position of the tangential component and the peak position of the normal component after determining the zero-crossing position of the tangential component and the peak position of the normal component based on the noise-reduced signal.

[0037] In some embodiments of this application, when determining the spatial phase delay based on the zero-crossing position of the tangential component and the peak position of the normal component, the wire rope breakage identification device based on the magnetic memory effect can determine the absolute value of the spatial distance between the zero-crossing position of the tangential component and the peak position of the normal component as the spatial phase delay.

[0038] For example, the spatial phase delay can be expressed as: ,in, These are the spatial coordinates corresponding to the zero-crossing points of the tangential component. The coordinates are the spatial coordinates corresponding to the peak position of the normal component.

[0039] In the embodiments of this application, the spatial phase delay can be regarded as a physical indicator of the wire breakage depth—the external wire breakage is directly exposed to the air due to the magnetic field, and the tangential zero-crossing point and the normal peak almost coincide ( Smaller); the magnetic field generated by the internal broken wire needs to pass through the undamaged steel wire layer, and due to magnetic shielding and eddy current effects, the normal peak will be delayed ( (relatively large)

[0040] Step 105: If the spatial phase delay is greater than or equal to a preset threshold, perform deconvolution recovery processing on the denoised signal to obtain the recovered magnetic memory detection signal, and determine the multidimensional feature vector based on the recovered magnetic memory detection signal; otherwise, determine the multidimensional feature vector based on the denoised signal.

[0041] In the embodiments of this application, the wire rope breakage identification device based on magnetic memory effect can determine the spatial phase delay amount after determining the zero-crossing position of the tangential component and the peak position of the normal component.

[0042] In the embodiments of this application, the deconvolution recovery process is a preprocessing operation to address the problem of waveform broadening and amplitude attenuation caused by magnetic shielding of internal broken wires. It uses the standard magnetic pulse response signal of the lossless region as the kernel function to perform deconvolution operation on the detection signal to recover the intrinsic waveform of the signal.

[0043] In some embodiments of this application, when the wire rope broken wire identification device based on magnetic memory effect performs deconvolution recovery processing on the denoised signal to obtain the recovered magnetic memory detection signal, it can acquire the first pulse response signal of the wire rope under test and use the first pulse response signal as a kernel function; wherein, the first pulse response signal characterizes the magnetic pulse response signal of the wire rope under test in the lossless region; deconvolution operation is performed on the denoised signal and the kernel function to obtain the deconvolution result; frequency domain Wiener filtering processing is performed on the deconvolution result to obtain the recovered magnetic memory detection signal.

[0044] In the embodiments of this application, the kernel function can be understood as a standard magnetic pulse response signal (first pulse response signal) collected in the non-destructive region of the wire rope, which serves as a reference template for deconvolution operation.

[0045] In the embodiments of this application, frequency-domain Wiener filtering refers to filtering the deconvolution result in the frequency domain, which can be used to suppress noise amplification during the deconvolution process. The recovered magnetic memory detection signal has a narrow pulse shape similar to that of the external broken wire, and its peak value has a stable linear relationship with the intrinsic width and the number of broken wires.

[0046] In the embodiments of this application, the preset threshold can be 8mm; when When the thickness is ≥8mm, it indicates a significant magnetic shielding effect and severe signal distortion, requiring deconvolution recovery processing to restore the intrinsic waveform of the signal. When the value is less than 8mm, it indicates that the magnetic shielding effect is weak or in an intermediate transition state. In this case, deconvolution recovery is skipped, and the original signal after noise reduction is used directly for feature extraction. Regardless of whether deconvolution recovery is performed, the subsequent neural network outputs the number of broken wires.

[0047] For example, the noise-reduced signal With kernel function Perform deconvolution operation: The purpose of deconvolution is to recover the intrinsic waveform of the source signal from the observed signal, eliminating the waveform broadening and amplitude attenuation caused by magnetic shielding.

[0048] In the embodiments of this application, a multidimensional feature vector refers to a vector composed of multiple features extracted from the processed signal, which is a total of 6-dimensional feature vectors.

[0049] In some embodiments of this application, when determining a multidimensional feature vector based on the recovered magnetic memory detection signal, the wire rope broken wire identification device based on the magnetic memory effect can extract the first peak value, the first peak-to-peak value, the first waveform energy, the first zero-crossing rate, and the first waveform broadening coefficient of the recovered magnetic memory detection signal; and determine the multidimensional feature vector based on the first peak value, the first peak-to-peak value, the first waveform energy, the first zero-crossing rate, the first waveform broadening coefficient, and the spatial phase delay.

[0050] In some embodiments of this application, when determining a multidimensional feature vector based on the denoised signal, the wire rope breakage identification device based on the magnetic memory effect can extract the second peak value, the second peak-to-peak value, the second waveform energy, the second zero-crossing rate, and the second waveform broadening coefficient of the denoised signal; and determine the multidimensional feature vector based on the second peak value, the second peak-to-peak value, the second waveform energy, the second zero-crossing rate, the second waveform broadening coefficient, and the spatial phase delay.

[0051] In the embodiments of this application, the peak value can reflect the maximum leakage magnetic field strength generated by the broken wire. The peak-to-peak value is the difference between the maximum and minimum values ​​of the signal, reflecting the dynamic range of the magnetic field and the waveform energy. The total leakage flux is reflected by the zero-crossing rate, which is the number of times the signal crosses the zero point per unit time. It represents the frequency of magnetic field polarity reversal. The waveform broadening factor W = half width at half maximum / standard pulse width. For internal broken wires, W should be close to 1 after deconvolution recovery. If it is still significantly greater than 1, it indicates that multiple broken wires are stacked.

[0052] Step 106: Determine the number of broken wires in the steel wire rope under test using the trained neural network model and multidimensional feature vectors.

[0053] In the embodiments of this application, the wire rope broken wire identification device based on magnetic memory effect can perform deconvolution recovery processing on the denoised signal when the spatial phase delay is greater than or equal to a preset threshold, to obtain the recovered magnetic memory detection signal, and determine the multidimensional feature vector based on the recovered magnetic memory detection signal; otherwise, after determining the multidimensional feature vector based on the denoised signal, the number of broken wires of the wire rope to be tested is determined by the trained neural network model and the multidimensional feature vector.

[0054] In the embodiments of this application, a multidimensional feature vector is input into the trained neural network model, and the model outputs a continuous value through forward inference, which is then rounded to obtain the number of broken wires.

[0055] In some embodiments of this application, the wire rope broken wire identification device based on magnetic memory effect can construct an initial neural network model; wherein, the initial neural network model includes an input layer, a hidden layer, and an output layer, the number of nodes in the input layer is equal to the dimension of the multidimensional feature vector, and the number of nodes in the output layer is 1; the initial weights and initial thresholds of the initial neural network model are optimized by a genetic algorithm to obtain optimized weights and optimized thresholds; a first neural network model is determined based on the optimized weights, optimized thresholds, and the initial neural network model; the first neural network model is trained using a training sample set to obtain a trained neural network model; wherein, the training sample set includes multiple samples, each sample containing a feature vector sample and a corresponding broken wire count label.

[0056] In the embodiments of this application, the initial neural network model is a BP neural network, whose structure includes an input layer, a hidden layer and an output layer; the number of nodes in the input layer is equal to the dimension of the multidimensional feature vector (6 nodes), the number of nodes in the hidden layer is 12 nodes, and the number of nodes in the output layer is 1 node (outputting the number of broken wires).

[0057] In the embodiments of this application, the optimal initial weights and thresholds searched by the genetic algorithm are assigned to the initial neural network model to obtain the first neural network model. The first neural network model has obtained initial parameters that are beneficial for internal filament breakage identification.

[0058] In the embodiments of this application, when training the first neural network model using the training sample set, the gradient descent method with momentum can be used to train the first neural network model, with a learning rate of 0.008, a target error of 0.0005, and a maximum number of iterations of 500. After training, the trained neural network model is obtained and used for inference of the number of broken wires in the on-site detection stage.

[0059] In some embodiments of this application, when the wire rope broken wire identification device based on magnetic memory effect optimizes the initial weights and thresholds of the initial neural network model using a genetic algorithm to obtain optimized weights and thresholds, it can encode the initial weights and thresholds to generate an initial population; wherein each individual in the initial population contains a set of weight vectors and threshold vectors; the weight vectors and threshold vectors of each individual in the initial population are assigned to the initial neural network model, and the predicted values ​​of each sample in the training sample set are determined through the assigned initial neural network model; the initial population is then used to determine the predicted values ​​of each sample in the training sample set. The weights of each sample are determined; based on the predicted value of each sample, the corresponding broken wire number label of each sample, and the weight of each sample, the weighted mean square error is determined; the reciprocal of the weighted mean square error is used as the fitness function to calculate the fitness value of each individual in the initial population; selection, crossover, and mutation operations are performed on the initial population based on the fitness value to generate a new generation population; the operation of generating a new generation population based on the fitness value is repeated until the target population that meets the convergence condition is obtained; the weight vector and threshold vector corresponding to the optimal individual in the target population are determined as the optimized weight and optimized threshold.

[0060] In the embodiments of this application, the weighted mean square error refers to the loss function that averages the error of each sample by multiplying it by different weights based on the mean square error. It is used to make the model pay more attention to high-weight samples (such as internal broken wire samples) in the optimization of genetic algorithms.

[0061] In some embodiments of this application, when determining the weight of each sample in the initial population, the wire rope broken wire identification device based on magnetic memory effect can set the weight of the sample with the broken wire type of "internal" as the first weight value and set the weight of the sample with the broken wire type of "external" as the second weight value; wherein, the first weight value is greater than the second weight value.

[0062] It should be noted that this application has made a weighted improvement to address the imbalance between internal and external broken wire samples—introducing a sample weighting mechanism into the fitness function to impose a higher penalty on the prediction error of internal broken wires.

[0063] In the embodiments of this application, all initial weights and initial thresholds of the BP neural network are encoded as real number vectors, serving as the genes of an individual in the genetic algorithm. Each individual contains a complete set of weight vectors and threshold vectors.

[0064] In the embodiments of this application, for each individual in the population, its weight vector and threshold vector are assigned to the initial neural network model, and then forward computation is performed on each sample in the training sample set to obtain the predicted value of the number of broken wires for that sample.

[0065] In some embodiments of this application, the first weight value is 3 and the second weight value is 1. This can amplify the prediction error of the internal broken wire samples by 3 times, thereby enabling the genetic algorithm to prioritize reducing the prediction error of the internal broken wire samples during the optimization process, that is, to prioritize optimizing the recognition accuracy of the internal broken wires.

[0066] For example, the weighted mean square error can be expressed as: WMSE = (1 / N) × (w) i ×(y i -ŷ i )²), where w i Let y be the weight of the i-th sample. i The actual number of broken wires, ŷ i These are predicted values. The fitness function is the reciprocal of WMSE, meaning a higher fitness value indicates a better weight / threshold combination for that individual.

[0067] In the embodiments of this application, the selection operation retains individuals with high fitness (the selection probability is proportional to the fitness value); the crossover operation exchanges some genes of two individuals with a preset probability (crossover probability is 0.75); the mutation operation randomly changes some gene values ​​of an individual with a certain probability.

[0068] In the embodiments of this application, a new generation of population is continuously generated iteratively until the convergence condition is met (such as reaching the maximum number of generations 150, or the fitness value no longer significantly improving). The weight vector and threshold vector corresponding to the individual with the highest fitness in the final population are determined as the optimized weight and optimized threshold.

[0069] In the embodiments of this application, an adaptive identification scheme is constructed with the quantification of broken wires as the ultimate goal and the physical mechanism of magnetic memory effect as the driving force. First, this application abandons the traditional method of treating "distinguishing between internal and external broken wires" and "quantifying the number of broken wires" as two independent tasks, and instead utilizes the spatial phase delay between the tangential zero-crossing point and the normal peak of the magnetic memory signal. As a physical indicator of the degree of signal distortion, this physical prior adaptively determines whether to perform deconvolution recovery processing on the signal—when... When the thickness is ≥8mm, the magnetic shielding effect is considered significant, and deconvolution recovery is performed on the signal to eliminate waveform broadening and amplitude attenuation; when For wire breakages <8mm, this process is skipped, ensuring that internal and external wire breakage signals are as comparable as possible before feature extraction. The neural network then outputs a unified number of broken wires. This design increases the correlation coefficient between the internal wire breakage signal and the number of broken wires from 0.61 to 0.89, fundamentally improving the overall accuracy of wire breakage quantification. A sample weighting mechanism is introduced into the training framework of the genetic algorithm-optimized BP neural network. Addressing the weakness of internal wire breakage signals and the difficulty in sample identification, the weight of internal wire breakage samples is set to three times that of external wire breakage samples. The reciprocal of the weighted mean square error is used as the fitness function, allowing the genetic algorithm to prioritize searching the weight parameter space sensitive to internal wire breakages during optimization. This effectively overcomes the inherent defects of traditional GA-BP models—overfitting strong signals and underfitting weak signals—and achieves, for the first time, targeted optimization of unbalanced samples in the task of quantifying wire breakages in steel wire ropes. A complete technical closed loop was constructed, encompassing signal acquisition and the quantification of broken wires. This loop covers triaxial magnetic memory signal acquisition, wavelet threshold denoising, spatial phase delay calculation and determination, deconvolution recovery, multi-dimensional feature extraction (peak value, peak-to-peak value, waveform energy, zero-crossing rate, waveform broadening coefficient, and spatial phase delay, totaling 6 dimensions), and weighted GA-BP quantization recognition. Experimental results show that the average recognition accuracy for 1 to 6 broken wires reaches 96.78%, the false alarm rate for internal broken wires is reduced from 10.0% in traditional methods to 3.8%, and the missed alarm rate is reduced from 11.1% to 5.0%. Furthermore, the trained model can perform a single detection in less than 0.5 seconds after field deployment, combining high precision and high real-time performance. It is suitable for online detection of broken wires in steel wire ropes in high-noise industrial scenarios such as mine hoisting and elevator traction.

[0070] Based on the above embodiments, in another embodiment of this application, a standard steel wire rope sample is used, the sample contains 1 to 6 broken wires, and two cases are set up: external broken wires and simulated internal broken wires. 300 samples are collected for each case, for a total of 600 samples.

[0071] Intermediate spatial phase delay The calculation and decision-making steps serve only as a basis for preprocessing branch decisions, based on... The degree to which the signal is affected by magnetic shielding is assessed to determine whether a deconvolution recovery operation is necessary. The purpose of this assessment is to ensure that internal and external broken wire signals are as comparable as possible before feature extraction, thereby improving the network's overall quantization accuracy for broken wire counts. The network ultimately outputs a uniform number of broken wires, without distinguishing between different types of broken wires.

[0072] Step 1: Signal acquisition and multi-component extraction based on magnetic memory effect.

[0073] A high-precision triaxial magnetic memory sensor is used to scan along the axial direction of the wire rope, simultaneously acquiring the tangential component Bx (parallel to the wire rope axis) and the normal component By. The sampling frequency is set to 2kHz, and the scanning speed is 0.2m / s. The original noisy signal is as follows: Figure 2 As shown.

[0074] Step 2: Wavelet thresholding for noise reduction.

[0075] To address on-site electromagnetic interference, the db4 wavelet basis function was selected to perform three-level decomposition on both the tangential and normal components of the signal. High-frequency coefficients at each level were quantized using a soft thresholding function, with the threshold determined based on unbiased risk estimation. The reconstructed signal yielded a high signal-to-noise ratio denoised signal, laying the foundation for subsequent feature extraction. Figure 3 The image shows the signal waveform after noise reduction using wavelet basis functions.

[0076] Step 3: Determining internal and external wire breakage based on the physical mechanism of magnetic memory.

[0077] For each suspected broken wire signal segment after noise reduction, accurately calculate the spatial phase delay between the zero-crossing position of the tangential component and the peak position of the normal component. .

[0078] when When the diameter is ≥8mm, it indicates a significant magnetic shielding effect and severe signal distortion, which is determined to be an internal wire breakage type, requiring step four for deconvolution recovery processing; when When the value is less than 8mm, it indicates that the magnetic shielding effect is weak or in an intermediate transition state. Skip the deconvolution recovery step and proceed directly to step five to extract features.

[0079] Regardless of the determination result, the BP network will uniformly output the number of broken wires. Even without deconvolution recovery, samples smaller than 8mm It itself is still used as the 6th dimension feature input network, and the network adaptively learns its mapping contribution to the number of broken wires during training, without the need to manually set the subdivision rules for the transition type.

[0080] Step 4: Deconvolution waveform recovery for internal broken wires.

[0081] For signals identified as having internal wire breaks, the waveform broadens and amplitude attenuates due to the wires passing through multiple layers of steel wire, resulting in severe distortion when directly extracting features. This application introduces deconvolution recovery processing: First, a standard magnetic pulse response signal is acquired in the undamaged area of ​​the steel wire rope as a kernel function. The noise-reduced signal With kernel function Perform deconvolution operation.

[0082] Frequency-domain Wiener filtering is employed to suppress noise amplification. The recovered signal exhibits a narrow pulse shape similar to that of external broken wires, and its peak value and intrinsic width show a stable linear relationship with the number of broken wires, significantly improving quantization reliability. For external broken wires, the deconvolution step is skipped, and the original signal after noise reduction is used directly.

[0083] Step 5: Multi-feature extraction and dataset construction.

[0084] For the processed signal segment, extract the following five key features: Peak value: Reflects the maximum leakage magnetic field strength generated by the broken wire. Peak-to-peak value: Reflects the dynamic range of the magnetic field. Waveform energy: Reflects the total leakage magnetic flux. Zero-crossing rate: The number of times the signal crosses zero points per unit time, characterizing the frequency of magnetic field polarity reversal. Waveform broadening factor W = half-width at half maximum (FWHM) / standard pulse width. For internal broken wires, W should be close to 1 after deconvolution recovery; if it is still significantly greater than 1, it indicates multiple broken wires stacked together. Simultaneously, the phase delay obtained in step three... It is also used as an auxiliary feature input.

[0085] In some embodiments of this application, the feature vector of each sample used for training may include six dimensions: peak value, peak-to-peak value, waveform energy, zero-crossing rate, waveform broadening coefficient, and spatial phase delay. A dataset containing 600 samples is constructed and randomly divided into a training set and a test set in a 7:3 ratio, ensuring a balanced proportion of internal and external broken wire samples in the training set and the test set.

[0086] Step 6: Optimize the construction of the BP neural network using a weighted genetic algorithm.

[0087] (1) Network structure design. Input layer: 6 nodes. Hidden layer: 12 nodes. Output layer: 1 node, outputting the number of broken wires.

[0088] (2) Genetic algorithm for weighted optimization of internal and external wire breakage imbalance problem. Conventional GA uses the inverse of the network mean square error as the fitness function.

[0089] In addition, to overcome the defect of the model favoring strong signals, this application designs a weighted mean square error.

[0090] Step 7: Model Validation and Results.

[0091] The trained weighted GA-BP model proposed in this application was evaluated using a test set. The average results after 5 independent replicate experiments are as follows: Overall accuracy: 96.78%; average precision: 96.2%; average recall: 95.1%; F1 score: 95.6%.

[0092] Internal wire breakage false alarm rate: 3.8%, false alarm rate: 5.0%; compared with the conventional GA-BP model without deconvolution recovery and weighted optimization, this application reduces the false alarm rate of internal wire breakage by 62% and the false alarm rate by 55%. Figure 4 The figure shows the prediction results of the number of broken fibers for each sample in the test set using the model proposed in this application. Table 1 below shows the confusion matrix of the classification results on the test set using the model proposed in this application; the evaluation metrics include accuracy, precision, recall, and F1 score. The confusion matrix shows that the model proposed in this application has reached a practical level for the quantitative identification of 1 to 6 broken fibers.

[0093] Table 1:

[0094] Steps S1 to S6 above constitute the complete offline training phase. After the trained model is deployed to the field, only signal acquisition, noise reduction, and... Calculation and determination, deconvolution (only) The detection process consists of six sub-steps: sample size ≥8mm, feature extraction, and forward inference via a BP network. The time for a single detection does not exceed 0.5 seconds.

[0095] The working principle of this application can be understood as a hybrid driving strategy that combines the physical mechanism of magnetic memory effect with deep learning, which can be divided into the following four levels: When a steel wire rope is under load, stress concentration at the broken wire causes irreversible reorientation of the magnetic domain structure of the ferromagnetic material, generating a leakage magnetic field. The tangential component of this leakage magnetic field exhibits a zero-crossing point at the center of the broken wire, while the normal component shows a peak value. For external broken wires, since the magnetic field is directly exposed to the air, the peak value of the tangential component almost coincides with the zero-crossing point of the normal component; however, the magnetic field generated by internal broken wires must pass through the intact steel wire layer. Due to the magnetic shielding and eddy current effect of the steel wire layer, the peak value of the normal component is delayed. It can reach over 8mm. This application is the first to utilize... As a physical indicator of wire breakage depth, it enables coarse classification of internal / external wire breakage.

[0096] To address the waveform broadening and amplitude attenuation caused by the transmission path of internal broken wires, this application employs a deconvolution method, using the standard magnetopulse response of the lossless region as the kernel function to recover the intrinsic waveform of the source signal from the observed signal. This processing significantly reduces the nonlinear distortion caused by magnetic shielding, restoring an approximately linear relationship between the characteristics of internal broken wires and the number of broken wires, laying the foundation for subsequent quantization and identification.

[0097] Traditional GA-BP models treat all training samples equally, leading them to tend to learn easily identifiable external broken wire features while ignoring weak signals from internal broken wires. This application introduces a sample weighting mechanism into the GA fitness function, imposing a higher penalty on the prediction error of internal broken wires. Simultaneously, through dynamic mutation probabilities, the search range is expanded in the early stages of evolution to cover the parameter space corresponding to internal broken wires, with fine-tuning in the later stages. This enables the genetic algorithm to provide the BP network with a set of initial weights and thresholds that are particularly sensitive to internal broken wires.

[0098] The processed 6-dimensional feature vector is input into the optimized BP network. The network learns the complex mapping relationship between the above features and the number of broken filaments through multiple nonlinear transformations. Finally, it outputs a continuous value, which, after rounding, yields a quantized result of 1 to 6 broken filaments.

[0099] In the embodiments of this application, according to The physical determination of wire breakage depth utilizes the unique multi-component spatial relationship of the magnetic memory effect, possessing clear physical significance. A deconvolution recovery preprocessing technique addressing internal wire breakage waveform distortion uses the standard magnetic pulse response of the lossless region as the kernel function to perform deconvolution operations on the detected internal wire breakage signal, restoring its intrinsic width and amplitude, thereby eliminating nonlinear distortion caused by magnetic shielding. This technique provides a clean feature foundation for internal wire breakage quantification. A weighted genetic algorithm-optimized BP neural network model is employed; the fitness function uses weighted mean square error, assigning higher weights to internal wire breakage samples, allowing the genetic algorithm to prioritize optimizing the identification accuracy of internal wire breaks during evolution. A dynamic adjustment strategy is adopted for mutation probability: high mutation in the early stages to broadly search the parameter space corresponding to weak signals, and low mutation in the later stages for fine-grained convergence. The network structure and the number of hidden layer nodes are specifically designed based on the feature dimension of the magnetic memory signal and the sample imbalance. Feature vectors are constructed by integrating multi-dimensional features such as phase delay, waveform broadening coefficient, and energy. These features include six dimensions: peak value, peak-to-peak value, waveform energy, zero-crossing rate, waveform broadening coefficient, and phase delay. Among these, the waveform broadening coefficient and phase delay are new features specifically introduced to distinguish between internal and external wire breaks. A complete closed-loop method from signal acquisition to quantization output is developed, including triaxial magnetic memory signal acquisition, wavelet denoising, coarse phase delay classification, internal wire breakage deconvolution recovery, multi-dimensional feature extraction, and weighted GA-BP quantization recognition. This forms a technical system specifically for high-precision quantization of internal and external wire breaks in steel wire ropes, suitable for high-noise industrial scenarios such as mining hoists and lifting machinery.

[0100] In summary, compared with the prior art, this application mainly has the following technical effects: (1) With the quantification of broken wires as the ultimate goal, the physical prior of phase delay is used to guide adaptive preprocessing. By measuring the spatial phase delay between the tangential zero-crossing point and the normal peak, it is used as a physical indicator of the degree of signal distortion. The larger the spatial phase delay, the more significant the magnetic shielding and the more severe the signal distortion. Based on this, it is adaptively determined whether to apply deconvolution recovery to the signal. After deconvolution recovery, the correlation coefficient between the internal broken wire signal and the number of broken wires is increased from 0.61 to 0.89, which significantly improves the quantification accuracy of the overall number of broken wires. (2) An effective deconvolution recovery preprocessing method is proposed to address the distortion of internal broken wire signals. By using frequency-domain Wiener filtering deconvolution with the standard magnetic pulse response of the lossless region as the kernel function, the characteristic distortion problem caused by magnetic shielding and waveform broadening of internal broken wires is solved. Compared with the conventional GA-BP model without deconvolution and weighted optimization, this application reduces the false alarm rate of internal broken wires from 10.0% to 3.8% and the missed alarm rate from 11.1% to 5.0%, significantly improving the overall accuracy of broken wire root quantification. (3) A weighted genetic algorithm optimization strategy for imbalanced samples was designed: For the first time, a sample weight mechanism was introduced in the quantitative identification of broken wire roots in steel wire rope, so that the optimization process automatically favors weak signals and high-value samples such as internal broken wires. Combined with a dynamic mutation probability strategy (high mutation in the early stage of evolution to extensively search the parameter space corresponding to weak signals, and low mutation in the later stage to fine convergence), it overcomes the inherent defects of conventional GA / PSO optimized BP network in overfitting strong signals and underfitting weak signals. (4) High accuracy and robustness in quantifying the total number of broken wires: After verification with standard samples, the average recognition accuracy for 1 to 6 broken wires reached 96.78%, precision 96.2%, recall 95.1%, and F1 score 95.6%. All indicators are significantly better than traditional BP neural network and unweighted GA-BP model. (5) Complete technical closed loop, suitable for complex industrial sites: from multi-component magnetic memory signal acquisition, wavelet noise reduction, The entire method, from calculation and judgment, deconvolution recovery, and weighted feature extraction to the quantification output of broken wires in GA-WBP, has a good ability to suppress working noise. After the trained model is deployed to the field, the single detection time does not exceed 0.5 seconds, making it particularly suitable for online detection of the number of broken wires in steel wire ropes such as those used in coal mine hoists and elevator traction ropes.

[0101] Based on the above embodiments, another embodiment of this application provides a wire rope broken wire identification device based on magnetic memory effect, such as... Figure 5As shown, the wire rope broken wire identification device 1 based on magnetic memory effect proposed in this application embodiment may include a processor 11 and a memory 12 storing instructions executable by the processor 11; further, the wire rope broken wire identification device 1 based on magnetic memory effect may also include a communication interface 13 and a bus 14 for connecting the processor 11, the memory 12 and the communication interface 13.

[0102] In the embodiments of this application, the processor 11 can be at least one of the following: Application-Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), Central Processing Unit (CPU), controller, microcontroller, and microprocessor. It is understood that for different devices, the electronic device used to implement the above-mentioned processor function can also be other, and this application embodiment does not specifically limit it. The memory 12 can be connected to the processor 11, wherein the memory 12 is used to store executable program code, which includes computer operation instructions. The memory 12 may include high-speed RAM memory, and may also include non-volatile memory, such as at least two disk drives.

[0103] In embodiments of this application, bus 14 is used to connect communication interface 13, processor 11 and memory 12 to enable communication between these devices.

[0104] In embodiments of this application, memory 12 is used to store instructions and data.

[0105] In practical applications, the aforementioned memory 12 can be volatile memory, such as random-access memory (RAM), or non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); or a combination of the above types of memory, and provide instructions and data to the processor 11.

[0106] Furthermore, in this embodiment, the functional modules can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional module.

[0107] If the integrated unit is implemented as a software functional module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the method of this embodiment.

[0108] Specifically, the program instructions corresponding to the wire rope broken wire identification method based on magnetic memory effect in this embodiment can be stored on storage media such as optical discs and hard disks. When the program instructions corresponding to the wire rope broken wire identification method based on magnetic memory effect in the storage medium are read or executed by a wire rope broken wire identification device based on magnetic memory effect, the following steps are included: Acquire the magnetic memory detection signal of the steel wire rope under test; The magnetic memory detection signal is denoised to obtain the denoised signal. The zero-crossing position of the tangential component and the peak position of the normal component are determined based on the denoised signal. The zero-crossing position of the tangential component represents the spatial coordinates of the tangential component signal crossing the zero point in the magnetic memory detection signal, and the peak position of the normal component represents the spatial coordinates of the normal component signal reaching its maximum value in the magnetic memory detection signal. The spatial phase delay is determined based on the zero-crossing position of the tangential component and the peak position of the normal component. If the spatial phase delay is greater than or equal to a preset threshold, the denoised signal is deconvolutionally recovered to obtain the recovered magnetic memory detection signal, and the multidimensional feature vector is determined based on the recovered magnetic memory detection signal; otherwise, the multidimensional feature vector is determined based on the denoised signal. The number of broken wires in the steel wire rope under test is determined by using a trained neural network model and multidimensional feature vectors.

[0109] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0110] This application is described with reference to schematic diagrams and / or block diagrams of implementations of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each step and / or block in the schematic diagrams and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more steps of the schematic diagrams and / or one or more blocks in the block diagrams.

[0111] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in the implementation flow diagram. Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0112] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0113] The above embodiments are merely preferred embodiments provided to fully illustrate this application, and the scope of protection of this application is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on this application are all within the scope of protection of this application.

Claims

1. A method for identifying broken wires in steel wire ropes based on magnetic memory effect, characterized in that, The method includes: Acquire the magnetic memory detection signal of the steel wire rope under test; The magnetic memory detection signal is subjected to noise reduction processing to obtain the noise-reduced signal; The zero-crossing position of the tangential component and the peak position of the normal component are determined based on the denoised signal; wherein, the zero-crossing position of the tangential component represents the spatial coordinates of the tangential component signal crossing the zero point in the magnetic memory detection signal, and the peak position of the normal component represents the spatial coordinates of the normal component signal reaching its maximum value in the magnetic memory detection signal; The spatial phase delay is determined based on the zero-crossing position of the tangential component and the peak position of the normal component. If the spatial phase delay is greater than or equal to a preset threshold, the denoised signal is deconvolutionally recovered to obtain the recovered magnetic memory detection signal, and a multidimensional feature vector is determined based on the recovered magnetic memory detection signal; otherwise, a multidimensional feature vector is determined based on the denoised signal. The number of broken wires in the steel wire rope under test is determined by the trained neural network model and the multidimensional feature vector.

2. The method for identifying broken wires in a steel wire rope based on magnetic memory effect according to claim 1, characterized in that, Determining the spatial phase delay based on the zero-crossing position of the tangential component and the peak position of the normal component includes: The absolute value of the spatial distance between the zero-crossing position of the tangential component and the peak position of the normal component is determined as the spatial phase delay.

3. The method for identifying broken wires in a steel wire rope based on magnetic memory effect according to claim 1, characterized in that, The step of performing deconvolution recovery processing on the denoised signal to obtain the recovered magnetic memory detection signal includes: The first pulse response signal of the steel wire rope under test is obtained, and the first pulse response signal is used as a kernel function; wherein, the first pulse response signal characterizes the magnetic pulse response signal of the steel wire rope under test in the non-destructive region; The denoised signal and the kernel function are deconvolved to obtain the deconvolution result. The deconvolution result is subjected to frequency domain Wiener filtering to obtain the recovered magnetic memory detection signal.

4. The method for identifying broken wires in a steel wire rope based on magnetic memory effect according to claim 1, characterized in that, The determination of the multidimensional feature vector based on the recovered magnetic memory detection signal includes: Extract the first peak value, first peak-to-peak value, first waveform energy, first zero-crossing rate, and first waveform broadening coefficient of the recovered magnetic memory detection signal; The multidimensional feature vector is determined based on the first peak value, the first peak-to-peak value, the first waveform energy, the first zero-crossing rate, the first waveform broadening factor, and the spatial phase delay. Accordingly, determining the multidimensional feature vector based on the denoised signal includes: Extract the second peak value, second peak-to-peak value, second waveform energy, second zero-crossing rate, and second waveform broadening factor of the noise-reduced signal; The multidimensional feature vector is determined based on the second peak value, the second peak-to-peak value, the second waveform energy, the second zero-crossing rate, the second waveform broadening coefficient, and the spatial phase delay.

5. The method for identifying broken wires in a steel wire rope based on magnetic memory effect according to claim 1, characterized in that, The method further includes: Construct an initial neural network model; wherein the initial neural network model includes an input layer, a hidden layer and an output layer, the number of nodes in the input layer is equal to the dimension of the multidimensional feature vector, and the number of nodes in the output layer is 1; The initial weights and initial thresholds of the initial neural network model are optimized using a genetic algorithm to obtain the optimized weights and optimized thresholds. The first neural network model is determined based on the optimized weights, the optimized thresholds, and the initial neural network model. The first neural network model is trained using a training sample set to obtain the trained neural network model; wherein, the training sample set includes multiple samples, each sample containing a feature vector sample and a corresponding label for the number of broken wires.

6. The method for identifying broken wires in a steel wire rope based on magnetic memory effect according to claim 5, characterized in that, The step of optimizing the initial weights and initial thresholds of the initial neural network model using a genetic algorithm to obtain optimized weights and optimized thresholds includes: The initial weights and the initial thresholds are encoded to generate an initial population; wherein each individual in the initial population contains a set of weight vectors and threshold vectors; The weight vector and threshold vector of each individual in the initial population are assigned to the initial neural network model, and the predicted value of each sample in the training sample set is determined by the assigned initial neural network model. Determine the weight of each sample in the initial population; Based on the predicted value of each sample, the number of broken wires corresponding to each sample, and the weight of each sample, the weighted mean square error is determined. The reciprocal of the weighted mean square error is used as the fitness function to calculate the fitness value of each individual in the initial population. Based on the fitness value, the initial population is subjected to selection, crossover, and mutation operations to generate a new generation population. Repeat the operation of generating a new generation of population based on fitness values ​​until a target population that meets the convergence condition is obtained. The weight vector and threshold vector corresponding to the best individual in the target population are determined as the optimized weight and optimized threshold.

7. The method for identifying broken wires in a steel wire rope based on magnetic memory effect according to claim 6, characterized in that, Determining the weight of each sample in the initial population includes: In the initial population, the weight of samples with internal broken wire type is set as the first weight value, and the weight of samples with external broken wire type is set as the second weight value. Wherein, the first weight value is greater than the second weight value.

8. The method for identifying broken wires in a steel wire rope based on magnetic memory effect according to claim 1, characterized in that, The step of denoising the magnetic memory detection signal to obtain a denoised signal includes: Multi-level wavelet decomposition is performed on the tangential and normal components of the magnetic memory detection signal using wavelet basis functions to obtain the high-frequency coefficients of each level. The high-frequency coefficients of each layer are quantized by a soft threshold function to obtain quantized high-frequency coefficients; wherein, the soft threshold function is a function that sets coefficients with absolute values ​​less than a threshold to zero and shrinks coefficients with absolute values ​​greater than or equal to the threshold to zero; The quantized high-frequency coefficients are reconstructed to obtain the denoised signal.

9. A wire rope broken wire identification device based on magnetic memory effect, characterized in that, The wire rope broken wire identification device based on magnetic memory effect includes a processor, a memory, and a memory storing executable instructions of the processor; when the instructions are executed by the processor, the wire rope broken wire identification method based on magnetic memory effect as described in any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the wire rope broken wire identification method based on magnetic memory effect as described in any one of claims 1 to 8.

Images