A dynamic threshold drum wire rope arranging monitoring method and system

Abstract

The application discloses a kind of dynamic threshold reel steel wire rope row monitoring method and system;It is related to hoisting equipment safety monitoring technical field, the present application includes the following steps: S1, obtains steel wire rope winding image picture data;S2, image picture data is preprocessed;S3, feature extraction is carried out to image, and real-time feature vector is output;S4, real-time feature vector is compared with standard feature vector with cosine similarity, finds out the most similar standard image with current test image, and calculates feature deviation distance;S5, determine current state;S6, trigger corresponding control linkage instruction;S7, data back and keep a record;The present application visual identification obtains steel wire rope real-time image data, generates real-time feature vector after data processing, then it is compared with the standard feature vector that it is pre-constructed, so that steel wire rope row state can be monitored in real time accurately, overcome the shortcomings of existing hoisting equipment steel wire rope row monitoring technology, such as poor real-time performance of manual inspection, experience-dependent, high early abnormality missed judgment rate.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring technology for lifting equipment, and in particular to a dynamic threshold drum wire rope arrangement monitoring method and system. Background Technology

[0002] The rope winding system of crawler cranes, truck cranes, tower cranes, and mine winches is a critical load-bearing component of lifting equipment. Under variable load conditions such as wind turbine tower erection and heavy equipment hoisting, the wire rope is prone to problems such as rope biting, rope skipping, and rope stacking due to load fluctuations, speed changes, and hoisting eccentricity. Continuous operation with these defects will exacerbate wire breakage and may even lead to broken ropes causing falling objects or equipment overturning accidents. Currently, rope winding faults mainly rely on manual inspection, which cannot be monitored 24 hours a day. Faults are easily missed during nighttime or high-speed operation. Fault diagnosis relies on human experience, resulting in many early-stage anomalies being missed or misdiagnosed, and there is no data retention for faults, making it difficult to trace the cause. Existing monitoring methods are divided into mechanical limit switches and conventional visual monitoring: mechanical switches only alarm after a severe fault contact, resulting in a significant delay in early warning and failing to meet the intelligent monitoring needs under complex and variable working conditions on construction sites. Summary of the Invention

[0003] To enable more accurate and agile early assessment of wire rope condition, this application provides a dynamic threshold drum wire rope arrangement monitoring method and system.

[0004] Firstly, this application provides a dynamic threshold method for monitoring the rope arrangement of a drum wire rope, employing the following technical solution:

[0005] A method for monitoring the rope arrangement of a steel wire rope on a dynamic threshold drum includes the following steps:

[0006] S1. Obtain image data of the wire rope winding;

[0007] S2. Perform noise reduction and enhancement preprocessing on the image data to obtain the processed data;

[0008] S3. Use a pre-trained convolutional neural network to extract features from the pre-processed image and output a real-time feature vector representing the winding state of the wire rope.

[0009] S4. Compare the extracted real-time feature vectors with the standard feature vectors in the pre-established standard rope-winding feature library using cosine similarity, find the standard image that is most similar to the current test image, and calculate the feature deviation distance.

[0010] S5. Compare the calculated feature deviation distance with the preset fixed warning threshold and speed-related danger threshold to determine the current status;

[0011] S6. Trigger corresponding control linkage commands based on the judgment result: output different commands according to different states;

[0012] S7 stores real-time monitoring images, abnormal screenshots, abnormal videos, and identification logs in the database, and displays the rope-laying status in real time on the monitor, supporting historical data tracing by time or fault type.

[0013] Optionally, step S2 is specifically described as follows: First, a bilateral filtering algorithm is used to denoise the original image. The bilateral filtering formula is... for:

[0014] ;

[0015] in, For pixel position, For neighborhood windows, For spatial domain standard deviation, The standard deviation of the range, As the normalization factor, The pixel brightness function representing the original image;

[0016] Then, Laplacian sharpening is used to enhance image edges. The sharpening formula is... for:

[0017] ;

[0018] in, These are the pixel values ​​of the original input image; For the Laplacian sharpening operator, and This is the enhancement coefficient.

[0019] Optionally, the convolutional neural network in step three includes at least three convolutional layers, pooling layers, and two fully connected layers, wherein... The formula for calculating the feature map of a convolutional layer is:

[0020] ;

[0021] in, For the first The feature map output by the layer is the input of the stop circle convolutional layer; For convolution kernel, Represents a two-dimensional convolution operation; For bias terms; The ReLU activation function is used; after the last fully connected layer, the output is a real-time feature vector. .

[0022] Optionally, step four can be specifically described as follows: the standard rope winding feature library stores K images acquired under normal rope winding conditions and their corresponding standard feature vectors; for real-time feature vectors... Calculate its cosine similarity with each standard feature vector. ;

[0023] ;

[0024] In the formula, This indicates the index number of the sample within the standard rope-wound feature library. The first in the standard rope feature library The standard feature vector corresponding to each normal rope-laying sample has a value range of [value range missing]. The closer the value is to 1, the more similar the two feature vectors are. Indicates modulus;

[0025] Find out Largest subscript Then calculate the feature offset distance with the most similar image. ;

[0026] ;

[0027] In the formula, Represents the maximum cosine similarity; The larger the value, the more severe the deviation.

[0028] Optionally, in step five, the anomaly determination rule is specifically described as follows:

[0029] when At that time, the result of the anomaly determination This indicates that it is normal.

[0030] when At that time, the result of the anomaly determination , indicating a general abnormality;

[0031] when At that time, the result of the anomaly determination This indicates a serious abnormality;

[0032] in, The preset fixed warning threshold, The speed-related hazard threshold is calculated using the following formula:

[0033] ;

[0034] In the formula, The preset basic hazard threshold is met. , This is a speed correction factor, usually a negative number. The speed of the wire rope.

[0035] Optionally, step six is ​​specifically described as follows: when the anomaly determination result is a general anomaly, the control linkage system triggers an audible and visual alarm and sends a deceleration command to the crane control system to reduce the wire rope running speed to a preset safety value; when the anomaly determination result is a serious anomaly, the control linkage system triggers an audible and visual alarm and sends a stop command to immediately stop the wire rope winding and unwinding operation.

[0036] Optionally, the real-time display module shows the current rope winding scene on the monitor and overlays parameters such as the deviation distance of the display feature, the anomaly judgment result, and the current speed; the data archiving module automatically stores the original image of the moment the anomaly occurred, the anomaly screenshot, the video clips 10 seconds before and after, and the identification log to the local or remote database; the historical traceability module allows operators to retrieve historical anomaly records by time range and fault type, and can play back the corresponding video clips.

[0037] Secondly, this application also provides a dynamic threshold drum wire rope alignment monitoring system to implement the monitoring method in the first aspect, employing the following technical solution.

[0038] A dynamic threshold drum wire rope arrangement monitoring system includes the following modules:

[0039] Data acquisition module: used to acquire image data of the wire rope winding;

[0040] The data preprocessing module is used to perform noise reduction and enhancement preprocessing on the acquired image data;

[0041] The feature extraction module is used to extract features from the preprocessed image data and output a real-time feature vector representing the winding state of the wire rope.

[0042] Offset calculation module: used to compare the real-time feature vector with the standard feature vector one by one using cosine similarity and to calculate the feature deviation distance;

[0043] The judgment output module is used to determine and output the warning level based on the feature deviation distance;

[0044] The instruction sending module is used to send control linkage instructions according to the warning level;

[0045] The archive module is used to store log data in the database.

[0046] In summary, this application includes the following beneficial technical effects:

[0047] This application uses visual recognition to acquire real-time image data of the wire rope, processes the data to generate a real-time feature vector, and then compares it with a pre-constructed standard feature vector, so that the wire rope arrangement status can be monitored in real time and accurately. This overcomes the shortcomings of existing wire rope arrangement monitoring technologies for lifting equipment, such as poor real-time performance of manual inspection, reliance on experience, and high rate of early anomaly misses. Attached Figure Description

[0048] Figure 1 This is the overall flowchart of the monitoring method in this application;

[0049] Figure 2 This is a diagram of the convolutional neural network structure in the monitoring method of this application;

[0050] Figure 3 This is a schematic diagram of feature matching and most similar image retrieval in the monitoring method of this application. Detailed Implementation

[0051] The following is in conjunction with the appendix Figure 1-3 This application will be described in further detail.

[0052] This application discloses a dynamic threshold method for monitoring the arrangement of steel wire ropes on a drum, comprising the following steps:

[0053] S1. Install a camera above the wire rope drum to acquire image data of the wire rope winding.

[0054] S2. Perform noise reduction and enhancement preprocessing on the image data to obtain the processed data;

[0055] The specific image processing steps are as follows: First, the original image is denoised using a bilateral filtering algorithm. The bilateral filtering formula is... for:

[0056] ;

[0057] in, For pixel position, For neighborhood windows, For spatial domain standard deviation, The standard deviation of the range, The pixel brightness function represents the original image. The normalization factor is expressed as:

[0058] ;

[0059] Then, Laplacian sharpening is used to enhance image edges. The sharpening formula is... for:

[0060] ;

[0061] in, These are the pixel values ​​of the original input image; For the Laplacian sharpening operator, and For enhancement coefficient

[0062] S3. Use a pre-trained convolutional neural network to extract features from the pre-processed image and output a real-time feature vector representing the winding state of the wire rope.

[0063] Among them, the first in the convolutional neural network The formula for calculating the feature map of a convolutional layer is:

[0064] ;

[0065] in, For the first The feature map output by the current convolutional layer is the input to the current convolutional layer. For convolution kernel, Represents a two-dimensional convolution operation; For bias terms; The ReLU activation function is used; after the last fully connected layer, the output is a real-time feature vector. ;

[0066] After each convolutional layer, a pooling layer is typically added to reduce the spatial size of the feature map and enhance translation invariance. Max pooling is used, defined as follows:

[0067] ;

[0068] in, The pooling window size is the feature map output by the current convolutional layer. Step size is . This means taking the maximum value within the pooling window as the output feature map. In position Pixel values;

[0069] After several convolutional and pooling layers, the feature map is flattened into a one-dimensional vector and fed into a fully connected layer; the calculation formula for the fully connected layer is:

[0070] ;

[0071] in, This is the feature map output by the last pooling layer. To perform the flattening operation, the multidimensional feature map is converted into a one-dimensional vector; This is the weight matrix of the fully connected layer. This is the bias term. After the last fully connected layer, the final high-dimensional feature vector is obtained. .

[0072] S4. Compare the extracted real-time feature vectors with the standard feature vectors in the pre-established standard rope-winding feature library using cosine similarity, find the standard image that is most similar to the current test image, and calculate the feature deviation distance.

[0073] Specifically, the standard rope-winding feature library stores K images acquired under normal rope-winding conditions and their corresponding standard feature vectors; for real-time feature vectors... Calculate its cosine similarity with each standard feature vector. ;

[0074] ;

[0075] In the formula, This indicates the index number of the sample within the standard rope-wound feature library. The first in the standard rope feature library The standard feature vectors corresponding to each normal rope-laying sample are used. Cosine similarity is used to measure the closeness of two vectors in direction, and its value ranges from 1 to 2. The closer the value is to 1, the more similar the two feature vectors are. Indicates modulus;

[0076] Find out Largest subscript Then calculate the feature offset distance with the most similar image. ;

[0077] ;

[0078] In the formula, Represents the maximum cosine similarity; The larger the value, the more severe the deviation.

[0079] S5. Compare the calculated feature deviation distance with the preset fixed warning threshold and speed-related danger threshold to determine the current status;

[0080] The specific description of the anomaly detection rules is as follows:

[0081] when At that time, the result of the anomaly determination This indicates that it is normal.

[0082] when At that time, the result of the anomaly determination , indicating a general abnormality;

[0083] when At that time, the result of the anomaly determination This indicates a serious abnormality;

[0084] in, The preset fixed warning threshold, The speed-related hazard threshold is calculated using the following formula:

[0085] ,

[0086] In the formula, The preset basic hazard threshold is met. , This is a speed correction factor, usually a negative number. For the speed of the wire rope

[0087] S6. Trigger corresponding control linkage commands based on the abnormality judgment result: output different commands according to different states; output deceleration command when there is a general abnormality, output stop command when there is a serious abnormality, and trigger audible and visual alarms at the same time.

[0088] S7 stores real-time monitoring images, abnormal screenshots, abnormal videos, and identification logs in the database, and displays the rope-laying status in real time on the monitor, supporting historical data tracing by time or fault type.

[0089] This application also provides a dynamic threshold drum wire rope alignment monitoring system, which includes:

[0090] Data acquisition module: used to acquire wire rope winding image data, specifically a high-definition camera, power supply module and wired transmission component deployed on the side or above the drum, used to acquire wire rope winding images in real time and transmit them to the data preprocessing module via wired means;

[0091] The data preprocessing module is used to perform noise reduction and enhancement preprocessing on the acquired image data;

[0092] The feature extraction module is used to extract features from the preprocessed image data and output a real-time feature vector representing the winding state of the wire rope.

[0093] Offset calculation module: used to compare the real-time feature vector with the standard feature vector one by one using cosine similarity and to calculate the feature deviation distance;

[0094] The judgment output module is used to determine and output the warning level based on the feature deviation distance;

[0095] The instruction sending module is used to send control linkage instructions according to the warning level;

[0096] The archive module is used to store log data in the database.

[0097] Taking crawler crane hoisting construction as an example, the wire rope drum operates at high speed, with a travel speed of... The video capture system's camera is deployed above the reel, capturing the rope winding scene without blind spots. The data preprocessing module performs bilateral filtering to denoise the image and video data. For spatial domain standard deviation, (Standard deviation of range) and Laplace sharpening ( This results in an enhanced image.

[0098] The deep learning detection unit uses a pre-trained CNN model to extract real-time feature vectors. The feature matching analysis unit will The cosine similarity is calculated one by one with all normal feature vectors in the standard rope feature library to obtain the maximum similarity. The most similar image is image number 47 in the database. Then the feature deviation distance... .

[0099] The system has a preset fixed warning threshold. Basic risk threshold Speed ​​correction factor Calculation speed correlated with danger threshold: anomaly detection model output (Serious abnormality).

[0100] Upon receiving a severe anomaly signal, the control linkage system immediately triggers an audible and visual alarm. Simultaneously, it sends a stop command to the crawler crane's PLC controller via the equipment interlocking unit, halting the wire rope drum's winding and unwinding operations within 0.5 seconds. The information storage and display system automatically captures images of the anomaly, records 10 seconds of video before and after it, and generates an identification log containing parameters such as feature distance, similarity, and speed, storing it in a local database. Simultaneously, the abnormal area is highlighted on the real-time display, and a pop-up message appears: "Severe Anomaly: Number of Rope Turns Exceeds Standard, Emergency Stopped." Operators can retrieve the anomaly record by time using the historical tracking module, replaying video clips to analyze the cause of the fault as wear on the drum's guide wheel causing a deviation in the wire rope entry angle, leading to winding errors. Based on the analysis, on-site personnel promptly replaced the guide wheel, preventing further accidents.

[0101] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for monitoring the arrangement of steel wire ropes on a dynamic threshold drum, characterized in that: Includes the following steps: S1. Obtain image data of the wire rope winding; S2. Perform noise reduction and enhancement preprocessing on the image data to obtain the processed data; S3. Use a pre-trained convolutional neural network to extract features from the pre-processed image and output a real-time feature vector representing the winding state of the wire rope. S4. Compare the extracted real-time feature vectors with the standard feature vectors in the pre-established standard rope-winding feature library using cosine similarity, find the standard image that is most similar to the current test image, and calculate the feature deviation distance. S5. Compare the calculated feature deviation distance with the preset fixed warning threshold and speed-related danger threshold to determine the current status; S6. Trigger corresponding control linkage commands based on the anomaly determination result: output different commands according to different states; S7 stores real-time monitoring images, abnormal screenshots, abnormal videos, and identification logs in the database, and displays the rope-laying status in real time on the monitor, supporting historical data tracing by time or fault type.

2. The method for monitoring the arrangement of steel wire rope on a dynamic threshold drum according to claim 1, characterized in that: The specific description of step S2 is as follows: First, a bilateral filtering algorithm is used to denoise the original image. The bilateral filtering formula is... for: ； in, For pixel position, For neighborhood windows, For spatial domain standard deviation, The standard deviation of the range, As the normalization factor, The pixel brightness function representing the original image; Then, Laplacian sharpening is used to enhance image edges. The sharpening formula is... for: ； in, These are the pixel values ​​of the original input image; For the Laplacian sharpening operator, and This is the enhancement coefficient.

3. The method for monitoring the arrangement of steel wire rope on a dynamic threshold drum according to claim 2, characterized in that: The convolutional neural network in step three contains at least three convolutional layers, pooling layers, and two fully connected layers. The formula for calculating the feature map of a convolutional layer is: ； in, For the first The feature map output by the layer is the input of the stop circle convolutional layer; For convolution kernel, Represents a two-dimensional convolution operation; For bias terms; The ReLU activation function is used; after the last fully connected layer, the output is a real-time feature vector. .

4. The method for monitoring the arrangement of steel wire rope on a dynamic threshold drum according to claim 1, characterized in that: The specific description of step four is as follows: The standard rope winding feature library stores multiple images acquired under normal rope winding conditions and their corresponding standard feature vectors. K is the number of samples in the database; for real-time feature vectors Calculate its cosine similarity with each standard feature vector. ; ； In the formula, This indicates the index number of the sample within the standard rope-wound feature library. Let represent the standard feature vector corresponding to the k-th normal rope-laying sample in the standard rope-laying feature library. Cosine similarity is used to measure the closeness of two vectors in a direction, and its value ranges from 1 to 2. The closer the value is to 1, the more similar the two feature vectors are. Indicates modulus; Find out Largest subscript Then calculate the feature offset distance with the most similar image. ; ； In the formula, Represents the maximum cosine similarity; The larger the value, the more severe the deviation.

5. The method for monitoring the arrangement of steel wire rope on a dynamic threshold drum according to claim 1, characterized in that: In step five, the anomaly detection rules are specifically described as follows: when At that time, the result of the anomaly determination This indicates that it is normal. when At that time, the result of the anomaly determination , indicating a general abnormality; when At that time, the result of the anomaly determination This indicates a serious abnormality; in, The preset fixed warning threshold, The speed-related hazard threshold is calculated using the following formula: ； In the formula, The preset basic hazard threshold is met. , This is a speed correction factor, usually a negative number. The speed of the wire rope.

6. The method for monitoring the arrangement of steel wire rope on a dynamic threshold drum according to claim 5, characterized in that: The specific description of step six is ​​as follows: When the abnormality determination result is a general abnormality, the control linkage system triggers an audible and visual alarm and sends a deceleration command to the crane control system to reduce the wire rope running speed to a preset safety value; when the abnormality determination result is a serious abnormality, the control linkage system triggers an audible and visual alarm and sends a stop command to immediately stop the wire rope winding and unwinding operation.

7. The method for monitoring the arrangement of steel wire rope on a dynamic threshold drum according to claim 1, characterized in that: The specific description of step seven is as follows: The real-time display module displays the current rope winding scene on the monitor and overlays parameters such as the deviation distance of the display feature, the anomaly judgment result, and the current speed; the data archiving module automatically stores the original image at the moment of the anomaly, the anomaly screenshot, the video clips 10 seconds before and after, and the identification log to the local or remote database; the historical traceability module allows operators to retrieve historical anomaly records by time range and fault type, and can replay the corresponding video clips.

8. A dynamic threshold drum wire rope alignment monitoring system, used to implement the dynamic threshold drum wire rope alignment monitoring method according to any one of claims 1 to 7, characterized in that: Includes the following modules: Data acquisition module: used to acquire image data of the wire rope winding; The data preprocessing module is used to perform noise reduction and enhancement preprocessing on the acquired image data; The feature extraction module is used to extract features from the preprocessed image data and output a real-time feature vector representing the winding state of the wire rope. Offset calculation module: used to compare the real-time feature vector with the standard feature vector one by one using cosine similarity and to calculate the feature deviation distance; The judgment output module is used to determine and output the warning level based on the feature deviation distance; The instruction sending module is used to send control linkage instructions according to the warning level; The archive module is used to store log data in the database.

Images