Anti-shearing alarm monitoring method and system for steel cable

Abstract

The invention relates to the technical field of steel cable anti-shearing alarm, and discloses an anti-shearing alarm monitoring method and system for a steel cable, and the method comprises the steps: building an anti-shearing monitoring circuit through arranging a resistance wire in the steel cable, and building a voltage collection circuit through a transistor and a divider resistance network. When the microcontroller outputs a control signal to switch on the transistor, two current paths are formed in the monitoring circuit. Voltage digital characteristic values are obtained by periodically sampling and digitally processing transistor source electrode voltage. And comparing a calculation result of the characteristic value with a preset standard value to obtain state characteristic data of the steel cable, and finally, distinguishing two abnormal states of steel cable shearing and wire short circuit by analyzing voltage distribution characteristics, so that the steel cable is effectively prevented from being damaged and short circuit cheating behaviors are effectively prevented.

Description

Technical Field

[0001] The present invention relates to the technical field of steel cable anti-shear alarm, and in particular to a steel cable anti-shear alarm monitoring method and system. Background Art

[0002] With the widespread application of smart locks in various scenarios, cable locks are widely used due to their convenience, safety and reliability. However, traditional cable locks face serious safety hazards in actual applications. Criminals can easily destroy the lock function by cutting the cable, causing the lock to fail and causing property losses.

[0003] Most of the wire cable locks on the market currently use simple mechanical structures or basic circuit detection solutions, which have problems such as low detection sensitivity and poor reliability. In particular, when the wire cable is cut, the conventional detection circuit cannot detect the abnormal state in time, and it is easy for criminals to deceive the system by short-circuiting ordinary wires, making the detection invalid. Although some improvement schemes have attempted to improve the anti-cutting performance by increasing the complexity of the detection circuit, these schemes often have disadvantages such as high cost, complex structure, and difficult installation and maintenance. In addition, the accuracy in identifying wire cable cutting and wire short-circuiting fraudulent behaviors still needs to be improved, and cannot meet the needs of actual applications. Summary of the invention

[0004] The invention provides a steel cable anti-shear alarm monitoring method and system, which are used to prevent the steel cable from being damaged and short-circuited fraudulently.

[0005] In a first aspect, the present invention provides a method for monitoring the anti-shear alarm of a steel cable, the method comprising: The steel cable is subjected to resistance wire laying processing, and one end of the resistance wire is connected to the system grounding terminal to obtain a steel cable structure including an anti-shear monitoring circuit; A voltage-dividing detection circuit is constructed according to the steel cable structure containing the anti-shear monitoring circuit, the drain of the transistor is connected to the power supply end through the first resistor, and the source of the transistor is connected to the analog conversion port of the microcontroller through the signal collection resistor to obtain a voltage collection circuit; Based on the voltage acquisition circuit, a control signal is outputted through the control terminal of the microcontroller to turn on the transistor, and a first current branch and a second current branch are formed in the steel cable structure containing the anti-shear monitoring circuit to obtain a detection loop; Periodically sampling the source voltage value of the transistor in the detection loop, performing digital-to-analog conversion on the sampled voltage value through the analog conversion port of the microcontroller to obtain a voltage digital characteristic value; According to the voltage digital characteristic value, the voltage division characteristic of the detection circuit is calculated and analyzed to obtain a calculation result, and the calculation result is compared with a preset standard resistance voltage division value to obtain the steel cable state characteristic data; Feature extraction is performed on the steel cable status feature data, and the steel cable cut state and the wire short circuit state are distinguished through voltage distribution characteristics, and an anti-cut alarm trigger signal is generated.

[0006] In a second aspect, the present invention provides an anti-shear alarm monitoring system for a steel cable, the anti-shear alarm monitoring system for a steel cable comprising: A processing module, used for performing resistance wire laying processing on the steel cable, connecting one end of the resistance wire to the system grounding terminal, and obtaining a steel cable structure including an anti-shear monitoring circuit; A construction module is used to construct a voltage-dividing detection circuit according to the steel cable structure containing the anti-shear monitoring circuit, connect the drain of the transistor to the power supply terminal through a first resistor, and connect the source of the transistor to the analog conversion port of the microcontroller through a signal collection resistor to obtain a voltage collection circuit; A control module, which is used to output a control signal through the control terminal of the microcontroller based on the voltage acquisition circuit to turn on the transistor, and to form a first current branch and a second current branch in the steel cable structure containing the anti-shear monitoring circuit to obtain a detection loop; A conversion module, used for periodically sampling the source voltage value of the transistor in the detection loop, performing digital-to-analog conversion on the sampled voltage value through the analog conversion port of the microcontroller to obtain a voltage digital characteristic value; A calculation module, used to calculate and analyze the voltage division characteristics of the detection circuit according to the voltage digital characteristic value, obtain a calculation result, and compare the calculation result with a preset standard resistance voltage division value to obtain the steel cable state characteristic data; The generation module is used to extract the characteristics of the steel cable status characteristic data, distinguish the steel cable cut state and the wire short circuit state through voltage distribution characteristics, and generate an anti-cut alarm trigger signal.

[0007] In the technical solution provided by the present invention, by setting a resistance wire with a specific resistance value inside the steel cable and using a voltage divider detection circuit to realize monitoring, the system can accurately sense the integrity state of the steel cable, effectively avoiding the limitations of the traditional mechanical detection scheme. The dual-branch detection structure constructed based on transistors and voltage divider networks realizes real-time monitoring of the steel cable state, and at the same time improves the stability and reliability of the detection signal through capacitor filtering and diode protection. By using digital processing and feature extraction technology to analyze the detection signal, the system can not only identify the shearing state of the steel cable, but also effectively distinguish the short-circuit deception behavior of the wire through the voltage distribution characteristics, thereby improving the safety of the anti-shearing scheme. The overall solution has a simple structure, low cost, and is easy to integrate into existing smart locks, with strong practicality and promotion value. By combining software and hardware, the complexity of the system implementation is reduced while ensuring the detection accuracy. By using periodic sampling and digital feature analysis, the system has strong anti-interference ability and environmental adaptability, and improves the reliability of the detection scheme in practical applications.

[0008] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0009] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a schematic diagram of an embodiment of a method for monitoring the anti-shearing alarm of a steel cable in an embodiment of the present invention; Figure 2 A schematic diagram of an embodiment of a detection circuit in an embodiment of the present invention; Figure 3 It is a schematic diagram of an embodiment of the anti-shear alarm monitoring system of the steel cable in the embodiment of the present invention. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0012] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device end including a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or device ends.

[0013] To facilitate understanding of this embodiment, a steel cable anti-shear alarm monitoring method disclosed in an embodiment of the present invention is first described in detail. Figure 1 As shown, the method comprises the following steps: It is understandable that the execution subject of the present invention may be a cable shearing alarm monitoring system, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject.

[0014] 101. Perform resistance wire laying processing on the steel cable, connect one end of the resistance wire to the system grounding terminal, and obtain a steel cable structure including an anti-shear monitoring circuit; 102. Construct a voltage division detection circuit according to the steel cable structure including the anti-shear monitoring circuit, connect the drain of the transistor to the power supply terminal through the first resistor, and connect the source of the transistor to the analog conversion port of the microcontroller through the signal collection resistor to obtain a voltage collection circuit; 103. Based on the voltage acquisition circuit, a control signal is outputted through the control terminal of the microcontroller to turn on the transistor, and a first current branch and a second current branch are formed in the steel cable structure containing the anti-shear monitoring circuit to obtain a detection loop; The detection circuit of the present invention is as follows Figure 2 As shown, Figure 2It is a schematic diagram of a detection loop provided by an embodiment of the present invention, and a resistance wire is arranged on a steel cable. Specifically, a resistance wire with a specific resistance value is arranged inside a conventional steel cable along its length direction, and one end of the resistance wire is electrically connected to the system ground wire GND, which is used to change the voltage-dividing characteristics of the system when the steel cable is cut. A field effect transistor Q1 is used as a core detection element, and the drain D of the transistor Q1 is connected to the 3.3V power supply terminal through a first resistor R1, and the source S of the transistor Q1 is connected to the analog-to-digital conversion interface ADC of the microcontroller MCU through a fourth resistor R4, thereby forming a voltage acquisition circuit. Among them, the voltage acquisition circuit includes: a transistor Q1, a first resistor R1, a fourth resistor R4, and a diode D1. The fourth resistor R4 is a signal acquisition resistor in this embodiment. A high-level signal is output through the control terminal CC_EN of the microcontroller MCU to turn on the transistor Q1. When Q1 is turned on, two current branches are formed in the steel cable monitoring circuit: the first current branch is the current from the 3.3V power supply through the first resistor R1, the drain D of the transistor Q1 to the source S, and then through the third resistor R3 and the resistance wire in the steel cable to the system ground GND; the second current branch is the current from the source S of the transistor Q1 through the capacitor C1, the fourth resistor R4 and the diode D1 and then into the ADC detection port of the MCU. These two branches together constitute a detection loop. A resistor is set in the steel cable lock to monitor the integrity of the steel cable; a high-level signal is output to the control circuit CC_EN through the microcontroller MCU to activate the current conduction process; the current enters the emitter D and collector S of the transistor Q1 from the resistor R1; the current flows through the resistor in the steel cable, part of it flows to the ground GND through the resistor R3, and the other part flows to the analog-to-digital conversion ADC detection port of the MCU through the capacitor C1, the resistor R4 and the diode D1; when the steel cable is cut or replaced, the system voltage divider resistance value changes, causing the voltage of the collector S of the transistor Q1 to change, and the change is transmitted to the MCU; the MCU determines whether the steel cable is cut based on the detected voltage change, and triggers an alarm after confirming the cutting behavior.

[0015] 104. Periodically sample the source voltage value of the transistor in the detection circuit, and perform digital-to-analog conversion on the sampled voltage value through the analog conversion port of the microcontroller to obtain a voltage digital characteristic value; Specifically, a sampling clock signal is applied to the transistor source to ensure that the sampling time point is accurate and stable. The sampling clock signal can ensure that the voltage value is collected consistently within the specified time interval, obtain a trigger signal for triggering voltage sampling, guide the system to sample at the correct time, and ensure that the voltage data is collected synchronously and has good time resolution. Based on the sampling trigger signal, the analog conversion port of the microcontroller performs analog-to-digital conversion on the voltage of the transistor source to obtain the original voltage sampling data, which reflects the actual voltage state of the transistor source at each sampling moment. The original voltage sampling data is processed to remove the interference components therein. The original sampling data is filtered by a multi-point moving average filtering method. The multi-point moving average filtering effectively reduces the random interference fluctuations in the voltage sampling data by averaging multiple sampling values ​​within a period of time, and obtains smoother and more stable filtered voltage data. A voltage sampling queue is established based on the filtered voltage data. The function of the voltage sampling queue is to store the voltage sampling data within a period of time so that the data can be analyzed later. Each time a new voltage sampling is performed, the newly sampled voltage data is stored in the voltage sampling queue in sequence, and the length of the queue is updated. By continuously updating the voltage sampling queue, a dynamic voltage data sequence is obtained, which is used to analyze the characteristics of voltage value change over time. The variance of the dynamic voltage data sequence is calculated to determine its fluctuation range and obtain the characteristic parameters of voltage fluctuation. Variance is an important indicator reflecting data stability and fluctuation amplitude. By calculating the variance, the variation amplitude of voltage data over a period of time is determined, so as to judge whether the state of the steel cable is within the normal range. The voltage threshold range is set based on the voltage fluctuation characteristic parameters to make a reasonable judgment on the voltage change. The dynamic voltage data sequence is compared with the set voltage threshold range to identify and extract valid sampling data, and invalid data caused by noise and short-term abnormalities are eliminated to ensure that the data used for judgment has a high degree of credibility. Based on the valid sampling data, its voltage mean is calculated, and the voltage mean is used as the characteristic voltage value of the current sampling period, reflecting the overall level of voltage in the entire sampling period. The periodic voltage characteristic value is digitally quantized. The continuous voltage value is converted into discrete values ​​so that the microcontroller can more easily store and calculate these values ​​to obtain the final voltage digital characteristic value.

[0016] 105. According to the voltage digital characteristic value, the voltage division characteristic of the detection circuit is calculated and analyzed to obtain a calculation result, and the calculation result is compared with the preset standard resistance voltage division value to obtain the steel cable state characteristic data; Specifically, the voltage digital characteristic value is calculated for the voltage division coefficient, and based on the known resistance values ​​of the first resistor and the signal acquisition resistor in the detection loop, the theoretical voltage division ratio is calculated by formula derivation to obtain the standard voltage division reference value. The standard voltage division reference value provides a benchmark for subsequent state judgment, representing the ideal voltage division situation in the loop under normal circumstances. According to the standard voltage division reference value, the actual measured voltage digital characteristic value is normalized to obtain the actual voltage division coefficient. The measured actual voltage value is mapped to the standard voltage division range, so that the data is compared and analyzed on a unified scale, eliminating the deviation introduced by the change of measurement conditions. Based on the actual voltage division coefficient, the Ohm's law calculation is performed in combination with the current distribution relationship between the first current branch and the second current branch of the detection loop to obtain the voltage division ratio data of the current detection state. The first current branch is a path through the steel cable resistance, and the second current branch includes the signal acquisition part. The current distribution and corresponding voltage division characteristics of the two reflect whether the state of the steel cable is normal. Through Ohm's law, the current voltage division ratio is calculated according to the relationship between the actual current and the resistance. The voltage division ratio data is calculated by difference with the preset standard resistance voltage division value to obtain the voltage division deviation. The voltage division deviation is an important judgment indicator, which reflects the difference between the current voltage division ratio and the normal state. A large deviation means that the steel cable is cut, replaced or shorted. According to the voltage division deviation, a deviation feature matrix is ​​established, and the deviation is filled into the matrix in sequence according to the sampling time sequence to form a voltage division feature vector. The voltage division feature vector represents the state of the voltage division deviation changing with time during the entire sampling process, which can reflect the dynamic characteristics of voltage change. The voltage division feature vector is subjected to principal component analysis to extract the most representative feature components from the high-dimensional data, eliminate relatively unimportant dimensions, simplify the data structure and retain the most critical information to obtain the feature component data. Based on the feature component data, the discrimination interval is set, and the voltage division feature vector is projected into the discrimination interval for feature space mapping to obtain feature space distribution data, which reflects the specific location and distribution of feature points under different states. Cluster analysis is performed on the feature space distribution data, and different state categories are divided by analyzing the aggregation form of feature points in the feature space. According to the distribution law of feature points in space, different state categories are identified and divided, such as the normal state, cut state or short-circuited state of the steel cable. Through cluster analysis, the state categories are clearly distinguished to obtain the state feature data of the steel cable.

[0017] Perform singular value decomposition on the feature component data, extract the main singular values ​​in the data, and obtain the feature decomposition matrix. Calculate the distribution interval of the eigenvalues ​​according to the eigendecomposition matrix. Determine the distribution range of the data through the maximum and minimum eigenvalues ​​in the eigendecomposition matrix, and obtain the boundary value of the feature space. The feature space boundary value is used to describe the distribution boundary of the data in the feature space. These boundary values ​​can help the system to clarify the normal distribution range of the data and the range of data that may be abnormal. Based on the boundary value of the feature space, construct the coordinate system of the discriminant space, map the boundary value of the feature space to the standardized coordinate axis, and establish a standardized discriminant interval. The standardized discriminant interval is the basis for analyzing and judging the relationship between different feature data. By standardizing the data, the deviation caused by different dimensions is eliminated, so that all data can be compared and analyzed on the same scale, improving the accuracy of feature analysis. Grid division is performed on the standardized discriminant interval, and the size of the grid is determined according to the distribution density of the feature component data in space, so as to construct the feature distribution grid. The feature space is divided more finely so that each feature point can correspond to a specific grid area. By analyzing the distribution density of the feature component data, the size of the grid is determined so that the grid can cover all the feature data and ensure that the density of data points in each grid area is reasonable. The partial pressure feature vector is projected into the feature distribution grid according to the direction of the principal component, and the projection coordinates of each vector in the grid are calculated to obtain the spatial mapping coordinates. Through projection, the main information of the data is effectively concentrated in a low-dimensional space. The spatial mapping coordinates are normalized and the coordinate values ​​are mapped into a unified numerical range to obtain normalized spatial coordinates, which reflect the relative position of the feature points in the feature space. Based on the normalized spatial coordinates, a feature distribution map is constructed, and the Euclidean distance between each coordinate point is calculated to obtain a spatial distance matrix. The Euclidean distance is used to measure the distance between two points in space. By calculating the Euclidean distance between each feature point, the relative distance relationship of the feature point in the feature space is obtained. The spatial distance matrix is ​​divided into regions, and similar feature points are divided into the same region according to the set distance threshold to obtain the distribution data of the feature space, which shows the aggregation of feature points in different states, such as normal state, cut state and short-circuited state. Each state forms a specific aggregation area in the feature space.

[0018] Two different density calculation models are established for the feature space distribution data, namely the first density calculation model and the second density calculation model. Through the first density calculation model, the local density value of the data point in the feature space is obtained to reflect the distribution of a specific data point in its neighborhood; through the second density calculation model, the global density value of the data point is obtained to represent the overall distribution of the data point in the entire feature space. Through these two density calculation models, double-layer density distribution data is obtained. Based on the double-layer density distribution data, the first round of cluster center selection is performed. By calculating the product of the local density value and the global density value, the importance index of each data point is obtained. The N data points with the largest product are selected as the candidate points of the initial cluster center to obtain a set of candidate center points. The candidate points represent the points with the highest density and the most representative in the feature space. The first spatial distance calculation is performed on the set of candidate center points. The candidate points with a spatial distance less than the threshold are eliminated through an adaptive distance threshold to ensure that the selected cluster centers have sufficient separation in space to avoid excessive concentration of multiple cluster centers and poor clustering effect. After this process, the optimized cluster center points are obtained, representing the main density area in the feature space. According to the optimized cluster center points, the first type of mixed Gaussian model is established to calculate the probability density function of each data point in the feature space. By calculating the probability density function, the first type of probability distribution matrix is ​​obtained to describe the probability distribution of the data points in the feature space. In this way, the relationship between the data points and different cluster centers is quantitatively described in the form of probability. The first type of probability distribution matrix is ​​subjected to singular value decomposition to extract the main eigenvectors, and the second type of mixed Gaussian model is constructed based on these eigenvectors to obtain the second type of probability distribution matrix. Through singular value decomposition, the dimension of the data is effectively reduced, while retaining the most important feature information in the data, so that the second type of probability distribution matrix can better describe the data distribution characteristics in the feature space. A dual probability fusion model is constructed based on the first type of probability distribution matrix and the second type of probability distribution matrix. The first and second type of probability distribution matrices are fused by weighted fusion to calculate the final attribution probability of each data point and obtain the probability attribution matrix. According to the probability attribution matrix, the second round of cluster center optimization is performed on the feature space distribution data. The iterative expectation maximization (EM) algorithm is used to update the position of the cluster centers so that these cluster centers better adapt to the actual distribution of the data. The EM algorithm is an optimization algorithm for probabilistic models. It gradually approaches the optimal solution by updating the estimated value of the cluster center in each iteration to obtain the target cluster center. The state of the target cluster center is calibrated, and all data points are assigned to the corresponding cluster center according to the maximum probability principle. Each data point will be assigned to the cluster center with the highest probability of belonging. In this way, the classification of data points is achieved, and the state category corresponding to each cluster center is determined according to the characteristic attributes of different cluster centers, and finally the state characteristic data of the steel cable is obtained.

[0019] 106. Extract the characteristic data of the steel cable status, distinguish the steel cable cut status and the wire short circuit status through the voltage distribution characteristics, and generate an anti-cut alarm trigger signal.

[0020] Specifically, the steel cable state feature data is input into the input layer of the deep convolutional neural network, and the steel cable state feature data is converted into a feature matrix to obtain an input data matrix. In this way, the original steel cable state data is represented as a matrix form that is easy to process and analyze as the input of the neural network. The input data matrix is ​​subjected to feature extraction through the encoding module of the deep convolutional neural network. The encoding module contains three convolution units, each of which consists of two convolution layers and one maximum pooling layer. In each convolution unit, the convolution layer is used to extract local patterns in the feature data, and the spatial features and change trends in the voltage data can be captured through the convolution operation. Each layer of convolution is followed by a ReLU activation function to increase the nonlinear representation ability of the model. At the same time, a batch normalization layer is used to normalize the data to reduce the problem of internal covariate shift. After this series of convolution and normalization processing, a deep feature map is obtained, which contains a multi-level feature representation of the input steel cable state data, which can effectively describe the change trend of the partial voltage deviation. The deep feature map is reconstructed through the decoding module. The decoding module consists of three deconvolution units, each of which contains two deconvolution layers and one upsampling layer. In the decoding process, the deconvolution operation is used to spatially expand the feature map and gradually restore the spatial characteristics of the original data. After each layer of deconvolution, the Leaky-ReLU activation function is used for activation processing to enhance the perception of negative value features and make the model more adaptable to the complex voltage change characteristics. Through deconvolution and upsampling operations, the distribution characteristics of the feature component data are reconstructed to obtain reconstructed feature data that can better restore the steel cable state data. The reconstructed feature data is input into the residual attention module for feature enhancement. The residual attention module consists of a channel attention submodule and a spatial attention submodule. The channel attention submodule processes the temporal change characteristics of the voltage digital feature value and strengthens the attention to important temporal features by assigning different weights. The spatial attention submodule is used to process the spatial distribution characteristics of the voltage divider characteristics and enhance the sensitivity to these features by weighting the features of specific spatial regions. The enhanced feature vector is obtained through the joint action of channel and spatial attention. The enhanced feature vector is input into the state classification layer, and the state classification layer classifies the input features through the Softmax activation function. The Softmax activation function maps the feature vector to a probability distribution, which is used to represent the probability of the detection circuit in three working states, including the normal state based on the standard resistor voltage divider value, the cut state based on the disconnection of the resistor wire, and the short-circuited state based on the replacement of the ordinary wire. In this way, the probability distribution of each state is obtained to determine the current working state of the steel cable. Based on the state classification results, the alarm module is triggered to determine whether an alarm signal needs to be issued.The alarm module determines the level of the anti-shear alarm by monitoring the duration of the shear state or the short-circuit state. If the shear or short-circuit state lasts for more than a period of time, an anti-shear alarm trigger signal of the corresponding level is generated according to different durations.

[0021] The deep feature map is input to the first deconvolution layer of the first deconvolution unit in the decoding module. The feature map is transposed convolution operation is performed using a 5×5 convolution kernel, and the feature map is upsampled to gradually restore the size of the original input in the spatial dimension. In order to enhance the model's ability to learn complex features, the Leaky-ReLU activation function is used to perform nonlinear transformation on the convolution result, so that the model can not only capture significant feature changes, but also perceive small negative features, increase the flexibility of the network, and obtain the first deconvolution feature. The first deconvolution feature is processed and input to the second deconvolution layer of the first deconvolution unit, and a transposed convolution operation is performed using a 3×3 convolution kernel. In order to ensure the stability of feature distribution and the uniformity of data, the features are processed in combination with a batch normalization layer. Batch normalization helps to reduce the difference between feature values, improve the convergence speed and training stability of the network, and obtain the first layer of feature maps. The nearest neighbor interpolation upsampling operation is performed based on the first layer of feature maps to double the size of the feature map, giving it a higher spatial resolution. While expanding the feature map, a jump connection is established with the features of the corresponding layer of the encoding module. The jump connection directly connects the features obtained in the decoding process with the features of the corresponding layer in the encoding process to fuse the feature representations of different stages and obtain the first layer of enhanced features. The first layer of enhanced features is input into the second deconvolution unit of the decoding module. Through the cascade processing of the double-layer deconvolution layer and the upsampling layer, the feature map is gradually expanded in space and reconstructed in detail. Combined with the jump connection, the features in the second deconvolution unit are fused with the corresponding features in the encoding module to maintain the consistency and integrity of the features. Through multi-level deconvolution and jump connection processing, the second layer of enhanced features is obtained. The second layer of enhanced features is input into the first deconvolution layer of the third deconvolution unit for feature mapping operation. In order to enhance the model's capture of global context information, the dilated convolution operation is used. The dilated convolution can effectively increase the receptive field of the convolution without increasing the number of parameters, so that the model can capture multi-scale information at the same time when processing the feature map, and obtain multi-scale feature representation. The multi-scale feature representation is input into the second deconvolution layer of the third deconvolution unit, and the feature map is processed by a 1×1 convolution kernel to compress the number of feature channels. The dimension of the feature map is reduced, the amount of calculation and the amount of parameters are reduced, and the information of different feature channels is integrated to obtain a compact and effective feature fusion result. An upsampling operation is performed based on the feature fusion result to restore the feature map to the size of the original input. The upsampling method is used to gradually restore the spatial size of the feature map so that it is finally consistent with the resolution of the input data, and the preliminary reconstructed features are obtained. In order to enhance the connection between the reconstructed features and the deep feature map, the preliminary reconstructed features are fused with the original deep feature map so that the reconstructed features retain the key information of the original input data and have the deep features learned by the network.

[0022] In the embodiment of the present invention, by setting a resistance wire with a specific resistance value inside the steel cable and using a voltage divider detection circuit to realize monitoring, the system can accurately sense the integrity state of the steel cable, effectively avoiding the limitations of the traditional mechanical detection scheme. The dual-branch detection structure constructed based on transistors and voltage divider networks realizes real-time monitoring of the steel cable state, and at the same time improves the stability and reliability of the detection signal through capacitor filtering and diode protection. By using digital processing and feature extraction technology to analyze the detection signal, the system can not only identify the shearing state of the steel cable, but also effectively distinguish the short-circuit deception behavior of the wire through the voltage distribution characteristics, thereby improving the safety of the anti-shear scheme. The overall scheme has a simple structure, low cost, and is easy to integrate into the existing smart locks, with strong practicality and promotion value. By combining software and hardware, the complexity of the system implementation is reduced while ensuring the accuracy of detection. By using periodic sampling and digital feature analysis, the system has strong anti-interference ability and environmental adaptability, and improves the reliability of the detection scheme in practical applications.

[0023] In a specific embodiment, the process of executing step 104 may specifically include the following steps: Applying a sampling clock signal to the transistor source, sampling the voltage of the transistor source according to the sampling clock signal, and obtaining a sampling trigger signal; Based on the sampling trigger signal, the voltage of the transistor source is converted into digital form through the analog conversion port of the microcontroller to obtain the original voltage sampling data; Perform multi-point moving average filtering on the original voltage sampling data to remove interference fluctuations in the original voltage sampling data and obtain filtered voltage data; A voltage sampling queue is established according to the filtered voltage data, and the newly sampled voltage data is sequentially stored in the voltage sampling queue and the queue length is updated to obtain a dynamic voltage data sequence; Calculate the variance of the dynamic voltage data sequence, determine the fluctuation range of the dynamic voltage data sequence, and obtain voltage fluctuation characteristic parameters; The voltage threshold range is set based on the voltage fluctuation characteristic parameters, and the dynamic voltage data sequence is compared with the voltage threshold range to obtain valid sampling data; Calculate the voltage mean value according to the valid sampling data, take the voltage mean value as the characteristic voltage value of the current sampling period, and obtain the periodic voltage characteristic value; The periodic voltage characteristic value is digitally quantized, and the quantized value is output as the voltage digital characteristic value to obtain the voltage digital characteristic value.

[0024] Specifically, a sampling clock signal is applied to the transistor source to ensure the timing and consistency of voltage sampling. By accurately controlling the sampling time interval, it is ensured that the sampled data sufficiently represents the actual voltage fluctuation. Assuming the sampling period is , every The voltage of the source of the transistor is sampled at a certain time. The sampled voltage value is used as the raw data for the next step of processing to form a stable time series, namely the sampling trigger signal. The sampling trigger signal is a pulse signal used to drive the microcontroller to measure the voltage of the transistor source. Based on the sampling trigger signal, the voltage of the transistor source is converted into an analog-to-digital conversion through the analog conversion port of the microcontroller, and the analog signal is converted into a digital signal for subsequent digital signal processing and calculation. In this process, the analog signal is converted by the analog conversion port to generate the original voltage sampling data. Assuming the voltage value is ,in Indicates The voltage value obtained at the sampling moment. The accuracy of the analog conversion port determines the resolution of the voltage measurement. For example, a 10-bit analog conversion port indicates that the voltage is divided into 1024 levels, realizing the digital representation of the voltage value. The original voltage sampling data is filtered to remove interference signals and extract stable voltage characteristics. The original sampling data is smoothed by using the multi-point moving average filtering method. The formula for multi-point moving average filtering is: ; in, For the The filtering result after sampling is is the size of the filter window, representing the number of sampling points involved in the average. It is before The voltage value of each sampling point. Through this filtering process, the influence of random noise is effectively reduced, and smoother and more stable filtered voltage data is obtained. A voltage sampling queue is established based on the filtered voltage data, and the newly sampled voltage data is sequentially stored in the voltage sampling queue and the queue length is updated in real time to form a dynamic voltage data sequence. Assume that the data sequence in the voltage sampling queue is ,in is the current queue length. The dynamic voltage data sequence is used to describe the voltage changes over a period of time. In order to analyze the stability and fluctuation range of the voltage, the variance of the dynamic voltage data sequence is calculated to determine the fluctuation range of the voltage data over this period of time. The variance calculation formula is: ; in, is the variance of the voltage data series, is the mean value of the voltage data, expressed as: ; The variance value can reflect the degree of fluctuation of voltage data. The larger the variance, the more severe the voltage fluctuation and the worse the stability of the system. Based on the voltage fluctuation characteristic parameters, the voltage threshold range is set to judge and classify the voltage data. The dynamic voltage data sequence is compared with the preset voltage threshold range to extract valid sampled data. If the voltage data is within the threshold range, the data is considered valid and can be used for subsequent average calculations. Based on the valid sampled data, its voltage mean is calculated and used as the characteristic voltage value of the current sampling period. The mean calculation effectively smoothes the short-term fluctuations of the voltage so that the obtained characteristic voltage value can represent the voltage state within the time period. Assume that the valid sampled data is , then the mean voltage is: ; The calculated voltage mean is used as the characteristic voltage value of the current sampling period, which represents the overall voltage level of the transistor source during the sampling period. The periodic voltage characteristic value is digitally quantized to facilitate system storage and further analysis. During the quantization process, a fixed quantization level is used to map the continuous voltage value to a discrete numerical range, so that integers can be used to represent the voltage state. For example, the voltage value is mapped to a range of 0 to 255 for storage and calculation in a microcontroller. The quantized value is the voltage digital characteristic value, which represents the voltage state of the system during this period.

[0025] In a specific embodiment, the process of executing step 105 may specifically include the following steps: Calculate the voltage division coefficient for the voltage digital characteristic value, calculate the theoretical voltage division ratio based on the resistance values ​​of the first resistor of the detection loop and the signal acquisition resistor, and obtain the standard voltage division reference value; According to the standard voltage division reference value, the voltage digital characteristic value is normalized to obtain the actual voltage division coefficient; Based on the actual voltage division coefficient, Ohm's law calculation is performed in combination with the current distribution relationship between the first current branch and the second current branch of the detection loop to obtain the voltage division ratio data of the current detection state; Calculate the difference between the voltage division ratio data and the preset standard resistance voltage division value to obtain the voltage division deviation; A deviation characteristic matrix is ​​established according to the partial pressure deviation, and the partial pressure deviation is filled into the deviation characteristic matrix according to the sampling time sequence to obtain a partial pressure characteristic vector, and a principal component analysis is performed on the partial pressure characteristic vector to extract the main characteristic components of the partial pressure characteristic vector to obtain characteristic component data; The discrimination interval is set based on the characteristic component data, and the partial pressure characteristic vector is projected onto the discrimination interval for feature space mapping to obtain feature space distribution data; Cluster analysis is performed on the feature space distribution data, and the state categories are divided according to the distribution rules of the feature points to obtain the steel cable state feature data.

[0026] Specifically, the theoretical voltage divider ratio in the detection loop is calculated to provide a reference for subsequent normalization processing and state determination. Assume that the first resistor in the detection loop is , the signal acquisition resistor is , according to the resistor voltage division principle, calculate the theoretical voltage division ratio. The theoretical voltage division ratio is used to describe the voltage distribution relationship between two resistors under ideal conditions. The formula is: ; in, is the standard voltage reference value, is the input voltage, and They are the two resistors in the detection circuit. This formula means that when the input voltage Under this condition, after the voltage is divided by two series resistors, it falls to The standard voltage division reference value represents the theoretical voltage distribution under normal conditions. According to this standard voltage division reference value, the voltage digital characteristic value is normalized to obtain the actual voltage division coefficient. Mapping the actual measured voltage characteristic value into a standard range can eliminate the deviation under different sampling conditions. Assume that the voltage digital characteristic value is , the calculation formula of the actual voltage division coefficient is: ; in, is the normalized actual voltage division coefficient. Based on the actual voltage division coefficient, combined with the current distribution relationship between the first current branch and the second current branch of the detection loop, Ohm's law is applied to calculate the detection state. Assume that the current of the first current branch is , the current in the second current branch is According to Ohm's law, the voltage divider ratio data of the current detection state is calculated, and its expression is: ; in, Indicates the voltage division ratio under current conditions, and Represent the voltage drop of the first and second current branches respectively. Through this calculation, the distribution of voltage in the current detection circuit is obtained, so as to judge the working state of the steel cable. The voltage division deviation is obtained by calculating the difference between the voltage division ratio data and the preset standard resistance voltage division value. The voltage division deviation can effectively reflect the degree of deviation between the current detection state and the normal state. Assuming that the voltage division ratio data is The standard voltage divider value is , the voltage division deviation is expressed as: ; in, is the voltage division deviation, which indicates the difference between the actual measured voltage division and the theoretical value. The larger the voltage division deviation, the abnormal state of the steel cable, such as cutting or short circuit. A deviation characteristic matrix is ​​established according to the voltage division deviation, and the voltage division deviation is filled into the deviation characteristic matrix according to the sampling sequence to obtain the voltage division characteristic vector. The characteristic vector is used to describe the voltage change characteristics at different time points, representing the state change of the steel cable within a certain time range. By performing principal component analysis on the voltage division characteristic vector, the main characteristic components are extracted to obtain the characteristic component data. The discrimination interval is set based on the characteristic component data. By setting the discrimination interval, the voltage division characteristic vector is projected into a specific discrimination space to obtain the characteristic space distribution data. Assume that the characteristic vector is , the discrimination interval is a vector space, and the projection process is realized by vector dot product, so as to obtain a specific data distribution in the feature space. Mapping high-dimensional feature data to low-dimensional space can more intuitively observe its distribution characteristics and clustering rules. Cluster analysis is performed on the feature space distribution data, and the state categories are divided according to the distribution law of feature points in the feature space to obtain the state feature data of the steel cable. Cluster analysis is an unsupervised machine learning method that can cluster data points with similar characteristics together to form different categories according to the distance and distribution between data points. For example, by clustering feature points, the state of the steel cable is divided into normal state, cut state and short-circuited state. In the feature space, the feature points of the normal state are usually clustered in a specific area, while the abnormal state (such as cutting or short-circuiting) will form different clustering areas in the space. Through cluster analysis, the working state of the steel cable can be effectively classified.

[0027] In a specific embodiment, the execution step sets a discrimination interval based on the characteristic component data, projects the partial pressure characteristic vector to the discrimination interval for feature space mapping, and obtains the feature space distribution data, which may specifically include the following steps: Perform singular value decomposition on the characteristic component data, extract the main singular values ​​of the characteristic component data, and obtain the characteristic decomposition matrix; Calculate the eigenvalue distribution interval according to the eigendecomposition matrix, determine the data distribution range through the maximum eigenvalue and the minimum eigenvalue, and obtain the boundary value of the feature space; Construct a discriminant space coordinate system based on the feature space boundary value, map the feature space boundary value to the standardized coordinate axis, and obtain a standardized discriminant interval; Gridding is performed on the standardized discrimination interval, and the grid size is determined according to the distribution density of the characteristic component data to obtain the characteristic distribution grid; Project the partial pressure eigenvector to the characteristic distribution grid according to the principal component direction, calculate the projection coordinates of each vector in the grid, and obtain the spatial mapping coordinates; Perform normalization transformation on the spatial mapping coordinates, map the coordinate values ​​into a uniform numerical range, and obtain normalized spatial coordinates; Based on the normalized spatial coordinates, a feature distribution map is constructed, and the Euclidean distance between each coordinate point is calculated to obtain a spatial distance matrix; The spatial distance matrix is ​​divided into regions, and similar data points are divided into the same region according to the distance threshold to obtain feature space distribution data.

[0028] Specifically, singular value decomposition is performed on the characteristic component data. Singular value decomposition is a matrix decomposition tool that can decompose the original data into multiple independent components to extract the most important information in the data. Assume that the characteristic component data matrix is , the singular value decomposition is expressed as: ; in, Is an orthogonal matrix representing the basis vectors of the feature data in the left singular space, is a diagonal matrix containing singular values, which represent the energy distribution in the eigenmatrix. It is the basis vector matrix of the characteristic data in the right singular space. Through singular value decomposition, the most important singular values ​​in the characteristic component data are extracted to obtain the characteristic decomposition matrix, which reflects the main characteristics of the data and helps to reduce the data dimension while retaining the key information of the data. The distribution interval of the eigenvalue is calculated according to the characteristic decomposition matrix to determine the overall distribution range of the data. By calculating the singular value matrix The maximum singular value and the minimum singular value in determine the boundary value of the feature space of the data. Assume that the maximum singular value is , the minimum singular value is , the feature space boundary value is expressed as an interval , describing the distribution boundary of feature data in space. Based on the determined feature space boundary value, a discriminant space coordinate system is constructed, and the feature space boundary value is mapped to the standardized coordinate axis to obtain a standardized discriminant interval, so that data of different feature dimensions can be compared and analyzed on the same scale, eliminating the deviation caused by scale differences. Assume that the boundary value is and , the standardization process is achieved through the following formula: ; in, is the standardized data value, is the original data value, and are the minimum and maximum boundary values ​​of the feature space, respectively. Through the standardization process, the eigenvalues ​​are mapped to a unified range, such as (0, 1). The standardized discrimination interval is gridded, and the entire feature space is divided into several small areas to obtain a feature distribution grid. The size of the grid is determined according to the distribution density of the feature component data. Assuming that the feature data is very dense in a certain area, the corresponding grid division will be more detailed to capture the data characteristics in the area. Divide the feature space into multiple small blocks so that each feature point falls into a specific grid, which is convenient for subsequent analysis and classification. Project the partial pressure eigenvector onto the feature distribution grid according to the direction of the principal component. Assume that the eigenvector is , the principal component direction is p, and the projection process is implemented through the dot product to obtain the projection coordinates of each vector in the grid: ; in, is the projection coordinate, which represents the projection result of the eigenvector in the direction of the principal component. Through the projection process, the high-dimensional eigenvector is reduced to the feature distribution grid, so that its distribution in the feature space can be observed more intuitively. The spatial mapping coordinates obtained by projection are normalized to ensure that these coordinate values ​​can be processed within a unified numerical range. Based on the normalized spatial coordinates, a feature distribution map is constructed to describe the relative position relationship of the feature points in space. In order to quantitatively describe the distance between feature points, the Euclidean distance between each coordinate point is calculated. The Euclidean distance is a common way to measure the straight-line distance between two points in space. Assume that the two coordinate points are and , the Euclidean distance between them is expressed as: ; in, is the Euclidean distance between two coordinate points, and are the coordinates of two feature points respectively. By calculating the Euclidean distance between each feature point, we get the spatial distance matrix, which describes the relative position and distance relationship of the feature points in the feature space. We divide the spatial distance matrix into regions, and divide the similar feature points into the same region according to the set distance threshold, and get the distribution data of the feature space. Assume that the distance threshold is , if the distance between two feature points , they are considered to belong to the same area. The points with similar characteristics are classified into one category by clustering method to achieve effective classification of the cable status. Different feature areas represent different cable states, such as normal state, cut state or short-circuited state.

[0029] In a specific embodiment, the execution step performs cluster analysis on the feature space distribution data, divides the state categories according to the distribution law of the feature points, and obtains the process of the steel cable state feature data, which can specifically include the following steps: A first density calculation model and a second density calculation model are established for the feature space distribution data, local density values ​​of data points are obtained through the first density calculation model, and global density values ​​of data points are obtained through the second density calculation model to obtain double-layer density distribution data; The first round of cluster center selection is performed based on the double-layer density distribution data, and the N data points with the largest product of the local density value and the global density value are used as the initial cluster center candidate points to obtain a set of candidate center points; Perform the first spatial distance calculation on the candidate center point set, and use the adaptive distance threshold to eliminate candidate points whose spatial distance is less than the threshold to obtain the optimized cluster center point; A first-class mixed Gaussian model is established based on the optimized cluster center points, and the probability density function of each data point in the feature space distribution data is calculated to obtain the first-class probability distribution matrix; Perform singular value decomposition on the first type of probability distribution matrix, extract the main eigenvectors to construct the second type of mixed Gaussian model, and obtain the second type of probability distribution matrix; A dual probability fusion model is constructed based on the first-class probability distribution matrix and the second-class probability distribution matrix, and the final attribution probability of the data points is calculated by weighted fusion to obtain the probability attribution matrix. The second round of cluster center optimization is performed on the feature space distribution data according to the probability attribution matrix, and the cluster center position is updated through the iterative expectation maximization algorithm to obtain the target cluster center; The target cluster center is calibrated in state, and the data points are assigned to the corresponding cluster center according to the maximum probability principle. The corresponding state category is determined according to the characteristic attributes of each cluster center to obtain the steel cable state characteristic data.

[0030] Specifically, a first density calculation model and a second density calculation model are established. The first density calculation model is used to obtain the local density value of each data point, which reflects the concentration of the data point in its vicinity; the second density calculation model is used to obtain the global density value of the data point, which describes the distribution characteristics of the data point in the entire feature space. Suppose there is a set of data points , Indicates The location of the data point in the feature space. The local density value is calculated by the Gaussian kernel function, which is defined as: ; in, Indicates The local density value of the data point, It is a data point and The Euclidean distance between is the bandwidth parameter of the kernel function, which is used to control the range of the neighborhood. The higher the local density value, the denser the data points around the point. The global density value is calculated by a distance inverse weighted model, defined as: ; in, Indicates The global density value of the data point. When the global density value is large, it means that the average distance between the point and other data points is small, indicating that its distribution in the entire space is more concentrated. Through the above two density calculation models, the double-layer density distribution data of the data points is obtained, where the product of the local density value and the global density value is used to select the initial cluster center point. The point with the largest product of the local density value and the global density value is selected. The data points are used as the candidate point set for the initial cluster center, which is defined as: C candidate ={ x i ∣ ρ i ⋅ δ i ] ; In this way, the selected candidate points are ensured to have both high local density and good global distribution characteristics. The spatial distance of the candidate center point set is calculated to ensure that the final cluster centers have sufficient separation. The candidate points whose distances are less than the set threshold are eliminated through the adaptive distance threshold, and the points that are far away are retained to reduce the redundancy between cluster centers. Assume that the distance threshold is , then if the distance between two candidate points , then one of the candidate points will be removed. In this way, the optimized cluster center point set is obtained. The first type of mixed Gaussian model is established based on the optimized cluster center points. The mixed Gaussian model is a probability model used to describe the probability distribution of data in the feature space. By calculating the probability density function of each data point in the feature space distribution data, the first type of probability distribution matrix is ​​obtained. For each data point , its probability density function under the mixed high-period model is expressed as: ; in, is the number of Gaussian distributions, For the Gaussian distribution weights satisfy , and Respectively The mean and covariance matrix of a Gaussian distribution, Indicates Gaussian distribution at the data point The probability density value at . Perform singular value decomposition on the first probability distribution matrix to extract the main eigenvectors, and build the second type of mixed Gaussian model based on the main eigenvectors to obtain the second type of probability distribution matrix. Through singular value decomposition, redundant information is removed and the main components in the data are retained to build a more streamlined probability model. Based on the first and second type probability distribution matrices, a dual probability fusion model is constructed. The dual probability fusion model calculates the final attribution probability of each data point by weighted fusion of the two types of probability distributions. Assume that the first type of probability distribution is , the second probability distribution is , the fused belonging probability is expressed as: ; in, is a weighting coefficient used to balance the contribution of the two types of probability distributions. Through weighted fusion, the advantages of different models are comprehensively considered to more accurately determine the attribution of each data point. The second round of cluster center optimization is performed on the feature space distribution data according to the probability attribution matrix. Through the iterative expectation maximization (EM) algorithm, the position of the cluster center is continuously updated so that the attribution probability of the data point gradually converges to obtain the target cluster center. The "expectation" step of the EM algorithm calculates the probability that each data point belongs to each cluster, while the "maximization" step updates the position of the cluster center to maximize the attribution probability of the data point. The state of the target cluster center is calibrated. All data points are assigned to the corresponding cluster center according to the maximum probability principle. Suppose a data point is in the target cluster center. The probability of belonging to the data point is the largest, then the data point is classified as According to the characteristic attributes of each cluster center, the state category corresponding to each cluster center is determined, such as normal state, cut state or short-circuited state, and the state characteristic data of the steel cable is obtained.

[0031] In a specific embodiment, the process of executing step 106 may specifically include the following steps: Inputting the steel cable state feature data into the input layer of the deep convolutional neural network, converting the steel cable state feature data into a feature matrix to obtain an input data matrix; The input data matrix is ​​subjected to feature extraction through the encoding module of the deep convolutional neural network. The encoding module contains three convolution units, each of which includes two convolution layers and one maximum pooling layer. The convolution layer uses the ReLU activation function and the batch normalization layer to extract the feature of the change trend of the partial pressure deviation to obtain a deep feature map. The deep feature map is reconstructed through the decoding module of the deep convolutional neural network. The decoding module contains three deconvolution units, each of which includes two deconvolution layers and one upsampling layer. The deconvolution layer uses the Leaky-ReLU activation function to reconstruct the distribution characteristics of the feature component data to obtain the reconstructed feature data. The reconstructed feature data is input into the residual attention module for feature enhancement. The channel attention submodule in the residual attention module processes the temporal variation characteristics of the voltage digital feature value, and the spatial attention submodule processes the spatial distribution characteristics of the voltage division characteristics to obtain an enhanced feature vector. The enhanced feature vector is input into the state classification layer. The state classification layer calculates the probability distribution of the three working states of the detection circuit through the Softmax activation function: the normal state based on the standard resistor voltage divider value, the cut state based on the disconnection of the resistor wire, and the short-circuit state based on the replacement of the ordinary wire, and obtains the state classification result; The alarm module is triggered based on the state classification result. The alarm module determines the anti-shear alarm level according to the duration of the shear state or the short-circuit state, and generates an anti-shear alarm trigger signal of the corresponding level.

[0032] Specifically, the steel cable state feature data is input into the input layer of the deep convolutional neural network, and the data is converted into a feature matrix that can be processed by the neural network to obtain an input data matrix. Assume that the input steel cable state feature data is , including data of multiple dimensions such as voltage division deviation, voltage value, time, etc., through preprocessing, the features are constructed into a shape of The matrix form of represents the length of the time series, Represents the number of different features at each time point. After the input data matrix passes through the input layer of the deep convolutional neural network, it enters the encoding module for feature extraction. In the encoding module, the deep convolutional neural network uses three layers of convolution units, each of which consists of two convolution layers and one maximum pooling layer. The convolution layer extracts the spatial and temporal features in the input data, especially the changing trend of the voltage deviation in the cable state feature. Assume that the convolution kernel is , the input feature map is , the output of the convolutional layer is calculated by the formula: ; in, is the output feature map after convolution, is the convolution kernel, * indicates the convolution operation, is the bias term. After each convolution operation, the ReLU activation function is used for nonlinear transformation. The expression of the ReLU activation function is: ; Among them, the purpose of ReLU is to increase the nonlinearity of the model and improve the ability to capture complex patterns. The batch normalization layer is used to standardize the output of the convolution, reduce the statistical differences between different samples, accelerate the training process and improve the stability of the model. The convolution is followed by the maximum pooling layer to reduce the size of the feature map, reduce the computational complexity, and extract significant features in the data. After being processed by the three-layer convolution unit of the encoding module, the input data is extracted as a deep feature map. The deep feature map is input into the decoding module of the deep convolutional neural network for feature reconstruction. The decoding module contains three deconvolution units, each of which includes two deconvolution layers and one upsampling layer. The purpose of the deconvolution layer is to gradually amplify the convolutional feature map in order to restore the spatial characteristics of the input data. Assume that the input feature map is , the deconvolution process is expressed as: ; in, is the deconvolution kernel, is the bias term, It is the output feature map after deconvolution. The deconvolution layer uses the Leaky-ReLU activation function for nonlinear transformation. The expression of Leaky-ReLU is: ; in, is a small positive number, which is used to retain a certain gradient when the input is negative to avoid the phenomenon of neuron "death". Through the reconstruction of the deconvolution layer and the processing of the upsampling layer, the size of the feature map is gradually restored to obtain the reconstructed feature data. In order to improve the expressiveness of the feature, the reconstructed feature data is input into the residual attention module for feature enhancement. The residual attention module weights the features through the attention mechanism to enhance the representation ability of key features. The residual attention module includes a channel attention submodule and a spatial attention submodule. The channel attention submodule is used to process the time series change characteristics of the voltage digital feature value, and enhances the influence of the important part of the time series feature by assigning different weights to each channel. The spatial attention submodule processes the spatial distribution characteristics of the voltage division characteristics, and highlights the significant features at a specific position by weighting the features in the spatial dimension. Through the processing of the residual attention module, an enhanced feature vector is obtained. The enhanced feature vector is input into the state classification layer. The function of the state classification layer is to classify the working state of the steel cable, and the probability distribution of each state is calculated using the Softmax activation function. Assuming that the classification results are three states: normal state, cut state and short-circuited state, the expression of the Softmax function is: in, Represents input features Belong to The probability of a state, It is The linear combination result corresponding to the states is is the number of status categories, , representing three possible states. The role of the Softmax activation function is to convert the linear output into a probability distribution to ensure that the sum of the probabilities of all states is 1. Based on the state classification results, enter the alarm module. The alarm module is used to determine whether the steel cable is in an abnormal state based on the classification results and trigger the corresponding alarm mechanism. If the classification results show that the steel cable is in a cut state or a short circuit state, the alarm module analyzes the duration of these abnormal states to determine the level of the anti-cut alarm. Assume that the probability of the cut state is , the probability of the short circuit state is If the values ​​of these probabilities exceed the set threshold , and duration Exceeding the safety time threshold , different levels of alarm signals will be triggered.

[0033] In a specific embodiment, the execution step reconstructs the features of the deep feature map through a decoding module of a deep convolutional neural network, the decoding module includes three deconvolution units, each deconvolution unit includes two deconvolution layers and one upsampling layer, the deconvolution layer uses a Leaky-ReLU activation function to reconstruct the distribution characteristics of the feature component data, and the process of obtaining the reconstructed feature data can specifically include the following steps: The deep feature map is input into the first deconvolution layer of the first deconvolution unit, the feature map is transposed convolution operation is performed through a 5×5 convolution kernel, and the Leaky-ReLU activation function is used for nonlinear transformation to obtain the first deconvolution feature; The first deconvolution feature is subjected to a 3×3 convolution kernel transposition convolution operation through the second deconvolution layer of the first deconvolution unit, and the feature distribution is processed in combination with the batch normalization layer to obtain the first layer feature map; Based on the first-layer feature map, the nearest neighbor interpolation upsampling is performed to double the size of the feature map, and a jump connection is established with the features of the corresponding layer of the encoding module to obtain the first-layer enhanced features; The first layer enhanced features are input into the second deconvolution unit, and the second layer enhanced features are obtained through the cascade processing of the double-layer deconvolution layer and the upsampling layer and combined with the second skip connection; The second layer enhanced features are mapped through the first deconvolution layer of the third deconvolution unit, and the receptive field is enlarged using dilated convolution to obtain multi-scale feature representation; The multi-scale feature representation is input into the second deconvolution layer of the third deconvolution unit, and the number of feature channels is compressed by a 1×1 convolution kernel to obtain the feature fusion result; Based on the feature fusion result, a final upsampling operation is performed to restore the feature map to the original input size to obtain preliminary reconstructed features, and the preliminary reconstructed features are fused with the deep feature map to obtain reconstructed feature data.

[0034] Specifically, the deep feature map is input to the first deconvolution unit in the decoding module for processing. In this unit, the deep feature map passes through the first deconvolution layer and then passes through a The main purpose of transposed convolution is to upsample the feature map to restore it to the original input spatial dimension. Assume that the input feature map is , the convolution kernel is , the calculation formula of transposed convolution is expressed as: ; in, represents the output feature map after transposed convolution, * represents the convolution operation, is a bias term. In order to enhance the nonlinear expression capability, the Leaky-ReLU activation function is used to perform nonlinear transformation on the convolution feature map. The activation function expression of Leaky-ReLU is: ; in, is a small positive number used to maintain a certain gradient when the input is negative to avoid the "death" of neurons. After the deconvolution operation, the first deconvolution feature is obtained. The first deconvolution feature is transposed and convolved through the second deconvolution layer of the first deconvolution unit. The convolution kernel is used for calculation, and the batch normalization layer is used to process the distribution of features. The purpose of batch normalization is to standardize the feature map after convolution, thereby reducing the distribution difference between different samples and improving the training efficiency and stability of the model. Assume that the output feature after batch normalization is , and get the first layer feature map. Based on the first layer feature map, perform the nearest neighbor interpolation upsampling operation. The nearest neighbor interpolation is an upsampling method that doubles the spatial size of the feature map so as to gradually restore it to the resolution of the original input. Assume that the size of the input feature map is , after upsampling, the size of the feature map will become . In order to enhance the expressive power of features during upsampling, the feature map of this layer is jump-connected to the features of the corresponding layer of the encoding module. Jump connection is a feature fusion method that combines the features gradually recovered during the decoding process with the original features during the encoding process, retaining more detail information and obtaining the first layer of enhanced features. The first layer of enhanced features is input into the second deconvolution unit in the decoding module. This unit contains a double-layer deconvolution layer and an upsampling layer. The feature map is processed through two deconvolution operations to gradually restore the spatial size and detail information of the features. Assume that the output after deconvolution is , which is calculated in a similar way to the previous one, by using convolution kernels of different sizes for transposed convolution operations, and then performing nonlinear transformations through the Leaky-ReLU activation function. After deconvolution, another upsampling operation is performed to further restore the spatial size of the feature map, and combined with the second jump connection in the encoding module to obtain the second layer of enhanced features. The second layer of enhanced features is input into the first deconvolution layer of the third deconvolution unit for feature mapping. In order to increase the receptive field of the model, dilated convolution is used. Dilated convolution introduces "holes" between the convolution kernels to expand the receptive field of the convolution without increasing the number of parameters, so that the model can pay attention to both details and global features. Assume that the convolution kernel is , the void factor is , the output of the dilated convolution is expressed as: ; in, represents the second layer enhanced features, Represents the output feature map after the dilated convolution. Through the dilated convolution, a multi-scale feature representation is obtained, which contains the global information and local details of the input data. The multi-scale feature representation is input to the second deconvolution layer of the third deconvolution unit, and the The convolution kernel processes the feature map and compresses the number of feature channels. The convolution kernel effectively integrates the information of each channel while reducing the amount of calculation and parameters to obtain the feature fusion result. Assume that the number of channels of the input feature is , the number of channels of the output feature is , the feature fusion result is expressed as: ; in, for The weight of the convolution kernel, is the fused feature result. In this way, the model can effectively integrate the features of different channels and represent them as a more compact feature map. An upsampling operation is performed based on the feature fusion result to restore the spatial size of the feature map to the size of the original input to obtain preliminary reconstructed features. The nearest neighbor interpolation method is used to ensure that the final feature map is consistent with the spatial resolution of the input. The preliminary reconstructed features are fused with the original deep feature map to retain more detailed information and high-level feature expressions, and finally the reconstructed feature data is obtained.

[0035] The above describes the anti-shear alarm monitoring method of the steel cable in the embodiment of the present invention. The following describes the anti-shear alarm monitoring system of the steel cable in the embodiment of the present invention. Figure 3 In one embodiment of the present invention, an anti-shear alarm monitoring system for a steel cable includes: The processing module 301 is used to perform resistance wire laying processing on the steel cable, connect one end of the resistance wire to the system grounding terminal, and obtain a steel cable structure including an anti-shear monitoring circuit; A construction module 302 is used to construct a voltage-dividing detection circuit according to the steel cable structure including the anti-shear monitoring circuit, connect the drain of the transistor to the power supply terminal through the first resistor, and connect the source of the transistor to the analog conversion port of the microcontroller through the signal collection resistor to obtain a voltage collection circuit; The control module 303 is used to output a control signal through the control terminal of the microcontroller based on the voltage acquisition circuit to turn on the transistor, and form a first current branch and a second current branch in the steel cable structure containing the anti-shear monitoring circuit to obtain a detection loop; The conversion module 304 is used to periodically sample the source voltage value of the transistor in the detection loop, and perform digital-to-analog conversion on the sampled voltage value through the analog conversion port of the microcontroller to obtain a voltage digital characteristic value; The calculation module 305 is used to calculate and analyze the voltage division characteristics of the detection circuit according to the voltage digital characteristic value, obtain the calculation result, and compare the calculation result with the preset standard resistance voltage division value to obtain the steel cable state characteristic data; The generation module 306 is used to extract the characteristics of the steel cable status characteristic data, distinguish the steel cable cut state and the wire short circuit state through the voltage distribution characteristics, and generate an anti-cut alarm trigger signal.

[0036] Through the cooperation of the above components, by setting a resistor wire with a specific resistance value inside the steel cable and using a voltage divider detection circuit to achieve monitoring, the system can accurately sense the integrity of the steel cable, effectively avoiding the limitations of traditional mechanical detection schemes. The dual-branch detection structure based on transistors and voltage divider networks realizes real-time monitoring of the steel cable status, while improving the stability and reliability of the detection signal through capacitor filtering and diode protection. By using digital processing and feature extraction technology to analyze the detection signal, the system can not only identify the state of the steel cable cut, but also effectively distinguish the wire short-circuit deception behavior through voltage distribution characteristics, thereby improving the safety of the anti-cutting scheme. The overall scheme has a simple structure, low cost, and is easy to integrate into existing smart locks, with strong practicality and promotion value. By combining software and hardware, the complexity of system implementation is reduced while ensuring detection accuracy. The use of periodic sampling and digital feature analysis enables the system to have strong anti-interference ability and environmental adaptability, improving the reliability of the detection scheme in practical applications.

[0037] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0038] If the integrated unit is implemented in the form of a software functional unit and 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 the present invention is essentially or the part that contributes to the prior art or the whole 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, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0039] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring the anti-shear alarm of a steel cable, characterized in that: The method comprises: The steel cable is subjected to resistance wire laying processing, and one end of the resistance wire is connected to the system grounding terminal to obtain a steel cable structure including an anti-shear monitoring circuit; A voltage-dividing detection circuit is constructed according to the steel cable structure containing the anti-shear monitoring circuit, the drain of the transistor is connected to the power supply end through the first resistor, and the source of the transistor is connected to the analog conversion port of the microcontroller through the signal collection resistor to obtain a voltage collection circuit; Based on the voltage acquisition circuit, a control signal is outputted through the control terminal of the microcontroller to turn on the transistor, and a first current branch and a second current branch are formed in the steel cable structure containing the anti-shear monitoring circuit to obtain a detection loop; Periodically sampling the source voltage value of the transistor in the detection loop, performing digital-to-analog conversion on the sampled voltage value through the analog conversion port of the microcontroller to obtain a voltage digital characteristic value; According to the voltage digital characteristic value, the voltage division characteristic of the detection circuit is calculated and analyzed to obtain a calculation result, and the calculation result is compared with a preset standard resistance voltage division value to obtain the steel cable state characteristic data; Feature extraction is performed on the steel cable status feature data, and the steel cable cut state and the wire short circuit state are distinguished through voltage distribution characteristics, and an anti-cut alarm trigger signal is generated.

2. The anti-shear alarm monitoring method of the steel cable according to claim 1 is characterized in that: The periodic sampling of the transistor source voltage value in the detection loop and the digital-to-analog conversion of the sampled voltage value through the analog conversion port of the microcontroller to obtain the voltage digital characteristic value include: Applying a sampling clock signal to the source of the transistor, sampling the voltage of the source of the transistor according to the sampling clock signal, and obtaining a sampling trigger signal; Based on the sampling trigger signal, performing analog-to-digital conversion on the voltage of the transistor source through the analog conversion port of the microcontroller to obtain original voltage sampling data; Performing multi-point moving average filtering on the original voltage sampling data to remove interference fluctuations in the original voltage sampling data to obtain filtered voltage data; Establishing a voltage sampling queue according to the filtered voltage data, storing the newly sampled voltage data in the voltage sampling queue in sequence and updating the queue length to obtain a dynamic voltage data sequence; Performing variance calculation on the dynamic voltage data sequence to determine the fluctuation range of the dynamic voltage data sequence and obtain voltage fluctuation characteristic parameters; Setting a voltage threshold range based on the voltage fluctuation characteristic parameter, and comparing the dynamic voltage data sequence with the voltage threshold range to obtain valid sampling data; Calculate a voltage mean value according to the valid sampling data, use the voltage mean value as a characteristic voltage value of the current sampling period, and obtain a periodic voltage characteristic value; The periodic voltage characteristic value is digitally quantized, and the quantized value is output as the voltage digital characteristic value to obtain the voltage digital characteristic value.

3. The anti-shear alarm monitoring method of the steel cable according to claim 2 is characterized in that: The voltage digital characteristic value is used to calculate and analyze the voltage division characteristic of the detection circuit to obtain a calculation result, and the calculation result is compared with a preset standard resistance voltage division value to obtain the steel cable state characteristic data, including: Calculating a voltage division coefficient for the voltage digital characteristic value, calculating a theoretical voltage division ratio based on the resistance values ​​of the first resistor of the detection loop and the signal acquisition resistor, and obtaining a standard voltage division reference value; According to the standard voltage division reference value, the voltage digital characteristic value is normalized to obtain an actual voltage division coefficient; Based on the actual voltage division coefficient, Ohm's law calculation is performed in combination with the current distribution relationship between the first current branch and the second current branch of the detection loop to obtain voltage division ratio data of the current detection state; Calculating the difference between the voltage division ratio data and the preset standard resistance voltage division value to obtain a voltage division deviation; Establishing a deviation characteristic matrix according to the partial pressure deviation, filling the partial pressure deviation into the deviation characteristic matrix according to the sampling time sequence to obtain a partial pressure characteristic vector, and performing principal component analysis on the partial pressure characteristic vector to extract the main characteristic components of the partial pressure characteristic vector to obtain characteristic component data; Setting a discrimination interval based on the characteristic component data, projecting the partial pressure characteristic vector to the discrimination interval for feature space mapping, and obtaining feature space distribution data; Cluster analysis is performed on the characteristic spatial distribution data, and state categories are divided according to the distribution rules of characteristic points to obtain steel cable state characteristic data.

4. The anti-shear alarm monitoring method of the steel cable according to claim 3 is characterized in that: The step of setting a discrimination interval based on the characteristic component data, projecting the partial pressure characteristic vector to the discrimination interval for feature space mapping, and obtaining feature space distribution data comprises: Performing singular value decomposition on the characteristic component data, extracting main singular values ​​of the characteristic component data, and obtaining an eigendecomposition matrix; Calculate the eigenvalue distribution interval according to the eigendecomposition matrix, determine the data distribution range through the maximum eigenvalue and the minimum eigenvalue, and obtain the boundary value of the feature space; Constructing a discriminant space coordinate system based on the feature space boundary value, mapping the feature space boundary value to a standardized coordinate axis, and obtaining a standardized discriminant interval; Gridding the standardized discrimination interval, determining the grid size according to the distribution density of the characteristic component data, and obtaining a characteristic distribution grid; Projecting the partial pressure characteristic vector to the characteristic distribution grid according to the principal component direction, calculating the projection coordinates of each vector in the grid, and obtaining the spatial mapping coordinates; Performing a normalization transformation on the spatial mapping coordinates, mapping the coordinate values ​​into a uniform numerical interval, and obtaining normalized spatial coordinates; Constructing a feature distribution map based on the normalized spatial coordinates, calculating the Euclidean distance between each coordinate point, and obtaining a spatial distance matrix; The spatial distance matrix is ​​divided into regions, and similar data points are divided into the same region according to a distance threshold to obtain the feature space distribution data.

5. The anti-shear alarm monitoring method of the steel cable according to claim 4 is characterized in that: The cluster analysis is performed on the characteristic spatial distribution data, and the state categories are divided according to the distribution law of the characteristic points to obtain the steel cable state characteristic data, including: Establishing a first density calculation model and a second density calculation model for the feature space distribution data, obtaining local density values ​​of data points through the first density calculation model, and obtaining global density values ​​of data points through the second density calculation model, to obtain double-layer density distribution data; Based on the double-layer density distribution data, a first round of cluster center selection is performed, and N data points with the largest product of local density value and global density value are used as initial cluster center candidate points to obtain a set of candidate center points; Performing a first spatial distance calculation on the candidate center point set, eliminating candidate points whose spatial distance is less than the threshold by using an adaptive distance threshold, and obtaining an optimized cluster center point; A first-class mixed Gaussian model is established according to the optimized cluster center points, and a probability density function of each data point in the feature space distribution data is calculated to obtain a first-class probability distribution matrix; Performing singular value decomposition on the first probability distribution matrix, extracting main eigenvectors to construct a second Gaussian mixture model, and obtaining a second probability distribution matrix; A dual probability fusion model is constructed based on the first probability distribution matrix and the second probability distribution matrix, and the final attribution probability of the data points is calculated by weighted fusion to obtain a probability attribution matrix; Performing a second round of cluster center optimization on the feature space distribution data according to the probability attribution matrix, updating the cluster center position by an iterative expectation maximization algorithm, and obtaining a target cluster center; The target cluster center is calibrated in state, the data points are assigned to the corresponding cluster center according to the maximum probability principle, and the corresponding state category is determined according to the characteristic attributes of each cluster center to obtain the steel cable state characteristic data.

6. The anti-shear alarm monitoring method of the steel cable according to claim 5, characterized in that: The feature extraction of the steel cable state feature data, distinguishing the steel cable cut state and the wire short circuit state by voltage distribution features, and generating an anti-cut alarm trigger signal includes: Inputting the steel cable state feature data into the input layer of a deep convolutional neural network, converting the steel cable state feature data into a feature matrix to obtain an input data matrix; The input data matrix is ​​subjected to feature extraction through the encoding module of the deep convolutional neural network, wherein the encoding module comprises three convolution units, each of which comprises two convolution layers and one maximum pooling layer, and the convolution layer uses a ReLU activation function and a batch normalization layer to extract features of the change trend of the voltage divider deviation to obtain a deep feature map; The deep feature map is subjected to feature reconstruction by a decoding module of the deep convolutional neural network, wherein the decoding module comprises three deconvolution units, each of which comprises two deconvolution layers and one upsampling layer, and the deconvolution layer uses a Leaky-ReLU activation function to reconstruct the distribution characteristics of the feature component data to obtain reconstructed feature data; The reconstructed feature data is input into the residual attention module for feature enhancement, wherein the channel attention submodule in the residual attention module processes the temporal variation characteristics of the voltage digital feature value, and the spatial attention submodule processes the spatial distribution characteristics of the voltage division characteristics, to obtain an enhanced feature vector; The enhanced feature vector is input into the state classification layer, and the state classification layer calculates the probability distribution of three working states of the detection circuit through the Softmax activation function: the normal state based on the standard resistance voltage divider value, the cut state based on the disconnection of the resistance wire, and the short-circuited state based on the replacement of the ordinary wire, and obtains the state classification result; An alarm module is triggered based on the state classification result. The alarm module determines the anti-shear alarm level according to the duration of the shear state or the short-circuit state, and generates an anti-shear alarm trigger signal of a corresponding level.

7. The anti-shear alarm monitoring method of the steel cable according to claim 6, characterized in that: The deep feature map is subjected to feature reconstruction through a decoding module of the deep convolutional neural network, wherein the decoding module comprises three deconvolution units, each of which comprises two deconvolution layers and one upsampling layer, and the deconvolution layer uses a Leaky-ReLU activation function to reconstruct the distribution characteristics of the feature component data to obtain reconstructed feature data, including: Inputting the deep feature map into the first deconvolution layer of the first deconvolution unit, performing a transposed convolution operation on the feature map through a 5×5 convolution kernel, and performing a nonlinear transformation using a Leaky-ReLU activation function to obtain a first deconvolution feature; Performing a 3×3 convolution kernel transposition convolution operation on the first deconvolution feature through the second deconvolution layer of the first deconvolution unit, and processing the feature distribution in combination with a batch normalization layer to obtain a first layer feature map; Perform nearest neighbor interpolation upsampling based on the first layer feature map, double the size of the feature map, and establish a jump connection with the features of the corresponding layer of the encoding module to obtain the first layer enhanced features; Inputting the first layer of enhanced features into the second deconvolution unit, through the cascade processing of the double-layer deconvolution layer and the upsampling layer, and combining with the second skip connection, the second layer of enhanced features is obtained; The second layer of enhanced features is subjected to feature mapping through the first deconvolution layer of the third deconvolution unit, and the receptive field is enlarged using a dilated convolution to obtain a multi-scale feature representation; Input the multi-scale feature representation into the second deconvolution layer of the third deconvolution unit, compress the number of feature channels through a 1×1 convolution kernel, and obtain a feature fusion result; A final upsampling operation is performed based on the feature fusion result to restore the feature map to the original input size to obtain preliminary reconstructed features, and the preliminary reconstructed features are fused with the deep feature map to obtain reconstructed feature data.

8. A steel cable anti-shear alarm monitoring system, characterized in that: The system is used to implement the anti-shear alarm monitoring method of the steel cable according to any one of claims 1 to 7, the system comprising: A processing module, used for performing resistance wire laying processing on the steel cable, connecting one end of the resistance wire to the system grounding terminal, and obtaining a steel cable structure including an anti-shear monitoring circuit; A construction module is used to construct a voltage-dividing detection circuit according to the steel cable structure containing the anti-shear monitoring circuit, connect the drain of the transistor to the power supply terminal through a first resistor, and connect the source of the transistor to the analog conversion port of the microcontroller through a signal collection resistor to obtain a voltage collection circuit; A control module, which is used to output a control signal through the control terminal of the microcontroller based on the voltage acquisition circuit to turn on the transistor, and to form a first current branch and a second current branch in the steel cable structure containing the anti-shear monitoring circuit to obtain a detection loop; A conversion module, used for periodically sampling the source voltage value of the transistor in the detection loop, performing digital-to-analog conversion on the sampled voltage value through the analog conversion port of the microcontroller to obtain a voltage digital characteristic value; A calculation module, used to calculate and analyze the voltage division characteristics of the detection circuit according to the voltage digital characteristic value, obtain a calculation result, and compare the calculation result with a preset standard resistance voltage division value to obtain the steel cable state characteristic data; The generating module is used to extract the characteristics of the steel cable status characteristic data, distinguish the steel cable cut state and the wire short circuit state through the voltage distribution characteristics, and generate an anti-cut alarm trigger signal.

Images