An image recognition-based interface switch machine status monitoring system and method

Abstract

This invention discloses an interface switch machine status monitoring system and method based on image recognition, relating to the field of computer vision technology. The system consists of several functional modules, including: a data acquisition module, which acquires interface switch machine status and weather data; the interface switch machine status includes extension / retraction time, switch machine image, and laser vibration spectrum; the weather data includes temperature, humidity, and particulate matter concentration; a feature extraction module, which receives the interface switch machine status and weather data, performs data fusion, and generates a first feature using feature engineering; the first feature includes interface switch machine extension / retraction features, interface switch machine vibration features, and corrosion image features; and a data analysis and anomaly detection module, which extracts the interface switch machine extension / retraction features and interface switch machine vibration features, calculates the locking notch deviation and vibration amplitude, and compares them with notch thresholds and vibration thresholds, respectively; continuous monitoring occurs when the locking notch deviation is less than the notch threshold and the vibration amplitude is less than the vibration threshold.

Description

Technical Field

[0001] This invention relates to the field of computer vision technology, specifically to an interface switch machine status monitoring system and method based on image recognition. Background Technology

[0002] As the core execution equipment in the railway signaling system, the switch machine is responsible for controlling track switching, ensuring that trains can safely and accurately complete track changes. The accuracy of track switching is one of the important indicators for measuring railway operational efficiency, and the switch machine is key to ensuring this precision. Through mechanical operation, it moves the turnout from one position to another, guiding the train onto the correct track. The accuracy of track switching also affects the smoothness and safety of high-speed trains. For high-speed trains, even a slight track deviation can cause a huge impact, damaging not only the vehicle itself but also potentially threatening the lives of passengers.

[0003] Traditional interface switch machine status maintenance relies on manual periodic inspections and threshold alarms, which suffers from problems such as detection lag and strong subjectivity. When inexperienced technicians conduct regular inspections of the switch machine, they may fail to accurately identify minor signs of poor electrical contact. As the usage time increases, these poor contacts gradually worsen, eventually leading to signal transmission interruption and affecting the normal operation of the train dispatching system. At the same time, interface switch machines are exposed to complex environments for a long time, and the transmission components are susceptible to moisture corrosion, which can cause metal corrosion and lead to mechanical jamming, poor electrical contact, and other faults. If these faults are not identified in time, they can pose safety hazards to the use of the switch machine. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] To address the shortcomings of existing technologies, this invention provides an image recognition-based interface switch machine status monitoring system and method. This system aims to solve at least the problems in existing technologies where interface switch machine status maintenance relies on manual periodic inspections and threshold alarms, resulting in detection delays. Furthermore, the transmission components of interface switch machines are susceptible to moisture corrosion, leading to metal corrosion, mechanical jamming, poor electrical contact, and other faults. Failure to identify these faults in a timely manner poses safety hazards to the use of the switch machine.

[0006] (II) Technical Solution

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] An image recognition-based interface switch machine status monitoring system includes:

[0009] The data acquisition module acquires the status of the interface switch machine and weather data. The status of the interface switch machine includes extension and retraction time, switch machine image, and laser vibration spectrum. The weather data includes temperature, humidity, and particulate matter concentration.

[0010] The feature extraction module receives the interface switch machine status and weather data, performs data fusion, and generates the first feature using feature engineering; the first feature includes the interface switch machine extension and retraction features, the interface switch machine vibration features, and corrosion image features;

[0011] The data analysis and anomaly detection module extracts the extension and vibration characteristics of the interface switch machine, calculates the locking notch deviation and vibration amplitude, and compares them with the notch threshold and vibration threshold, respectively.

[0012] When the locking gap deviation is less than the gap threshold and the vibration amplitude is less than the vibration threshold, continuous monitoring is performed, and the difference score and corrosion evaluation index are obtained to calculate the switch machine condition score.

[0013] The prediction and optimization module issues multi-level warnings and intervenes based on the switch machine status score.

[0014] Furthermore, the process of obtaining the scaling characteristics of the interface switch machine is as follows:

[0015] The scaling time and switch machine images are obtained, and the scaling features of the interface switch machine are extracted using CNN. The scaling features of the interface switch machine include: switch machine damage images and switch machine displacement images.

[0016] CNN extracts interface switch machine scaling features:

[0017] By preprocessing the switch machine images, a multi-task network is designed using segmentation and regression algorithms to extract the switch machine damage images and switch machine displacement images, train the configuration, and verify the deployment.

[0018] Furthermore, the vibration characteristics of the interface switch machine are as follows:

[0019] By using data fusion and feature engineering, vibration spectrum data and weather data are synchronized on the timestamp, spectral features are extracted from the laser vibration spectrum, environmental features are extracted from the weather data, and the spectral features and environmental features are fused together. Using machine learning algorithms, vibration features of the interface switch machine are generated.

[0020] Furthermore, the characteristics of the corrosion image are as follows:

[0021] Color space analysis is used to extract rusted areas; texture colors are obtained and layered fusion is performed to generate rusted image features.

[0022] Furthermore, the process of calculating the locking notch deviation and vibration amplitude is as follows:

[0023] By calibrating the reference position during the extension and retraction process of the interface switch machine and acquiring the extension and retraction characteristics of the interface switch machine, degradation correction is performed using the following formula:

[0024]

[0025] In the formula, This is a degradation correction value; This represents the maximum number of actions. The current action time point; k is the current action count, and k = 1, 2, 3...M; N is the number of sampling points within the window; The displacement sensor reading for the k-th action;

[0026] The locking notch deviation is generated based on the degradation correction value, using the following formula:

[0027]

[0028] In the formula, This is due to the deviation of the locking notch; The time point of the last action; This is the steady-state value;

[0029] Calculate the vibration amplitude:

[0030]

[0031] In the formula, The amplitude of vibration; This refers to the side frequency amplitude; For the integration window time; For mechanical vibration weights; This refers to the dust damping coefficient; For temperature; The symbol for the derivative; The vibration time.

[0032] Furthermore, the locking notch deviation and vibration amplitude are compared with the notch threshold and vibration threshold, respectively:

[0033] Based on historical data, the notch threshold and vibration threshold are obtained respectively;

[0034] The comparison process is as follows:

[0035] An alarm will be issued when the locking gap deviation is greater than or equal to the gap threshold, or the vibration amplitude is greater than or equal to the vibration threshold.

[0036] When the locking gap deviation is less than the gap threshold and the vibration amplitude is less than the vibration threshold, continuous monitoring is performed and a difference score is obtained.

[0037] Furthermore, the process of obtaining the difference score is as follows:

[0038] The differences between the locking notch deviation and the notch threshold, and the differences between the vibration amplitude and the vibration threshold are obtained respectively. A nonlinear algorithm is used to calculate the difference score, as shown in the formula:

[0039]

[0040] In the formula, Score the difference; The threshold for the locking gap deviation; This is the vibration amplitude threshold.

[0041] Furthermore, the corrosion assessment index and switch machine condition score were obtained:

[0042] By acquiring the features of the rust image, the proportion of the reddish-rust area to the surface of the component is calculated using the following formula:

[0043]

[0044] In the formula, cx is the corrosion evaluation index; This represents the area of ​​rust. This refers to the area that is not corroded; Q represents the edge fragmentation weight; Q is the edge fractal dimension damage index.

[0045] The switch machine status score is obtained as follows:

[0046] The switch machine condition score is calculated by obtaining the difference score and corrosion assessment index, using the following formula:

[0047]

[0048] In the formula, R is the switch machine status score; Weights for difference scores; As the weight of the corrosion assessment index; The value of the feature mapping function for the difference scoring; This represents the feature mapping function value of the corrosion assessment index.

[0049] Furthermore, the process of issuing multi-level warnings and intervening based on the switch machine status score is as follows:

[0050] Acquire historical data and scoring ranges, set Eq alarm thresholds by importing the switch machine status score into the historical data scoring range, and execute preset handling strategies.

[0051] A method for monitoring the status of an interface switch machine based on image recognition includes the following steps:

[0052] Step 1: Obtain the interface switch machine status and weather data; the interface switch machine status includes extension and retraction time, switch machine image and laser vibration spectrum, and the weather data includes temperature, humidity and particulate matter concentration;

[0053] Step 2: Receive the interface switch machine status and weather data, perform data fusion, and generate the first feature using feature engineering; the first feature includes the interface switch machine extension and retraction features, the interface switch machine vibration features, and the corrosion image features;

[0054] Step 3: Extract the extension and contraction characteristics and vibration characteristics of the interface switch machine, calculate the locking notch deviation and vibration amplitude, and compare them with the notch threshold and vibration threshold respectively; when the locking notch deviation is less than the notch threshold and the vibration amplitude is less than the vibration threshold, continue monitoring, and obtain the difference score and corrosion evaluation index to calculate the switch machine status score.

[0055] Step 4: Issue multi-level warnings and intervene based on the switch machine status score.

[0056] (III) Beneficial Effects

[0057] This invention provides an interface switch machine status monitoring system and method based on image recognition, which has the following beneficial effects:

[0058] (1) By adopting advanced computer vision and deep learning algorithms, this system can automatically identify various states of the switch machine, including normal working state, slight wear, and serious damage. Compared with the traditional manual inspection method, this automated detection method not only improves the detection speed, but also greatly improves the accuracy of fault detection, and immediately issues an alarm when an abnormality is detected. This allows maintenance personnel to take action before the problem becomes more serious, thereby effectively avoiding traffic accidents caused by equipment failure.

[0059] (2) Through the analysis of historical data and pattern recognition, the system can predict which components will fail. This not only reduces unnecessary preventive maintenance work, but also ensures that key components are replaced or repaired in a timely manner, thus extending the service life of the equipment. In addition to image recognition technology, the system also integrates multiple data sources such as vibration sensors and temperature and humidity sensors to form a comprehensive monitoring platform. This platform not only provides more comprehensive status information, but also verifies the existence of faults from multiple perspectives, thereby enhancing the reliability of the system. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of the system flow of the present invention;

[0061] Figure 2 This is a schematic diagram of the overall method of the present invention. Detailed Implementation

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0063] Example 1:

[0064] Please see Figure 1 This embodiment provides an image recognition-based interface switch machine status monitoring system, which includes:

[0065] The data acquisition module acquires the interface switch machine status and weather data. The interface switch machine status includes extension / retraction time, switch machine image, and laser vibration spectrum; the weather data includes temperature, humidity, and particulate matter concentration; and outputs the interface switch machine status and weather data.

[0066] Interface switch machine status data acquisition:

[0067] Stretch time:

[0068] By using a photoelectric encoder installed on the switch machine's operating rod or motor drive shaft, the gear rotation angle or displacement change is detected, and the PLC timing module is used for timing to accurately record the start and end times of the action;

[0069] PLC timing module: Triggers timers by switching signals (such as "turnout switching action started" and "turnout switching action completed") to calculate the action time difference;

[0070] Switch machine image:

[0071] Using an industrial multispectral camera that supports visible light + infrared imaging (such as the FLIR series), fix it to the side bracket of the switch machine, cover the dynamic parts from a top angle (anti-obstruction design required), and cover the key parts of the switch machine (locking notch, indicator rod, gearbox), expand the dynamic range in strong light or backlight scenes, and avoid overexposure / underexposure; link with the switch machine action signal, automatically capture key frames when the switch switching action begins / completes;

[0072] Laser vibration spectrum:

[0073] A laser Doppler vibration meter is used to perform non-contact measurement of the vibration of the switch machine housing or gearbox surface. At the same time, a PCB Piezotronics triaxial accelerometer is installed in contact at the critical bearing position to obtain the vibration of the critical bearing. The time-domain vibration signal is converted into frequency-domain features by FFT (Fast Fourier Transform), and the natural frequency deviation of the switch machine (such as high-frequency harmonic anomalies caused by gear wear) is identified by resonance peak detection.

[0074] Weather data:

[0075] Temperature and humidity:

[0076] Using an industrial-grade temperature and humidity sensor with a built-in anti-condensation coating, it is placed close to the surface of the switch machine housing to avoid direct sunlight and collect temperature and humidity data in real time.

[0077] Particulate matter concentration:

[0078] A laser scattering particulate matter sensor is used to detect solid particulate matter in the range of 0.3~10μm. A sintered metal filter screen is installed at the air inlet of the laser scattering particulate matter sensor to prevent large particles from clogging it during sandstorms and to detect particulate matter concentration. At the same time, it is calibrated on-site regularly using a standard dust concentration detector.

[0079] The feature extraction module acquires the status of the interface switch machine and weather data, performs data fusion, and uses feature engineering to generate the first feature, which includes the interface switch machine extension and contraction features, the interface switch machine vibration features, and the corrosion image features.

[0080] Interface switch machine telescopic features:

[0081] The scaling time and switch machine images are obtained, and the scaling features of the interface switch machine are extracted using CNN. The scaling features of the interface switch machine include: switch machine damage images and switch machine displacement images.

[0082] Images of switch machine damage include cracks in the switch machine casing, peeling of the surface coating, corrosion of metal parts, and deformation of connection points. Among these, the cracks in the switch machine casing are mainly concentrated in areas of mechanical stress concentration, indicating potential structural fatigue problems during long-term use. The peeling of the surface coating and corrosion of metal parts reflect insufficient weather resistance of the equipment in harsh environments. The deformation of connection points may be caused by improper operation or external impact.

[0083] The switch machine displacement image visually reflects the real-time position changes of key components of the switch machine (such as push rods, connecting rods, etc.) during the turnout switching process, records key frames from the initial static state to the fully switched state, and shows the precise displacement trajectory and relative position relationship of each component during turnout switching;

[0084] CNN extracts interface switch machine scaling features:

[0085] By preprocessing the switch machine images, a multi-task network is designed using segmentation and regression algorithms to extract the switch machine damage images and switch machine displacement images, train the configuration, and verify the deployment.

[0086] Switch machine image preprocessing:

[0087] Four types of damaged areas (cracks in the switch machine housing, peeling of surface coating, corrosion of metal parts, and deformation of connection parts) are labeled and stored as Mask semantic segmentation tags. Classification tags are assigned according to the type of damaged area (e.g., crack_01 represents the crack number of the switch machine housing). The displacement coordinates of key components are regressed (e.g., push rod, connecting rod), and bounding boxes are marked. Combined with the extension and retraction time, the real-time position changes during the turnout switching process are named according to the timestamp.

[0088] Mask semantic segmentation label: Assign a category label to each pixel in the switch machine image to distinguish and mark different damage types and components, thereby supporting accurate image analysis and feature extraction;

[0089] Bounding Box Labeling: In switch machine images, rectangular boxes are used to mark the location and extent of specific objects or components to indicate their precise position and size in the switch machine image, facilitating subsequent detection and analysis;

[0090] Extract images of switch machine damage and switch machine displacement:

[0091] In the design of multi-task networks, two tasks are completed simultaneously through a shared CNN backbone network: segmentation of the damaged area of ​​the switch machine and regression and segmentation of the displacement coordinates of key components. This can not only share the underlying features to reduce computational redundancy, but also combine the correlation information of the two types of tasks to improve the robustness of the model.

[0092] A pre-trained ResNet-50 is used as a shared feature extractor to convert the input switch machine image into a high-dimensional feature map. The advantage of sharing is that the extracted switch machine damage image and switch machine displacement image are correlated in terms of static and dynamic features. For example, cracks often appear in areas of mechanical stress concentration with frequent displacement. The shared backbone network allows the feature learning of the two types of tasks to promote each other.

[0093] Training configuration:

[0094] Prepare a labeled dataset, including switch machine damage images with mask semantic segmentation labels and switch machine displacement images with bounding box labels; select a convolutional neural network (CNN) architecture, such as ResNet, U-Net, or EfficientNet, and adjust it according to task requirements; set training hyperparameters (such as learning rate, batch size, number of training epochs), and optimizer; to prevent overfitting, use data augmentation techniques (such as random cropping, rotation, flipping, etc.) and regularization methods (such as Dropout or L2 regularization).

[0095] Learning rate: determines the step size for updating model parameters, affecting convergence speed and stability;

[0096] Batch size: The number of samples used in each iteration, affecting memory usage and the accuracy of gradient estimation;

[0097] Training epochs: The number of times the entire dataset is trained, which affects the learning depth of the model;

[0098] Optimizer type: such as Adam or SGD, is a method used to minimize the loss function, affecting training efficiency and performance.

[0099] Regularization methods: By adding additional constraints or penalty terms to the loss function, the model's generalization ability is improved to prevent overfitting.

[0100] Verify deployment:

[0101] The model is integrated into the target system for evaluation and performance monitoring. Iterative improvements are made as needed to ensure the model consistently provides accurate predictions or analytical results. The trained model is then tested and validated in a real-world application environment to ensure its performance and stability meet expectations. Necessary adjustments and optimizations are made based on feedback to achieve reliable operation and maintenance. A lightweight code example of the model is provided below.

[0102] Python

[0103] from torchvision.models import mobilenet_v3_small

[0104] # Replace backbone with MobileNetV3

[0105] self.backbone = mobilenet_v3_small(pretrained=True).features

[0106] ```

[0107] - Edge reasoning acceleration:

[0108] bash

[0109] # Transform the model using TensorRT

[0110] trtexec --onnx=model.onnx --saveEngine=model.engine --fp16

[0111] ```

[0112] For example:

[0113] To detect long-term displacement anomalies in a switch machine at a railway hub, a sequence of displacement images continuously captured over 20 hours was used as input. A CNN segmentation network detected a 0.8mm crack (IoU=0.72) at the push rod connection. Displacement regression showed a 5% reduction in the maximum stroke of the push rod (the threshold alarm range was set to ±3%). The system was linked to vibration spectrum analysis, which confirmed that the abnormal vibration frequency was related to crack propagation. An early warning signal was output, indicating that the push rod assembly needed to be replaced, along with a heat map of the crack location.

[0114] Vibration characteristics of interface switch machines:

[0115] By using data fusion and feature engineering, vibration spectrum data and weather data are synchronized on the timestamp, spectral features are extracted from the laser vibration spectrum, environmental features are extracted from the weather data, and the spectral features and environmental features are fused. Machine learning algorithms (such as principal component analysis PCA, recursive feature elimination RFE, etc.) are used to generate vibration features of the interface switch machine.

[0116] Vibration spectrum data and weather data are synchronized in terms of timestamps:

[0117] Establish a unified time base, using UTC (Coordinated Universal Time) as the standard time, ensuring that all devices record times based on the same time standard and aligned by timestamps. If no records are found within a specific time window, data from the previous time period is used to fill the gaps.

[0118] Extracting spectral features and environmental features:

[0119] Spectral feature extraction:

[0120] The vibration spectrum data undergoes filtering and noise reduction preprocessing to remove unnecessary interference signals and obtain time-domain characteristics. The mean, variance, peak value, skewness, and kurtosis are then obtained from the processed vibration spectrum data. A temperature-vibration sensitivity matrix is ​​constructed and measured experimentally.

[0121]

[0122] Environmental feature extraction:

[0123] Acquire temperature, humidity, and particulate matter concentration from weather data; extract the mean, standard deviation, maximum, and minimum values ​​of temperature, humidity, and particulate matter concentration from the weather data; select a time window in seconds and calculate environmental characteristics within each time window;

[0124] Data fusion:

[0125] Vibration spectrum data and weather data are merged according to timestamps to form a comprehensive dataset;

[0126] If vibration spectrum data and weather data are recorded once per second, they can be directly merged into one or more records according to the timestamp; if the recording frequencies are different, appropriate resampling or interpolation processing is required to match the vibration spectrum data and weather data at the same time point according to the timestamp.

[0127] Machine learning algorithm to generate vibration characteristics of interface switch machine:

[0128] Using machine learning models, the importance of each feature is evaluated, and key features that significantly impact the model's predictive performance are identified. Principal Component Analysis (PCA) or other dimensionality reduction techniques are used to reduce the number of redundant features, simplify the model structure, and improve computational efficiency. Supervised learning models (such as support vector machines and neural networks) are built and trained using labeled historical data, with the goal of accurately predicting the switch machine's state (normal, slight wear, severe damage, etc.). Cross-validation is used to evaluate the model's generalization ability to prevent overfitting. Grid search or Bayesian optimization methods are employed to find the optimal combination of model hyperparameters.

[0129] Rust image features:

[0130] Using color space analysis, the rusted area is extracted in the HSV / Lab space, and the surface roughness is quantified using LBP local binary mode or GLCM gray-level co-occurrence matrix; texture color is obtained and hierarchical fusion is performed to extract rust features;

[0131] For example, red / orange hue threshold segmentation, as well as the proportion of rust spots, the size of the largest rust spot, and the perimeter / area ratio of the rusted area, can reflect the degree of diffusion and obtain the centroid location and spatial density of the rusted area;

[0132] For example, low-level features such as texture and color are embedded and concatenated with high-level features using deep learning, and spatial attention networks are used to dynamically weight the importance of different features.

[0133] The data analysis and anomaly detection module acquires the extension and contraction characteristics and vibration characteristics of the interface switch machine, calculates the locking notch deviation and vibration amplitude, and compares the locking notch deviation and vibration amplitude with the notch threshold and vibration threshold, respectively. When the locking notch deviation is less than the notch threshold and the vibration amplitude is less than the vibration threshold, the monitoring continues and the difference score is obtained. At the same time, the corrosion evaluation index is obtained and the switch machine status score is calculated.

[0134] Calculate the locking notch deviation:

[0135] By calibrating the reference position during the extension and retraction process of the interface switch machine and acquiring the extension and retraction characteristics of the interface switch machine, the locking notch deviation is calculated in real time. Under fault-free conditions, the average displacement value when the standard locking position and the actuating rod are in place is recorded. After each action, the steady-state displacement value is taken for degradation correction. The formula is as follows:

[0136]

[0137] In the formula, This is a degradation correction value; This represents the maximum number of actions. The current action time point; k is the current action count, with values ​​of k=1, 2, 3...M; N is the number of sampling points within the window, and when the sampling rate is ≥100Hz, N=300; The displacement sensor reading for the k-th action;

[0138] The locking notch deviation is obtained by acquiring the degradation correction value, using the following formula:

[0139]

[0140] In the formula, This is due to the deviation of the locking notch; The time point of the last action; The steady-state value represents the steady-state observation window.

[0141] Calculate the vibration amplitude:

[0142]

[0143] In the formula, The amplitude of vibration; This refers to the side frequency amplitude; For the integration window time; For mechanical vibration weights; This refers to the dust damping coefficient; For temperature; The symbol for the derivative is used to express the rate of change of a function over time; The vibration time is defined as a 3-second interval.

[0144] Historical data is based on past experiments, papers, and publicly available data;

[0145] Gap threshold:

[0146] Based on the 3σ principle of normal historical data, that is, mean ± 3 times standard deviation, the mean and standard deviation under normal operating conditions are recalculated every Y operations, and a temperature correction factor is introduced (the thermal expansion and contraction of metal affects the displacement reference).

[0147] For example:

[0148] A large amount of locking notch deviation data under normal operating conditions was collected, and preprocessed and statistically analyzed: the mean was 0.5 mm; the standard deviation was 0.1 mm; three times the standard deviation was selected as the threshold, and the formula is as follows:

[0149] Notch threshold = Mean + 3 * Standard deviation = 0.5mm + 3 * 0.1mm = 0.8mm

[0150] To ensure detection sensitivity while reducing the possibility of false alarms;

[0151] Vibration threshold:

[0152] Choose a fixed multiple of the standard deviation as the threshold; for example, if the mean of the vibration amplitude is 0.5g and the standard deviation is 0.1g, set the threshold to 0.5g + 3 * 0.1g = 0.8g.

[0153] A large amount of vibration amplitude data under normal operating conditions was collected, preprocessed, and statistically analyzed, yielding the following results: mean value 0.5g; standard deviation 0.1g; three times the standard deviation was selected as the threshold, and the formula is as follows:

[0154] Vibration threshold = Mean + 3 * Standard deviation = 0.5g + 3 * 0.1g = 0.8g

[0155] The vibration threshold is set to reduce the possibility of false alarms while ensuring detection sensitivity. The process of comparing the locking notch deviation and vibration amplitude with the notch threshold and vibration threshold, respectively, is as follows:

[0156] An alarm will be issued when the locking gap deviation is greater than or equal to the gap threshold, or the vibration amplitude is greater than or equal to the vibration threshold.

[0157] When the locking gap deviation is less than the gap threshold and the vibration amplitude is less than the vibration threshold, continuous monitoring is performed and a difference score is obtained.

[0158] The difference score is obtained as follows:

[0159] The differences between the locking notch deviation and the notch threshold, and the differences between the vibration amplitude and the vibration threshold are obtained respectively. A nonlinear algorithm is used to calculate the difference score, as shown in the formula:

[0160]

[0161] In the formula, Score the difference; The threshold for the locking gap deviation; The vibration amplitude threshold;

[0162] Obtaining the rust assessment index:

[0163] By acquiring the features of the rust image, the proportion of the reddish-rust area to the surface of the component is calculated using the following formula:

[0164]

[0165] In the formula, cx is the corrosion evaluation index; This represents the area of ​​rust. This refers to the area that is not corroded; Q represents the edge fragmentation weight; Q is the edge fractal dimension damage index.

[0166] The switch machine status score is obtained as follows:

[0167] The switch machine condition score is calculated by obtaining the difference score and corrosion assessment index, using the following formula:

[0168]

[0169] In the formula, R is the switch machine status score; Weights for difference scores; As the weight of the corrosion assessment index; The value of the feature mapping function for the difference scoring; The value of the feature mapping function for the corrosion assessment index;

[0170] The prediction and optimization module obtains the status score of the switch machine, issues multi-level warnings, and intervenes accordingly;

[0171] The status score of the switch machine is obtained, including historical data and score ranges. By substituting the score ranges from the historical data, an EQ alarm threshold is set, defined according to the equipment's normal operating range, minor anomalies, moderate anomalies, and severe anomalies. The historical data table is as follows:

[0172]

[0173] In this way, the status of the switch machine can be effectively monitored, and corresponding measures can be taken in different levels of abnormal situations to ensure the safe operation of the equipment.

[0174] Example 2:

[0175] Please see Figure 2 Based on Example 1, this embodiment also provides a method for monitoring the status of an interface switch machine based on image recognition, including the following specific steps:

[0176] Step 1: Obtain the interface switch machine status and weather data; the interface switch machine status includes extension and retraction time, switch machine image and laser vibration spectrum, and the weather data includes temperature, humidity and particulate matter concentration;

[0177] Step 2: Receive the interface switch machine status and weather data, perform data fusion, and generate the first feature using feature engineering; the first feature includes the interface switch machine extension and retraction features, the interface switch machine vibration features, and the corrosion image features;

[0178] Step 3: Extract the extension and contraction characteristics and vibration characteristics of the interface switch machine, calculate the locking notch deviation and vibration amplitude, and compare them with the notch threshold and vibration threshold respectively; when the locking notch deviation is less than the notch threshold and the vibration amplitude is less than the vibration threshold, continue monitoring, and obtain the difference score and corrosion evaluation index to calculate the switch machine status score.

[0179] Step 4: Issue multi-level warnings and intervene based on the switch machine status score.

[0180] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0181] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0182] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A status monitoring system for an interface switch machine based on image recognition, characterized in that, The system includes: The data acquisition module acquires the status of the interface switch machine and weather data. The status of the interface switch machine includes extension and retraction time, switch machine image, and laser vibration spectrum. The weather data includes temperature, humidity, and particulate matter concentration. The feature extraction module receives the interface switch machine status and weather data, performs data fusion, and generates the first feature using feature engineering; the first feature includes the interface switch machine extension and retraction features, the interface switch machine vibration features, and corrosion image features; The data analysis and anomaly detection module extracts the extension and vibration characteristics of the interface switch machine, calculates the locking notch deviation and vibration amplitude, and compares them with the notch threshold and vibration threshold, respectively. When the locking gap deviation is less than the gap threshold and the vibration amplitude is less than the vibration threshold, continuous monitoring is performed, and the difference score and corrosion evaluation index are obtained to calculate the switch machine condition score. The predictive optimization module provides multi-level warnings and interventions based on the switch machine status score; The process of calculating the locking notch deviation and vibration amplitude is as follows: By calibrating the reference position during the extension and retraction process of the interface switch machine and acquiring the extension and retraction characteristics of the interface switch machine, degradation correction is performed using the following formula: In the formula, This is a degradation correction value; This represents the maximum number of actions. The current action time point; k is the current action count, and k = 1, 2, 3...M; N is the number of sampling points within the window; The displacement sensor reading for the k-th action; The locking notch deviation is generated based on the degradation correction value, using the following formula: In the formula, This is due to the deviation of the locking notch; The time point of the last action; This is the steady-state value; Calculate the vibration amplitude: In the formula, The amplitude of vibration; This refers to the side frequency amplitude; For the integration window time; For mechanical vibration weights; This refers to the dust damping coefficient; For temperature; The symbol for the derivative; The vibration time.

2. The interface switch machine status monitoring system based on image recognition according to claim 1, characterized in that, The process of obtaining the scaling features of the interface switch machine is as follows: The scaling time and switch machine images are obtained, and the scaling features of the interface switch machine are extracted using CNN. The scaling features of the interface switch machine include: switch machine damage images and switch machine displacement images. CNN extracts interface switch machine scaling features: By preprocessing the switch machine images, a multi-task network is designed using segmentation and regression algorithms to extract the switch machine damage images and switch machine displacement images, train the configuration, and verify the deployment.

3. The interface switch machine status monitoring system based on image recognition according to claim 2, characterized in that, The vibration characteristics of the interface switch machine are as follows: By using data fusion and feature engineering, vibration spectrum data and weather data are synchronized on the timestamp, spectral features are extracted from the laser vibration spectrum, environmental features are extracted from the weather data, and the spectral features and environmental features are fused together. Using machine learning algorithms, vibration features of the interface switch machine are generated.

4. The interface switch machine status monitoring system based on image recognition according to claim 3, characterized in that: The characteristics of rust images are: Color space analysis is used to extract rusted areas; texture colors are obtained and layered fusion is performed to generate rusted image features.

5. The interface switch machine status monitoring system based on image recognition according to claim 1, characterized in that: The locking notch deviation and vibration amplitude are compared with the notch threshold and vibration threshold, respectively: Based on historical data, the notch threshold and vibration threshold are obtained respectively; The comparison process is as follows: An alarm will be issued when the locking gap deviation is greater than or equal to the gap threshold, or the vibration amplitude is greater than or equal to the vibration threshold. When the locking gap deviation is less than the gap threshold and the vibration amplitude is less than the vibration threshold, continuous monitoring is performed and a difference score is obtained.

6. The interface switch machine status monitoring system based on image recognition according to claim 5, characterized in that: The process of obtaining the difference score is as follows: The differences between the locking notch deviation and the notch threshold, and the differences between the vibration amplitude and the vibration threshold are obtained respectively. A nonlinear algorithm is used to calculate the difference score, as shown in the formula: In the formula, Score the difference; The threshold for the locking gap deviation; This is the vibration amplitude threshold.

7. The interface switch machine status monitoring system based on image recognition according to claim 6, characterized in that: Obtaining the corrosion assessment index and switch machine condition score: By acquiring the features of the rust image, the proportion of the reddish-rust area to the surface of the component is calculated using the following formula: In the formula, cx is the corrosion evaluation index; This represents the area of ​​rust. This refers to the area that is not corroded; Q represents the edge fragmentation weight; Q is the edge fractal dimension damage index. The switch machine status score is obtained as follows: The switch machine condition score is calculated by obtaining the difference score and corrosion assessment index, using the following formula: In the formula, R is the switch machine status score; Weights for difference scores; As the weight of the corrosion assessment index; The value of the feature mapping function for the difference scoring; This represents the feature mapping function value of the corrosion assessment index.

8. The interface switch machine status monitoring system based on image recognition according to claim 7, characterized in that: The process of issuing multi-level warnings and intervening based on switch machine status scores is as follows: Acquire historical data and scoring ranges, set Eq alarm thresholds by importing the switch machine status score into the historical data scoring range, and execute preset handling strategies.

9. A method for monitoring the status of an interface switch machine based on image recognition, applied to any one of the systems described in claims 1 to 8, characterized in that: Includes the following steps: Step 1: Obtain the interface switch machine status and weather data; the interface switch machine status includes extension and retraction time, switch machine image and laser vibration spectrum, and the weather data includes temperature, humidity and particulate matter concentration; Step 2: Receive the interface switch machine status and weather data, perform data fusion, and generate the first feature using feature engineering; the first feature includes the interface switch machine extension and retraction features, the interface switch machine vibration features, and the corrosion image features; Step 3: Extract the extension and contraction characteristics and vibration characteristics of the interface switch machine, calculate the locking notch deviation and vibration amplitude, and compare them with the notch threshold and vibration threshold respectively; when the locking notch deviation is less than the notch threshold and the vibration amplitude is less than the vibration threshold, continue monitoring, and obtain the difference score and corrosion evaluation index to calculate the switch machine status score. Step 4: Issue multi-level warnings and intervene based on the switch machine status score; The process of calculating the locking notch deviation and vibration amplitude is as follows: By calibrating the reference position during the extension and retraction process of the interface switch machine and acquiring the extension and retraction characteristics of the interface switch machine, degradation correction is performed using the following formula: In the formula, This is a degradation correction value; This represents the maximum number of actions. The current action time point; k is the current action count, and k = 1, 2, 3...M; N is the number of sampling points within the window; The displacement sensor reading for the k-th action; The locking notch deviation is generated based on the degradation correction value, using the following formula: In the formula, This is due to the deviation of the locking notch; The time point of the last action; This is the steady-state value; Calculate the vibration amplitude: In the formula, The amplitude of vibration; This refers to the side frequency amplitude; For the integration window time; For mechanical vibration weights; This refers to the dust damping coefficient; For temperature; The symbol for the derivative; The vibration time.

Images