Coal mine hoist hook unhooking and tail rope winding intelligent monitoring method based on RFID

Abstract

The application provides a coal mine hoist hook uncoupling and tail rope winding intelligent monitoring method based on RFID, installs an intelligent monitoring system, constructs a hook uncoupling gradient database, constructs a tail rope winding state library, obtains monitoring data online, sequentially carries out median filtering method and generates an adversarial network denoising process on the monitoring data, adopts a least square decoupling algorithm to separate the coupling interference between tags from the denoised uncoupling real-time RSSI matrix, obtains an independent coupling matrix, adopts a random forest classification algorithm to construct a classification model, and identifies the winding state of the tail rope by using the feature vector extracted from the real-time matrix; determines the uncoupling degree based on the independent coupling matrix; carries out hierarchical early warning based on the combined state of the uncoupling degree estimate value and the tail rope winding state, and carries out redundant verification of hierarchical early warning through the strain change rate. The method can realize quantitative evaluation of the hoist uncoupling state and automatic and accurate classification judgment of the tail rope winding state, and also has intelligent early warning control.

Description

Technical Field

[0001] This invention belongs to the field of intelligent monitoring technology, specifically an RFID-based intelligent monitoring method for hook detachment and tail rope entanglement in coal mine hoists. Background Technology

[0002] In underground coal mine hoisting systems, the hook connection status and tail rope operation are critical factors in ensuring the safety of hoisting operations. Abnormalities such as incomplete hook closure or tail rope entanglement can easily lead to major safety accidents such as load slippage and wire rope breakage. Traditional monitoring methods mainly rely on manual inspection, mechanical limit switches, or single-point measurements using strain gauges. These traditional methods suffer from numerous drawbacks, including poor real-time performance, limited coverage, and weak resistance to interference, making them insufficient to meet the inherent safety monitoring requirements of intelligent mines. With the rapid development of Radio Frequency Identification (RFID) technology, it has become possible to monitor the operational status of hoisting systems using its passive, interference-resistant, and easy-to-deploy characteristics. However, the complex underground environment, strong metal interference, and significant electromagnetic wave multipath effects pose challenges to the stability of tag signals and the reliability of data. There is an urgent need to introduce high-density array deployment and advanced signal processing algorithms to improve the system's anti-interference capabilities and discrimination accuracy. Therefore, a high-precision, intelligent monitoring method for underground hoists is urgently needed to achieve real-time perception and early warning control of hook disengagement status and tail rope entanglement. Summary of the Invention

[0003] To address the problems existing in the prior art, this invention provides an intelligent monitoring method for hook unhooking and tail rope entanglement of coal mine hoists based on RFID. This method has advantages such as high intelligence, high recognition accuracy, good reliability, and adaptability to complex environments. It effectively combines cosine similarity matching, Gaussian regression correction, multi-threshold control, and redundant physical quantity fusion technology. It can not only realize the quantitative assessment of the hoist unhooking state and the automatic and accurate classification and judgment of the tail rope state, but also has intelligent early warning control, which can realize dynamic response and rapid shutdown operation under different early warning conditions.

[0004] To achieve the above objectives, the present invention provides an intelligent monitoring method for hook unhooking and tail rope entanglement in coal mine hoists based on RFID, comprising the following steps:

[0005] Step 1: Install the intelligent monitoring system;

[0006] Multiple passive RFID tags are installed at the junction of the hook and the crossbar to form a two-dimensional sensing coupling array; multiple passive RFID tags are installed on each tail rope to form a linear array; a dense tag array is constructed through the two-dimensional sensing coupling array and the linear array; strain gauges are installed on the contact surface between the hook and the crossbar, and strain gauges are installed in the tail rope channel; at the same time, an explosion-proof RFID reader is installed above the hook, an edge computing module is installed near the hoist, a controller is installed in the monitoring center, and communication connections are established between the edge computing module and the dense tag array, strain gauges, and strain gauges; a communication connection is also established between the edge computing module and the controller.

[0007] Step Two: Data Acquisition and Benchmark Library Construction;

[0008] B1: Construct a hook-and-unhook gradient database;

[0009] B11: Closed-state reference acquisition; Set the initial distance between the hook and the crossbar to 0mm, continuously acquire 100 sets of RSSI matrix data using an RFID reader, and obtain the initial coupling matrix through matrix mean filtering. ;

[0010] B12: Unhooking gradient simulation; using a hydraulic traction device to pull the hook, reducing the distance between the hook and the crossbar from... to according to The step size is adjusted gradually, and each gradient represents Degree of decoupling, simulation arrive common Each state has a data collection point; data is collected under each state. The RSSI matrix data was processed by matrix mean filtering to obtain the gradient matrices. ;

[0011] B13: Standardization of the decoupling matrix; construction of a decoupling gradient database by combining the matrices obtained in A11 and A12. ,in, ; for decoupling gradient database Each matrix in the matrix is ​​normalized using gradient matrix processing to obtain a decoupling normalized matrix. Finally, a standardized decoupling gradient database is obtained. ,in, ;

[0012] B2: Construct a library of tail rope winding states;

[0013] B21: Reference State Acquisition; The tail tether is set to fully extended, the distance between adjacent tags is the initial reference distance, 100 sets of RSSI matrix data are continuously acquired using an RFID reader, and the reference matrix is ​​obtained through matrix mean filtering. ;

[0014] B22: Simulation of different winding states; by changing the state of the tail rope, the distance between adjacent tags is gradually compressed, decreasing each time. , used to simulate common Each entanglement state is sampled. The RSSI matrix data was processed by matrix mean filtering to obtain the corresponding RSSI matrices. ;

[0015] B23: Normalization of the winding matrix; constructing a winding gradient library by combining the matrices obtained in B21 and B22. ,in, ; For the winding gradient library Each matrix in the matrix is ​​normalized by gradient matrix to obtain the entanglement normalized matrix. Finally, a standardized entanglement state database is obtained. ,in, ;

[0016] Step 3: Obtain monitoring data online;

[0017] By polling and sampling the dense tag array at fixed intervals using an RFID reader, the real-time RSSI matrix of the decoupling tag is obtained. and the entanglement of real-time RSSI matrix Simultaneously, strain gauge 1 and strain gauge 2 are used to acquire uncoupling strain signals and winding strain signals in real time and send them to the edge computing module. The edge computing module obtains uncoupling strain data and winding strain data based on the uncoupling strain signals and winding strain signals, and obtains the strain change rate. ;

[0018] Step 4: Denoise the monitoring data;

[0019] First, median filtering is used for preliminary smoothing and denoising. Then, a generative adversarial network is introduced for secondary deep denoising to obtain the denoised RSSI matrix.

[0020] Step 5: Data decoupling;

[0021] The least squares decoupling algorithm is used to decouple the real-time RSSI matrix after denoising. The coupling interference between tags is separated to obtain the independent coupling matrix C;

[0022] Step Six: Identify the tangling state of the tail rope;

[0023] F1: A classification model is constructed using the random forest classification algorithm and pre-trained to obtain a classification model for the tail rope entanglement state.

[0024] F2: From the denoised, wrapped real-time RSSI matrix Extracting feature vectors ;

[0025] F3: Feature vectors The data is input into the tail rope entanglement classification model to classify the tail rope entanglement state and output the tail rope entanglement state φ, which includes three types: normal, slight entanglement and severe entanglement.

[0026] Step 7: Determine the degree of decoupling based on the independent coupling matrix;

[0027] G1: Calculate the independent coupling matrix C and the normalized decoupling gradient database sequentially using the cosine similarity matching algorithm. Find the cosine similarity of all matrices in the given matrix and identify the matrix with the highest similarity. and the matrix The corresponding degree of decoupling is used as the initial estimate of the degree of decoupling. ;

[0028] G2: Combining distance transformation and Gaussian process regression to correct the initial decoupling estimate Obtain the final estimate of the degree of decoupling ;

[0029] Step 8: Based on the estimated degree of decoupling The combined state of the tail rope entanglement φ is used to classify and warn of early warning, and the strain rate of change is used to determine the early warning. Redundancy verification of tiered early warning is performed.

[0030] As a preferred embodiment, in step one, the multiple passive RFID tags are evenly divided into two groups, with 10 UHF passive RFID tags in each group. The 10 UHF passive RFID tags in the first group are installed sequentially on the hook with a spacing of ≤5 cm, and the 10 UHF passive RFID tags in the second group are installed sequentially on the crossbar with a spacing of ≤5 cm, forming... Two-dimensional sensing coupling array; 10 passive RFID tags are installed longitudinally on the tail rope with a spacing of ≤10 cm; the RFID reader is located 1 meter above the hook;

[0031] The passive RFID tags 1 and 2 have the same structure, both including a UHF passive RFID tag body, an RF transparent epoxy resin protective layer, a buffer protective layer, a protective layer, and a high-viscosity silicone layer. The RF transparent epoxy resin protective layer wraps around the outside of the UHF passive RFID tag body; the buffer protective layer wraps around the outside of the RF transparent epoxy resin protective layer; the buffer protective layer adopts a composite structure of polyurethane foam and sparse magnets; the protective layer and the high-viscosity silicone layer are respectively bonded to opposite sides of the buffer protective layer; the protective layer is made of wear-resistant and corrosion-resistant material; the high-viscosity silicone layer has a low coefficient of friction. Industrial-grade high-viscosity silicone;

[0032] As a preferred embodiment, in steps B11 and B12 of step two, matrix mean filtering is performed according to formula (1), and a matrix is ​​constructed. In step B13 of step two, the decoupling normalization matrix is ​​obtained according to formula (2). ;

[0033] (1);

[0034] In the formula, Indicates the first In the second sampling, the first With the Signal strength between tags;

[0035] (2);

[0036] In the formula, For matrix The mean; For matrix Standard deviation; ;

[0037] In steps B21 and B22 of step two, matrix mean filtering is performed according to formula (3), and a matrix is ​​constructed. In step B13 of step two, the winding normalization matrix is ​​obtained according to formula (4). ;

[0038] (3);

[0039] (4);

[0040] In the formula, For matrix The mean; For matrix standard deviation .

[0041] As a preferred option, in step three, the strain rate of change is obtained according to formula (5). ;

[0042] (5);

[0043] In the formula, for Real-time response data.

[0044] As a preferred option, the noise reduction process for the monitoring data in step four is as follows:

[0045] D1: First, use median filtering for initial smoothing and noise reduction; for each element of the matrix... According to formula (6), the median of itself and its adjacent elements is taken as the filtered value. Construct the filtered RSSI matrix ;

[0046] (6);

[0047] In the formula: and They are respectively Two adjacent elements;

[0048] D2: Introducing a generative adversarial network to filter the RSSI matrix Deep denoising is performed to obtain a more accurate RSSI matrix. ;

[0049] D21: Building Generators and discriminator And obtain the discriminator according to formula (7). loss function The generator is obtained according to formula (8). loss function ;

[0050] (7);

[0051] (8);

[0052] In the formula, Represents the preprocessed RSSI matrix Distribution; Representing noise data Distribution;

[0053] D22: Using matrices As the training set, by minimizing the loss function and loss function Training Generator and discriminator ;

[0054] D3: Using a trained generator For real-time RSSI matrix Denoising is performed, and the denoised RSSI matrix is ​​obtained according to formula (9). ;

[0055] (9).

[0056] As a preferred embodiment, in step five, the independent coupling matrix is ​​obtained by minimizing the objective function according to formula (10). ;

[0057] (10).

[0058] As a preferred option, in step F1 of step six, the process of obtaining the tail rope entanglement state classification model is as follows:

[0059] F11: Using multiple decision trees Construct a random forest classification model;

[0060] F12: Standardize the winding state database Transform into the corresponding training feature set ,in, Using training feature sets The random forest classification model was trained, and the tail rope entanglement state classification model was obtained after training.

[0061] In step six, F2, the denoised, wrapped real-time RSSI matrix is ​​calculated according to formula (11). Obtaining feature vectors ;;

[0062] (11);

[0063] In step F3 of step six, the entanglement status label of the tail rope is obtained using a voting mechanism according to formula (12). , The three winding states φ correspond to normal, slight winding, and severe winding, respectively;

[0064] (12);

[0065] In the formula, Indicates the first Each decision tree is used to process the input feature vector. The classification results; This indicates the operation of taking the mode;

[0066] As a preferred embodiment, in step G1 of step seven, the initial decoupling degree estimate is... The process is as follows:

[0067] G11: Calculate the cosine similarity according to formula (13) ;

[0068] (13);

[0069] G12: Select the one with the highest similarity according to formula (14) The corresponding gradient is used as an estimate of the initial degree of decoupling. ,in, ;

[0070] (14).

[0071] As a preferred option, in step G2 of step seven, the initial degree of decoupling is corrected. The estimation process is as follows:

[0072] G21: Independent coupling matrix Converted to a relative distance matrix between the hook and the crossbar; based on a pre-calibrated RSSI-distance mapping function. According to formula (15), the matrix Each element in Convert to the corresponding relative distance ;

[0073] (15);

[0074] G22: Calculate the average of all relative distances according to formula (16). ;

[0075] (16);

[0076] G22: Building based on The two-dimensional input vector is input into the pre-trained Gaussian process regression model. And according to formula (17), the initial decoupling degree After making corrections, we obtain the corrected estimate of the degree of decoupling. ;

[0077] (17).

[0078] As a preferred option, the tiered early warning process in step eight is as follows:

[0079] if ,and If the indicator is normal, it can be initially determined that the hook and tail rope are in normal condition, and further verification is needed. ,like If no significant strain changes are detected, it indicates that the system is currently in normal operation and no warning action is required.

[0080] if ,and If the indication is slight tangling, it can be initially determined that the hook is in normal condition and the tail rope is slightly tangled. Further verification is needed. ,like ,and If the temperature continues to rise, it confirms that the hook is in normal condition but the tail rope is slightly tangled, triggering a Level 1 warning. This warning sends a message urging the user to check the tail rope for any minor tangling. Simultaneously, it displays the current status... value;

[0081] if ,and If the indicator shows severe tangling, the initial assessment is that the hook is in normal condition but the tail rope is severely tangled. Further verification is needed. ,like ,and If the hoist continues to rise, it confirms that the hook is in normal condition but the tail rope is severely tangled, triggering an emergency warning and stopping the hoist. Simultaneously, it issues a warning message: "Severe tail rope tangling detected; machine stopped; please immediately check for severe tail rope tangling." It also displays the current status... value;

[0082] if ,and If the indicator is normal, the initial judgment is that the hook has slightly come loose and the tail rope is in normal condition. Further verification is needed. ,like ,and If the temperature continues to rise, it confirms that the hook has slightly come undone and the tail rope is in normal condition, triggering a Level 1 warning. Simultaneously, a message is issued requesting a check for any partial detachment of the hook, along with the current status of the warning. value;

[0083] if ,and If the indicator shows slight entanglement, it initially suggests that the hook has slightly come off and the tail rope is slightly entangled; further verification is needed. ,like ,and and If both indicators continue to rise, it confirms a slight hook detachment and slight tail rope tangling, triggering a Level 1 warning. Simultaneously, a warning message is issued urging the user to check for potential issues with the hook and tail rope. The current status is also displayed. Value and value;

[0084] if ,and If the indicator shows severe tangling, it initially suggests that the hook has slightly come loose and the tail rope is severely tangled, requiring further verification. ,like ,and If the hoist continues to rise, it will confirm a slight hook detachment and severe tail rope entanglement, triggering an emergency warning and stopping the hoist. Simultaneously, it will issue a warning message stating that a slight hook detachment and severe tail rope entanglement have been detected, the hoist has been stopped, and the severe tail rope entanglement should be checked first. At the same time, it will display the current status... Value and value;

[0085] if ,and If the indicator is normal, the initial judgment is that the hook has seriously come off the hook, but the tail rope is in normal condition. Further verification is needed. ,like ,and A sharp rise indicates a confirmed serious hook detachment, while the tail rope remains in normal condition. This triggers an emergency warning and shuts down the hoist. Simultaneously, a message is issued urging immediate inspection of the hook detachment, stating that a serious hook detachment has been detected and the hoist has been stopped. The current status is also displayed. Value and value;

[0086] if ,and If the indication is slight entanglement, the initial judgment is that the hook is severely unhooked and the tail rope is slightly entangled. Further verification is needed. ,like ,and and If both conditions continue to rise, it confirms a severe hook detachment and slight entanglement of the tail rope, triggering an emergency warning and stopping the hoist. Simultaneously, a message is issued indicating that a severe hook detachment and slight tail rope entanglement have been detected, the hoist has been stopped, and priority should be given to checking for the severe hook detachment. The current status is also displayed. value;

[0087] if ,and If the indicator shows severe tangling, it initially suggests that the hook is severely unhooked and the tail rope is severely tangled, requiring further verification. ,like ,and and If both conditions continue to rise, it confirms a severe hook detachment and severe tail rope entanglement, triggering an emergency warning and stopping the hoist. Simultaneously, a warning message is issued: "Severe hook detachment and severe tail rope entanglement detected; hoist stopped; please immediately check for these conditions." The current status is also displayed. Value and value.

[0088] This invention discloses an RFID-based method for monitoring hook disengagement and tail rope entanglement in coal mine hoists. First, multiple UHF passive RFID tags are arranged on the hook and crossbar to form a two-dimensional sensing coupling array. Multiple UHF passive RFID tags are arranged on each tail rope to form a linear array. This dense tag array, formed by the two-dimensional sensing coupling array and the linear array, effectively covers all areas to be monitored, ensuring comprehensive monitoring coverage. By placing strain gauges at the hook contact interface and in the tail rope channel, strain data can be effectively detected simultaneously with RSSI data reading when mechanical deformation occurs between the hook and crossbar or when the tail rope deforms. Thus, strain data can be used as a redundancy verification method, effectively ensuring the accuracy and reliability of monitoring and avoiding false alarms. Second, a standardized disengagement gradient database is established by simulating different disengagement states using a hydraulic device. This not only effectively reduces the database construction cost and cycle but also facilitates the comparison of the matching degree of real-time matrices using a cosine similarity algorithm during subsequent disengagement degree matching, which is beneficial for quantitative judgment of the disengagement degree. In establishing a standardized decoupling gradient database, matrix mean filtering is used for denoising, eliminating environmental noise and significantly improving RSSI data quality, thus enabling the establishment of a more accurate baseline coupling matrix. Simultaneously, establishing a standardized winding state database by simulating different degrees of tail rope winding not only effectively reduces database construction costs and time but also provides accurate basic data for subsequent tail rope state classification. Furthermore, during online monitoring, polling and sampling the dense tag array at a fixed period effectively reduces signal overlap in dense tag environments, lowers the false read rate, and improves overall recognition efficiency and accuracy. During online monitoring, strain change rate is simultaneously obtained. This provides a reliable basis for subsequent redundancy verification, helping to ensure the accuracy of subsequent early warnings. Then, a dual denoising mechanism of median filtering + GAN denoising is applied to the RSSI matrix, which not only eliminates abrupt noise but also effectively filters nonlinear background interference, thereby obtaining high-quality RSSI feature data and providing a stable data foundation for distinguishing between hook unhooking and tail rope entanglement. Furthermore, a least-squares decoupling algorithm is used to separate independent signals from the mutually coupled tag signals, obtaining an independent coupling matrix. This reveals the actual signal relationship between each tag, providing reliable data support for determining whether the hoist hook has unhooked and whether the tail rope is knotted. Then, a random forest algorithm is used to construct a tail rope entanglement state classification model, which can achieve automatic, efficient, and accurate classification of the tail rope state based on feature vectors, thus significantly improving the accuracy and stability of classification. Finally, an initial unhooking degree estimate is obtained through cosine similarity matching, and then the final unhooking degree is obtained by combining distance transformation and Gaussian process regression correction, which can efficiently and accurately quantify the unhooking degree of the hoist hook. Finally, in the graded early warning process, graded early warnings are carried out based on the combined state of the estimated degree of decoupling and the state of the tail rope entanglement. The redundancy verification of graded early warnings is carried out by the strain change rate, which significantly improves the reliability and accuracy of the early warning and avoids false alarms.

[0089] This method boasts advantages such as high intelligence, high recognition accuracy, high reliability, and adaptability to complex environments. It effectively combines cosine similarity matching, Gaussian regression correction, multi-threshold control, and redundant physical quantity fusion technologies. It not only enables quantitative assessment of the hoist's unhooking state and automatic, accurate classification of the tail rope state, but also features intelligent early warning control, allowing for dynamic response and rapid shutdown under different early warning conditions. Compared to traditional single-point strain gauge structural monitoring methods, this invention, based on a dense RFID array and machine learning algorithms, offers significant advantages in robustness, accuracy, and applicability. It is suitable for safety monitoring and control of critical connection structures in intelligent mines, high-risk mechanical structures, and port machinery. Attached Figure Description

[0090] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0091] This invention proposes an RFID-based method for monitoring hook detachment and tail rope entanglement in coal mine hoists. By combining a high-density passive RFID array, an industrial-grade reader / writer, signal processing, and intelligent identification algorithms, it achieves high-precision sensing and real-time early warning of critical connection states of the hoist. The technical solution of this invention is described in detail below with reference to actual implementation. The invention is further illustrated below with reference to the accompanying drawings.

[0092] like Figure 1As shown, this invention provides an intelligent monitoring method for hook detachment and tail rope entanglement in coal mine hoists based on RFID, comprising the following steps:

[0093] Step 1: Install an intelligent monitoring system; to achieve high-precision status perception of key nodes of the hoist, this invention designs an intelligent monitoring system based on UHF passive RFID technology. The core of the system includes: a dense tag array, an industrial-grade RFID reader, an edge computing module, and a controller, forming a complete signal acquisition and processing path. Preferably, the system supports the collaborative parallel deployment of multiple RFID readers and accesses the upper-level monitoring platform through local edge nodes, possessing good scalability and deployment flexibility. The specific structure is as follows:

[0094] Multiple passive RFID tags are installed at the junction of the hook and the crossbar to form a two-dimensional sensing coupling array. This deployment scheme can effectively cover... The sensor resolution reaches the level of detecting minute structural displacements. Multiple passive RFID tags are installed on each tail rope to form a linear array for monitoring its tension and entanglement. This deployment method facilitates the capture of relative displacement and signal anomalies between tags caused by overall and local entanglement of the tail rope. A dense tag array is constructed using a two-dimensional sensing coupling array and a linear array. To reduce false alarm rates and improve system safety, high-sensitivity strain gauges are installed at the hook contact interface and in the tail rope channel. Strain gauge one is installed at the hook-crossbar contact surface, and strain gauge two is installed in the tail rope channel. Simultaneously, an explosion-proof RFID reader is installed above the hook. This RFID reader can effectively cover all passive RFID tags to minimize blind spots and dead zones. As a priority, to ensure stable tag information reading and system reliability, an industrial-grade explosion-proof UHF RFID reader module is selected. An edge computing module is installed near the hoist, and a controller is installed at the monitoring center. Communication connections are established between the edge computing module and the dense tag array, strain gauge one and strain gauge two, and between the edge computing module and the controller. As a priority, the edge computing module uses an industrial Ethernet protocol (such as Modbus TCP) to communicate with the controller. More preferably, the controller is a PLC controller. The edge computing module is used for data preprocessing, decoupling, identification, and early warning logic judgment.

[0095] As a preferred embodiment, multiple passive RFID tags are evenly divided into two groups, with 10 UHF passive RFID tags in each group. The 10 UHF passive RFID tags in the first group are installed sequentially on the hook with a spacing of ≤5 cm, and the 10 UHF passive RFID tags in the second group are installed sequentially on the crossbar with a spacing of ≤5 cm, forming... A two-dimensional sensing coupling array; a dense array of 20 tags (spaced ≤ 5cm) enables the system to capture minute displacements of 0.5mm between the hook and the crossbar. Combined with a subsequent standardized gradient database, it can achieve high-precision hook release status monitoring. Ten passive RFID tags are installed longitudinally on the tail rope at a spacing ≤ 10cm; the RFID reader is located 1 meter above the hook to ensure stable reading of all tag signals.

[0096] The passive RFID tags 1 and 2 have identical structures and operate within a frequency range of 860–960 MHz. Both passive RFID tags 1 and 2 include a UHF passive RFID tag body, a radio frequency transparent epoxy resin protective layer, a buffer protective layer, a protective layer, and a high-adhesion silicone layer. The radio frequency transparent epoxy resin protective layer wraps around the outside of the UHF passive RFID tag body and is sealed with radio frequency transparent epoxy resin to ensure that the tag achieves an IP68 waterproof and dustproof rating, thus guaranteeing stable read / write conditions in environments with water mist, coal dust, and oil stains. The buffer protective layer wraps around the outside of the radio frequency transparent epoxy resin protective layer. The buffer protective layer adopts a composite structure of polyurethane foam and sparse magnets, which effectively absorbs mechanical vibration and enhances the adhesion of the tag, providing safer protection for the tag. Among them, the polyurethane foam structure provides good cushioning performance and can absorb mechanical vibration; the magnet assembly is used to enhance adhesion stability and reduce the probability of tail rope entanglement through magnetic field repulsion. The weak repulsive magnetic field formed between the magnets helps prevent localized stacking caused by rope entanglement; the protective layer and the high-viscosity silicone layer are respectively bonded to opposite sides of the buffer protective layer; the protective layer is made of wear-resistant and corrosion-resistant material, which is made of Kevlar or Teflon coating, thus effectively enhancing the overall tagging system's wear resistance and chemical corrosion resistance, adapting to the requirements of long-term, high-intensity use in underground drilling, and effectively resisting damage to the tags caused by rope friction; the high-viscosity silicone layer uses a friction coefficient of... This industrial-grade high-viscosity silicone binds tightly to metal parts or ropes, effectively preventing slippage or displacement when firmly bonded. It is particularly suitable for oily and damp surfaces. The high-viscosity silicone layer ensures stability during adhesion, facilitating the creation of accurate standardized unhooking gradient and entanglement state databases initially, and enabling more precise monitoring data during online monitoring. This composite structure effectively adapts to the high humidity, dust, oil, and vibration conditions of underground mining environments, significantly improving its stability, adhesion, and resistance to environmental disturbances.

[0097] This passive RFID tag structure has the following advantages: high-frequency, high-density arrays can achieve millimeter-level structural change detection; multi-layer materials work together to ensure adhesion and read / write stability under extreme conditions; the encapsulation structure takes into account both mechanical and electromagnetic performance to ensure signal quality; it is easy to deploy, requires no external power supply, and is suitable for space-constrained downhole equipment structures; and when used with edge intelligence modules, it can build a distributed real-time monitoring network.

[0098] Step Two: Data Acquisition and Benchmark Library Construction;

[0099] B1: Construct a hook-and-unhook gradient database;

[0100] B11: Closed-state baseline acquisition; Set the initial distance between the hook and the crossbar to 0mm, i.e., the hook and crossbar are set to a fully closed state. Use an RFID reader to continuously acquire 100 sets of RSSI matrix data, and obtain the initial coupling matrix through matrix mean filtering. ;

[0101] B12: Unhooking gradient simulation; using a hydraulic traction device to pull the hook, reducing the distance between the hook and the crossbar from... to according to The step size is adjusted gradually, and each gradient represents Degree of decoupling, simulation arrive common Each state has a data collection point; data is collected under each state. The RSSI matrix data was processed by matrix mean filtering to obtain the gradient matrices. ;

[0102] B13: Standardization of the decoupling matrix; construction of a decoupling gradient database by combining the matrices obtained in A11 and A12. ,in, To eliminate individual response differences between different labels, the decoupled gradient database was analyzed. Each matrix in the matrix is ​​normalized using gradient matrix processing to obtain a decoupling normalized matrix. Finally, a standardized decoupling gradient database is obtained. ,in, This serves as the core foundation for subsequent decoupling degree matching and regression estimation;

[0103] B2: Construct a library of tail rope winding states;

[0104] B21: Reference State Acquisition; The tail tether is set to fully extended, and the distance between adjacent tags is the initial reference distance, preferably 10cm. 100 sets of RSSI matrix data are continuously acquired using an RFID reader, and the reference matrix is ​​obtained through matrix mean filtering. ;

[0105] B22: Simulation of different winding states; by changing the state of the tail rope, the distance between adjacent tags is gradually compressed, decreasing each time. , used to simulate common Each entanglement state is sampled. The RSSI matrix data was processed by matrix mean filtering to obtain the corresponding RSSI matrices. ;

[0106] B23: Normalization of the winding matrix; constructing a winding gradient library by combining the matrices obtained in B21 and B22. ,in, To eliminate individual response differences between different labels, the entangled gradient library was modified. Each matrix in the matrix is ​​normalized by gradient matrix to obtain the entanglement normalized matrix. Finally, a standardized entanglement state database is obtained. ,in, ;

[0107] By constructing the two types of standard databases mentioned above, the collected RSSI matrix can be matched with the samples in the standard library for similarity and state estimation during the real-time monitoring phase. This enables high-precision identification of the hook unhooking state and tail rope entanglement state of the hoist, accurately determining the current status of the hook and tail rope, and achieving comprehensive perception and assessment of the hoist's operational safety.

[0108] Step 3: Obtain monitoring data online;

[0109] By polling and sampling the dense tag array at fixed intervals using an RFID reader, the real-time RSSI matrix of the decoupling tag is obtained. and the entanglement of real-time RSSI matrix Simultaneously, strain gauge 1 and strain gauge 2 are used to acquire uncoupling strain signals and winding strain signals in real time and send them to the edge computing module. The edge computing module obtains uncoupling strain data and winding strain data based on the uncoupling strain signals and winding strain signals, and obtains the strain change rate. ;

[0110] Specifically, the sampling period is set to 200 milliseconds. In each period, a set of RSSI data is obtained by scanning multiple UHF passive RFID tags installed on the hook and crossbar. Each set of RSSI data contains the RSSI values ​​between multiple UHF passive RFID tags, forming a real-time RSSI matrix for hook release. By scanning multiple UHF passive RFID tags mounted on the tail rope, a set of RSSI data is obtained. Each set of RSSI data contains the RSSI values ​​between multiple UHF passive RFID tags, forming a real-time RSSI matrix. The matrix above records the signal strength information between all tags at the current moment, which is the basic data for subsequent processing.

[0111] Step 4: Denoise the monitoring data;

[0112] To improve the accuracy and robustness of subsequent recognition, noise suppression processing is required for the real-time matrix. The entire denoising process includes two stages: median filtering and deep learning (GAN) denoising. First, median filtering is used for preliminary smoothing and denoising, and then a generative adversarial network is introduced for secondary deep denoising to obtain the denoised RSSI matrix.

[0113] Step 5: Data decoupling;

[0114] Due to spatial proximity effects and multipath channel interference among tags, the elements in the original RSSI matrix are not completely independent, exhibiting significant inter-tag coupling. Directly used RSSI data will contain interference information, severely affecting the accurate determination of the tether entanglement state and failing to accurately reflect the actual signal strength relationship between tags. Therefore, a least-squares decoupling algorithm is needed, aiming to minimize the difference between actual measurements and the predictions of the coupling model.

[0115] From the decoupled real-time RSSI matrix after denoising The coupling interference between labels is separated to obtain an independent coupling matrix C, which serves as the basis for subsequent feature recognition.

[0116] Step Six: Identify the tangling state of the tail rope; To achieve automatic identification of the tangling state of the tail rope, this invention constructs a random forest classifier to analyze the feature vector. Multi-class classification is performed. Random forest is an ensemble learning method that improves classification accuracy and stability by constructing multiple decision trees and combining their predictions. In tail rope entanglement monitoring, the random forest algorithm utilizes data from the RSSI matrix... The feature vectors extracted are used to classify the entanglement state of the tail rope.

[0117] F1: A classification model is built using the Random Forest classification algorithm and pre-trained to obtain a classification model for the tail rope entanglement state.

[0118] F2: From the denoised, wrapped real-time RSSI matrix Extracting feature vectors These feature vectors contain coupling information between RFID tags on the tail tether;

[0119] F3: Feature vectors The data is input into the tail rope entanglement classification model to classify the tail rope entanglement state and output the tail rope entanglement state φ, which includes three types: normal, slight entanglement and severe entanglement.

[0120] Step 7: Determine the degree of decoupling based on the independent coupling matrix;

[0121] G1: Calculate the independent coupling matrix C and the normalized decoupling gradient database sequentially using the cosine similarity matching algorithm. Find the cosine similarity of all matrices in the given matrix and identify the matrix with the highest similarity. and the matrix The corresponding degree of decoupling is used as the initial estimate of the degree of decoupling. This step utilizes a pre-built decoupling gradient database to initially determine the degree of decoupling by matching the current matrix with the matrices in the database.

[0122] G2: Combining distance transformation and Gaussian process regression to correct the initial decoupling estimate Obtain the final estimate of the degree of decoupling ;

[0123] Step 8: To achieve a controllable and tiered security response strategy, based on the estimated degree of decoupling... The combined state of the tail rope entanglement φ is used to classify and warn of early warning, and the strain rate of change is used to determine the early warning. Redundancy verification of tiered early warning is performed.

[0124] As a preferred embodiment, in steps B11 and B12 of step two, matrix mean filtering is performed according to formula (1), and a 10×10 matrix is ​​constructed. In step B13 of step two, the decoupling normalization matrix is ​​obtained according to formula (2). ;

[0125] (1);

[0126] In the formula, Indicates the first In the second sampling, the first With the Signal strength between tags;

[0127] (2);

[0128] In the formula, For matrix The mean; For matrix Standard deviation; ;

[0129] As a preferred option, in steps B21 and B22 of step two, matrix mean filtering is performed according to formula (3), and a matrix is ​​constructed. In step B13 of step two, the winding normalization matrix is ​​obtained according to formula (4). ;

[0130] (3);

[0131] (4);

[0132] In the formula, For matrix The mean; For matrix standard deviation .

[0133] As a preferred option, in step three, the strain rate of change is obtained according to formula (5). ;

[0134] (5);

[0135] In the formula, for Real-time response data.

[0136] As a preferred option, the noise reduction process for the monitoring data in step four is as follows:

[0137] D1: Initial smoothing and denoising are performed using median filtering. Median filtering is a non-linear filtering technique that effectively removes abrupt spikes, interference values, and isolated signal jumps in the data. It is a suitable non-linear denoising method for RFID RSSI data. Median filtering of the real-time RSSI matrix can effectively eliminate the impact of environmental noise and occasional interference on the RSSI data. For each element in the matrix... According to formula (6), the median of itself and its adjacent elements is taken as the filtered value. Construct the filtered RSSI matrix ;

[0138] (6);

[0139] In the formula: and They are respectively Two adjacent elements;

[0140] D2: Due to the complex downhole environment, RSSI signals are also subject to nonlinear interference such as multipath effects, metal reflections, and mechanical vibrations. To further improve the quality of RSSI data, a Generative Adversarial Network (GAN) is introduced into the median filtering process to further improve the quality of the filtered RSSI matrix. Deep denoising is performed to obtain a more accurate RSSI matrix. A Generative Adversarial Network (GAN) consists of two parts: a generator and a discriminator. The generator is responsible for generating denoised RSSI data that approximates real RSSI data based on the input noisy data, while the discriminator distinguishes between the real RSSI data and the denoised data generated by the generator, outputting a probability value. Through adversarial training between the two, the generator can generate high-quality denoised RSSI data, thereby improving the accuracy of the data.

[0141] D21: Building Generators and discriminator And obtain the discriminator according to formula (7). loss function The generator is obtained according to formula (8). loss function ;

[0142] (7);

[0143] (8);

[0144] In the formula, Represents the preprocessed RSSI matrix Distribution; Representing noise data Distribution;

[0145] D22: Using matrices As the training set, by minimizing the loss function and loss function Training Generator and discriminator The generator is designed to produce denoised data that is difficult to distinguish from real RSSI data. The training process is as follows: First, the RSSI matrix obtained from the data preprocessing step is used... As a training set, corresponding noisy data is generated. Next, build the generator. and discriminator generator Receive noise data and the preprocessed RSSI matrix Output the denoised RSSI matrix Discriminator Receive the preprocessed RSSI matrix or RSSI matrix generated by the generator Output a probability value , which represents the probability that the input RSSI matrix is ​​real data.

[0146] D3: Using a trained generator For real-time RSSI matrix Denoising is performed, and the denoised RSSI matrix is ​​obtained according to formula (9). This matrix is ​​more stable and has greater physical meaning than the original matrix, which can significantly enhance the reliability of subsequent state recognition.

[0147] (9).

[0148] Through the above steps, the GAN denoising algorithm can effectively remove the RSSI matrix. This reduces noise in the data, improving its accuracy and reliability, and providing a cleaner data foundation for subsequent calculations of decoupling and entanglement.

[0149] As a preferred embodiment, in step five, the independent coupling matrix is ​​obtained by minimizing the objective function according to formula (10). ;

[0150] (10).

[0151] This optimization problem can be solved using numerical optimization methods, such as gradient descent or Newton's method. The resulting matrix... This is the decoupled independent coupling matrix, which eliminates mutual interference between tags and more accurately reflects the actual signal strength relationship between each tag and other tags. This step is a crucial part of the entire monitoring system, providing a clean data foundation for subsequent decoupling degree calculations.

[0152] As a preferred option, in step F1 of step six, the process of obtaining the tail rope entanglement state classification model is as follows:

[0153] F11: Using multiple decision trees A random forest classification model is constructed; the random forest model outputs the final classification result through a voting mechanism of multiple decision trees; the random forest algorithm has the advantages of handling high-dimensional features, strong noise resistance, and good scalability, and is suitable for nonlinear classification tasks in complex downhole environments; the random forest algorithm can effectively handle complex feature relationships and improve the accuracy of entangled state classification.

[0154] F12: Standardize the winding state database Transform into the corresponding training feature set ,in, Using training feature sets The random forest classification model was trained, and the tail rope entanglement state classification model was obtained after training.

[0155] In step six, F2, the denoised, wrapped real-time RSSI matrix is ​​calculated according to formula (11). Obtaining feature vectors ; Eigenvector Used to characterize the entanglement state of the tail rope, this feature vector reflects the trend of signal coupling changes between adjacent tags and can sensitively capture the physical changes caused by local bending or compression of the tail rope.

[0156] (11);

[0157] In step F3 of step six, the entanglement status label of the tail rope is obtained using a voting mechanism according to formula (12). ; The three entanglement states φ are: normal (tail rope straight, tag spacing uniform, RSSI trend stable); slight entanglement (local tag spacing contracted, coupling strength slightly increasing trend); and severe entanglement (multiple tags close together or overlapping, signal coupling concentrated and enhanced, matrix shift significant). This identification result will be used to assist in the subsequent decoupling early warning mechanism and will be input as an independent dimension into the regression model for joint state estimation.

[0158] (12);

[0159] In the formula, Indicates the first Each decision tree is used to process the input feature vector. The classification results; This indicates the mode operation, which selects the classification result that appears most frequently as the final output. Using the random forest classification algorithm, this invention can accurately identify the tangling state of the tail rope, providing strong protection for the safe operation of mine hoists.

[0160] Cosine similarity is a measure of the similarity in direction between two vectors; the closer the value is to 1, the more similar the two vectors are. For matrices, the cosine similarity can be calculated after flattening the matrix into vectors. Specifically, in step G1 of step seven, the initial decoupling degree estimate... The process is as follows:

[0161] G11: Calculate the cosine similarity according to formula (13) ;

[0162] (13);

[0163] G12: Select the one with the highest similarity according to formula (14) The corresponding gradient is used as an estimate of the initial degree of decoupling. ,in, ;

[0164] (14).

[0165] As a preferred option, in step G2 of step seven, in order to further improve the accuracy of the decoupling degree estimation, it is necessary to adjust the initial decoupling degree. Make corrections to adjust the initial degree of decoupling. The estimation process is as follows:

[0166] G21: Independent coupling matrix Converted to a relative distance matrix between the hook and the crossbar; based on a pre-calibrated RSSI-distance mapping function. This function describes the relationship between RSSI values ​​and actual distances, and the matrix is ​​calculated according to formula (15). Each element in Convert to the corresponding relative distance ;

[0167] (15);

[0168] G22: Calculate the average of all relative distances according to formula (16). ;

[0169] (16);

[0170] G22: To further improve the accuracy of decoupling estimation, a system based on... The two-dimensional input vector is input into the pre-trained Gaussian process regression model. And according to formula (17), the initial decoupling degree After making corrections, we obtain the corrected estimate of the degree of decoupling. Gaussian process regression is a probability-based regression method that can predict new data points based on existing data points, and it also introduces the uncertainty of the prediction. In this scenario, the Gaussian process regression model is based on... and The relationship between them corrects the initial degree of decoupling.

[0171] (17).

[0172] Corrected estimate of decoupling degree It takes into account both matrix matching results and actual distance information, resulting in higher accuracy.

[0173] As a preferred option, the tiered early warning process in step eight is as follows:

[0174] if ,and If the indicator is normal (and continues for 3 detection cycles, each detection cycle is 200ms), then it is initially determined that the hook and tail rope are in normal condition, and further verification is needed. ,like If no significant strain changes are detected, it indicates that the system is currently in normal operation and no warning action is required.

[0175] if ,and If the indication is slight tangling, it can be initially determined that the hook is in normal condition and the tail rope is slightly tangled. Further verification is needed. ,like ,and If the temperature continues to rise, it confirms that the hook is in normal condition but the tail rope is slightly tangled, triggering a Level 1 warning to remind the operator to check the tail rope. A warning message is issued stating that a slight tangling has been detected in the tail rope and requesting the operator to check for any minor tangling. Simultaneously, the current status is displayed. value;

[0176] if ,and If the indicator shows severe tangling, the initial assessment is that the hook is in normal condition but the tail rope is severely tangled. Further verification is needed. ,like ,and If the hoist continues to rise, it confirms that the hook is in normal condition but the tail rope is severely tangled, triggering an emergency warning and stopping the hoist to prevent accidents caused by tail rope entanglement. Simultaneously, a warning message is issued: "Severe tail rope entanglement detected, machine stopped, please immediately check for severe tail rope entanglement." The current status is also displayed. value;

[0177] if (And it lasts for 3 detection cycles, each detection cycle is 200ms), and If the indicator is normal, the initial judgment is that the hook has slightly come loose and the tail rope is in normal condition. Further verification is needed. ,like ,and If the temperature continues to rise, it confirms that the hook has slightly come undone and the tail rope is in normal condition, triggering a Level 1 warning. Simultaneously, a message is issued requesting a check for any partial detachment of the hook, along with the current status of the warning. value;

[0178] if (And it lasts for 3 detection cycles, each detection cycle is 200ms), and If the indicator shows slight entanglement (and this continues for 3 detection cycles, each lasting 200ms), it can be preliminarily determined that the hook has slightly come off and the tail rope is slightly entangled. Further verification is needed. ,like ,and and If both values ​​continue to rise, and a double anomaly is detected, it confirms a slight hook detachment and a slight entanglement of the tail rope, triggering a Level 1 warning. Simultaneously, a warning message is issued urging the user to check for potential problems with the hook and tail rope. The current status is also displayed. Value and value;

[0179] if (And it lasts for 3 detection cycles, each detection cycle is 200ms), and If the indicator shows severe tangling, it initially suggests that the hook has slightly come loose and the tail rope is severely tangled, requiring further verification. ,like ,and If the hoist continues to rise, it will confirm a slight hook detachment and severe tail rope entanglement, triggering an emergency warning and stopping the hoist. Simultaneously, it will issue a warning message stating that a slight hook detachment and severe tail rope entanglement have been detected, the hoist has been stopped, and the severe tail rope entanglement should be checked first. At the same time, it will display the current status... Value and value;

[0180] if ,and If the indicator is normal, the initial judgment is that the hook has seriously come off the hook, but the tail rope is in normal condition. Further verification is needed. ,like ,and A sudden rise confirms severe hook detachment, while the tail rope remains in normal condition. This triggers an emergency warning and stops the hoist to prevent hook detachment. Simultaneously, a warning message is issued: "Severe hook detachment detected; machine stopped; please immediately check hook detachment." The current status is also displayed. Value and value;

[0181] if ,and If the indication is slight entanglement (and continues for 3 detection cycles, each detection cycle being 200ms), then the initial judgment is that the hook is severely unhooked and the tail rope is slightly entangled, requiring further verification. ,like ,and and If both conditions continue to rise, it confirms a severe hook detachment and slight entanglement of the tail rope, triggering an emergency warning and stopping the hoist. Simultaneously, a message is issued indicating that a severe hook detachment and slight tail rope entanglement have been detected, the hoist has been stopped, and priority should be given to checking for the severe hook detachment. The current status is also displayed. value;

[0182] if ,and If the indicator shows severe tangling, it initially suggests that the hook is severely unhooked and the tail rope is severely tangled, requiring further verification. ,like ,and and If both conditions continue to rise, it confirms a severe hook detachment and severe tail rope entanglement, triggering an emergency warning and stopping the hoist. Simultaneously, a warning message is issued: "Severe hook detachment and severe tail rope entanglement detected; hoist stopped; please immediately check for these conditions." The current status is also displayed. Value and value.

[0183] This invention discloses an RFID-based method for monitoring hook disengagement and tail rope entanglement in coal mine hoists. First, multiple UHF passive RFID tags are arranged on the hook and crossbar to form a two-dimensional sensing coupling array. Multiple UHF passive RFID tags are arranged on each tail rope to form a linear array. This dense tag array, formed by the two-dimensional sensing coupling array and the linear array, effectively covers all areas to be monitored, ensuring comprehensive monitoring coverage. By placing strain gauges at the hook contact interface and in the tail rope channel, strain data can be effectively detected simultaneously with RSSI data reading when mechanical deformation occurs between the hook and crossbar or when the tail rope deforms. Thus, strain data can be used as a redundancy verification method, effectively ensuring the accuracy and reliability of monitoring and avoiding false alarms. Second, a standardized disengagement gradient database is established by simulating different disengagement states using a hydraulic device. This not only effectively reduces the database construction cost and cycle but also facilitates the comparison of the matching degree of real-time matrices using a cosine similarity algorithm during subsequent disengagement degree matching, which is beneficial for quantitative judgment of the disengagement degree. In establishing a standardized decoupling gradient database, matrix mean filtering is used for denoising, eliminating environmental noise and significantly improving RSSI data quality, thus enabling the establishment of a more accurate baseline coupling matrix. Simultaneously, establishing a standardized winding state database by simulating different degrees of tail rope winding not only effectively reduces database construction costs and time but also provides accurate basic data for subsequent tail rope state classification. Furthermore, during online monitoring, polling and sampling the dense tag array at a fixed period effectively reduces signal overlap in dense tag environments, lowers the false read rate, and improves overall recognition efficiency and accuracy. During online monitoring, strain change rate is simultaneously obtained. This provides a reliable basis for subsequent redundancy verification, helping to ensure the accuracy of subsequent early warnings. Then, a dual denoising mechanism of median filtering + GAN denoising is applied to the RSSI matrix, which not only eliminates abrupt noise but also effectively filters nonlinear background interference, thereby obtaining high-quality RSSI feature data and providing a stable data foundation for distinguishing between hook unhooking and tail rope entanglement. Furthermore, a least-squares decoupling algorithm is used to separate independent signals from the mutually coupled tag signals, obtaining an independent coupling matrix. This reveals the actual signal relationship between each tag, providing reliable data support for determining whether the hoist hook has unhooked and whether the tail rope is knotted. Then, a random forest algorithm is used to construct a tail rope entanglement state classification model, which can achieve automatic, efficient, and accurate classification of the tail rope state based on feature vectors, thus significantly improving the accuracy and stability of classification. Finally, an initial unhooking degree estimate is obtained through cosine similarity matching, and then the final unhooking degree is obtained by combining distance transformation and Gaussian process regression correction, which can efficiently and accurately quantify the unhooking degree of the hoist hook. Finally, in the graded early warning process, graded early warnings are carried out based on the combined state of the estimated degree of decoupling and the state of the tail rope entanglement. The redundancy verification of graded early warnings is carried out by the strain change rate, which significantly improves the reliability and accuracy of the early warning and avoids false alarms.

[0184] This method boasts advantages such as high intelligence, high recognition accuracy, high reliability, and adaptability to complex environments. It effectively combines cosine similarity matching, Gaussian regression correction, multi-threshold control, and redundant physical quantity fusion technologies. It not only enables quantitative assessment of the hoist's unhooking state and automatic, accurate classification of the tail rope state, but also features intelligent early warning control, allowing for dynamic response and rapid shutdown under different early warning conditions. Compared to traditional single-point strain gauge structural monitoring methods, this invention, based on a dense RFID array and machine learning algorithms, offers significant advantages in robustness, accuracy, and applicability. It is suitable for safety monitoring and control of critical connection structures in intelligent mines, high-risk mechanical structures, and port machinery.

Claims

1. A smart monitoring method for hook unhooking and tail rope entanglement in coal mine hoists based on RFID, characterized in that, Includes the following steps: Step 1: Install the intelligent monitoring system; Multiple passive RFID tags are installed at the junction of the hook and the crossbar to form a two-dimensional sensing coupling array; multiple passive RFID tags are installed on each tail rope to form a linear array; a dense tag array is constructed through the two-dimensional sensing coupling array and the linear array; strain gauges are installed on the contact surface between the hook and the crossbar, and strain gauges are installed in the tail rope channel; at the same time, an explosion-proof RFID reader is installed above the hook, an edge computing module is installed near the hoist, a controller is installed in the monitoring center, and communication connections are established between the edge computing module and the dense tag array, strain gauges, and strain gauges; a communication connection is also established between the edge computing module and the controller. Step Two: Data Acquisition and Benchmark Library Construction; B1: Construct a hook-and-unhook gradient database; B11: Closed-state reference acquisition; Set the initial distance between the hook and the crossbar to 0mm, continuously acquire 100 sets of RSSI matrix data using an RFID reader, and obtain the initial coupling matrix through matrix mean filtering. ; B12: Unhooking gradient simulation; using a hydraulic traction device to pull the hook, reducing the distance between the hook and the crossbar from... to according to The step size is adjusted gradually, and each gradient represents Degree of decoupling, simulation arrive common Each state has a data collection point; data is collected under each state. The RSSI matrix data was processed by matrix mean filtering to obtain the gradient matrices. ; B13: Standardization of the decoupling matrix; construction of a decoupling gradient database by combining the matrices obtained in A11 and A12. ,in, ; for decoupling gradient database Each matrix in the matrix is ​​normalized using gradient matrix processing to obtain a decoupling normalized matrix. Finally, a standardized decoupling gradient database is obtained. ,in, ; B2: Construct a library of tail rope winding states; B21: Reference State Acquisition; The tail tether is set to fully extended, the distance between adjacent tags is the initial reference distance, 100 sets of RSSI matrix data are continuously acquired using an RFID reader, and the reference matrix is ​​obtained through matrix mean filtering. ; B22: Simulation of different winding states; by changing the state of the tail rope, the distance between adjacent tags is gradually compressed, decreasing each time. , used to simulate common Each entanglement state is sampled. The RSSI matrix data was processed by matrix mean filtering to obtain the corresponding RSSI matrices. ; B23: Normalization of the winding matrix; constructing a winding gradient library by combining the matrices obtained in B21 and B22. ,in, ; For the winding gradient library Each matrix in the matrix is ​​normalized by gradient matrix to obtain the entanglement normalized matrix. Finally, a standardized entanglement state database is obtained. ,in, ; Step 3: Obtain monitoring data online; By polling and sampling the dense tag array at fixed intervals using an RFID reader, the real-time RSSI matrix of the decoupling tag is obtained. and the entanglement of real-time RSSI matrix Simultaneously, strain gauge 1 and strain gauge 2 are used to acquire uncoupling strain signals and winding strain signals in real time and send them to the edge computing module. The edge computing module obtains uncoupling strain data and winding strain data based on the uncoupling strain signals and winding strain signals, and obtains the strain change rate. ; Step 4: Denoise the monitoring data; First, median filtering is used for preliminary smoothing and denoising. Then, a generative adversarial network is introduced for secondary deep denoising to obtain the denoised RSSI matrix. Step 5: Data decoupling; The least squares decoupling algorithm is used to decouple the real-time RSSI matrix after denoising. The coupling interference between tags is separated to obtain the independent coupling matrix C; Step Six: Identify the tangling state of the tail rope; F1: A classification model is constructed using the random forest classification algorithm and pre-trained to obtain a classification model for the tail rope entanglement state. F2: From the denoised, wrapped real-time RSSI matrix Extracting feature vectors ; F3: Feature vectors The data is input into the tail rope entanglement classification model to classify the tail rope entanglement state and output the tail rope entanglement state φ, which includes three types: normal, slight entanglement and severe entanglement. Step 7: Determine the degree of decoupling based on the independent coupling matrix; G1: Calculate the independent coupling matrix C and the normalized decoupling gradient database sequentially using the cosine similarity matching algorithm. Find the cosine similarity of all matrices in the given matrix and identify the matrix with the highest similarity. and the matrix The corresponding degree of decoupling is used as the initial estimate of the degree of decoupling. ; G2: Combining distance transformation and Gaussian process regression to correct the initial decoupling estimate Obtain the final estimate of the degree of decoupling ; Step 8: Based on the estimated degree of decoupling The combined state of the tail rope entanglement φ is used to classify and warn of early warning, and the strain rate of change is used to determine the early warning. Redundancy verification of tiered early warning is performed.

2. The intelligent monitoring method for hook unhooking and tail rope entanglement of a coal mine hoist based on RFID according to claim 1, characterized in that, In step one, multiple passive RFID tags are evenly divided into two groups, with 10 UHF passive RFID tags in each group. The 10 UHF passive RFID tags in the first group are installed on the hook with a spacing of ≤5 cm, and the 10 UHF passive RFID tags in the second group are installed on the crossbar with a spacing of ≤5 cm, forming... Two-dimensional sensing coupling array; 10 passive RFID tags are installed longitudinally on the tail rope with a spacing of ≤10 cm; the RFID reader is located 1 meter above the hook; The passive RFID tags 1 and 2 have the same structure, both including a UHF passive RFID tag body, an RF transparent epoxy resin protective layer, a buffer protective layer, a protective layer, and a high-viscosity silicone layer. The RF transparent epoxy resin protective layer wraps around the outside of the UHF passive RFID tag body; the buffer protective layer wraps around the outside of the RF transparent epoxy resin protective layer; the buffer protective layer adopts a composite structure of polyurethane foam and sparse magnets; the protective layer and the high-viscosity silicone layer are respectively bonded to opposite sides of the buffer protective layer; the protective layer is made of wear-resistant and corrosion-resistant material; the high-viscosity silicone layer has a low coefficient of friction. Industrial-grade high-viscosity silicone.

3. The intelligent monitoring method for hook unhooking and tail rope entanglement of a coal mine hoist based on RFID according to claim 1, characterized in that, In steps B11 and B12 of step two, matrix mean filtering is performed according to formula (1), and a matrix is ​​constructed. In step B13 of step two, the decoupling normalization matrix is ​​obtained according to formula (2). ; （1）； In the formula, Indicates the first In the second sampling, the first With the Signal strength between tags; （2）； In the formula, For matrix The mean; For matrix Standard deviation; ; In steps B21 and B22 of step two, matrix mean filtering is performed according to formula (3), and a matrix is ​​constructed. In step B13 of step two, the winding normalization matrix is ​​obtained according to formula (4). ; （3）； （4）； In the formula, For matrix The mean; For matrix standard deviation .

4. The intelligent monitoring method for hook unhooking and tail rope entanglement of a coal mine hoist based on RFID according to claim 1, characterized in that, In step three, the strain rate of change is obtained according to formula (5). ; （5）； In the formula, for Real-time response data.

5. A method for intelligent monitoring of hook unhooking and tail rope entanglement in a coal mine hoist based on RFID, as described in claim 1, characterized in that, In step four, the noise reduction process for the monitoring data is as follows: D1: First, use median filtering for initial smoothing and noise reduction; for each element of the matrix... According to formula (6), the median of itself and its adjacent elements is taken as the filtered value. Construct the filtered RSSI matrix ; （6）； In the formula: and They are respectively Two adjacent elements; D2: Introducing a generative adversarial network to filter the RSSI matrix Deep denoising is performed to obtain a more accurate RSSI matrix. ; D21: Building Generators and discriminator And obtain the discriminator according to formula (7). loss function The generator is obtained according to formula (8). loss function ; （7）； （8）； In the formula, Represents the preprocessed RSSI matrix Distribution; Representing noise data Distribution; D22: Using matrices As the training set, by minimizing the loss function and loss function Training Generator and discriminator ; D3: Using a trained generator For real-time RSSI matrix Denoising is performed, and the denoised RSSI matrix is ​​obtained according to formula (9). ; （9） 。 6. The intelligent monitoring method for hook unhooking and tail rope entanglement of a coal mine hoist based on RFID according to claim 1, characterized in that, In step five, the independent coupling matrix is ​​obtained by minimizing the objective function according to formula (10). ; （10） 。 7. A method for intelligent monitoring of hook unhooking and tail rope entanglement in a coal mine hoist based on RFID, as described in claim 1, characterized in that, In step six, F1, the process of obtaining the tail rope entanglement state classification model is as follows: F11: Using multiple decision trees Construct a random forest classification model; F12: Standardize the winding state database Transform into the corresponding training feature set ,in, Using training feature sets The random forest classification model was trained, and the tail rope entanglement state classification model was obtained after training. In step six, F2, the denoised, wrapped real-time RSSI matrix is ​​calculated according to formula (11). Obtaining feature vectors ; （11）； In step F3 of step six, the entanglement status label of the tail rope is obtained using a voting mechanism according to formula (12). , The three winding states φ correspond to normal, slight winding, and severe winding, respectively; （12）； In the formula, Indicates the first Each decision tree is used to process the input feature vector. The classification results; This indicates the operation of taking the mode.

8. A method for intelligent monitoring of hook unhooking and tail rope entanglement in a coal mine hoist based on RFID, as described in claim 1, characterized in that, In step seven, G1, the initial decoupling degree estimate The process is as follows: G11: Calculate the cosine similarity according to formula (13) ; （13）； G 12: Select the one with the highest similarity according to formula (14) The corresponding gradient is used as an estimate of the initial degree of decoupling. ,in, ; （14） 。 9. A method for intelligent monitoring of hook unhooking and tail rope entanglement in a coal mine hoist based on RFID, as described in claim 1, characterized in that, In step seven, G2, the initial degree of decoupling is corrected. The estimation process is as follows: G21: Independent coupling matrix Converted to a relative distance matrix between the hook and the crossbar; based on a pre-calibrated RSSI-distance mapping function. According to formula (15), the matrix Each element in Convert to the corresponding relative distance ; （15）； G22: Calculate the average of all relative distances according to formula (16). ; （16）； G22: Building based on The two-dimensional input vector is input into the pre-trained Gaussian process regression model. And according to formula (17), the initial decoupling degree After making corrections, we obtain the corrected estimate of the degree of decoupling. ; （17） 。 10. A method for intelligent monitoring of hook unhooking and tail rope entanglement in a coal mine hoist based on RFID, as described in claim 1, characterized in that, In step eight, the tiered early warning process is as follows: if ,and If the indicator is normal, it can be initially determined that the hook and tail rope are in normal condition, and further verification is needed. ,like If no significant strain changes are detected, it indicates that the system is currently in normal operation and no warning action is required. if ,and If the indication is slight tangling, it can be initially determined that the hook is in normal condition and the tail rope is slightly tangled. Further verification is needed. ,like ,and If the temperature continues to rise, it confirms that the hook is in normal condition but the tail rope is slightly tangled, triggering a Level 1 warning. This warning sends a message urging the user to check the tail rope for any minor tangling. Simultaneously, it displays the current status... value; if ,and If the indicator shows severe tangling, the initial assessment is that the hook is in normal condition but the tail rope is severely tangled. Further verification is needed. ,like ,and If the hoist continues to rise, it confirms that the hook is in normal condition but the tail rope is severely tangled, triggering an emergency warning and stopping the hoist. Simultaneously, it issues a warning message: "Severe tail rope tangling detected; machine stopped; please immediately check for severe tail rope tangling." It also displays the current status... value; if ,and If the indicator is normal, the initial judgment is that the hook has slightly come loose and the tail rope is in normal condition. Further verification is needed. ,like ,and If the temperature continues to rise, it confirms that the hook has slightly come undone and the tail rope is in normal condition, triggering a Level 1 warning. Simultaneously, a message is issued requesting a check for any partial detachment of the hook, along with the current status of the warning. value; if ,and If the indicator shows slight entanglement, it initially suggests that the hook has slightly come off and the tail rope is slightly entangled; further verification is needed. ,like ,and and If both indicators continue to rise, it confirms a slight hook detachment and slight tail rope tangling, triggering a Level 1 warning. Simultaneously, a warning message is issued urging the user to check for potential issues with the hook and tail rope. The current status is also displayed. Value and value; if ,and If the indicator shows severe tangling, it initially suggests that the hook has slightly come loose and the tail rope is severely tangled, requiring further verification. ,like ,and If the hoist continues to rise, it will confirm a slight hook detachment and severe tail rope entanglement, triggering an emergency warning and stopping the hoist. Simultaneously, it will issue a warning message stating that a slight hook detachment and severe tail rope entanglement have been detected, the hoist has been stopped, and the severe tail rope entanglement should be checked first. At the same time, it will display the current status... Value and value; if ,and If the indicator is normal, the initial judgment is that the hook has seriously come off the hook, but the tail rope is in normal condition. Further verification is needed. ,like ,and A sharp rise indicates a confirmed serious hook detachment, while the tail rope remains in normal condition. This triggers an emergency warning and shuts down the hoist. Simultaneously, a message is issued urging immediate inspection of the hook detachment, stating that a serious hook detachment has been detected and the hoist has been stopped. The current status is also displayed. Value and value; if ,and If the indication is slight entanglement, the initial judgment is that the hook is severely unhooked and the tail rope is slightly entangled. Further verification is needed. ,like ,and and If both conditions continue to rise, it confirms a severe hook detachment and slight entanglement of the tail rope, triggering an emergency warning and stopping the hoist. Simultaneously, a message is issued indicating that a severe hook detachment and slight tail rope entanglement have been detected, the hoist has been stopped, and priority should be given to checking for the severe hook detachment. The current status is also displayed. value; if ,and If the indicator shows severe tangling, it initially suggests that the hook is severely unhooked and the tail rope is severely tangled, requiring further verification. ,like ,and and If both conditions continue to rise, it confirms a severe hook detachment and severe tail rope entanglement, triggering an emergency warning and stopping the hoist. Simultaneously, a warning message is issued: "Severe hook detachment and severe tail rope entanglement detected; hoist stopped; please immediately check for these conditions." The current status is also displayed. Value and value.

Images