Crane safety monitoring system and method

Abstract

The application provides a crane safety monitoring system and method, the system comprising: an airborne monitoring terminal for real-time acquisition of multi-dimensional operation data of the crane; an operation and maintenance monitoring platform comprising an operation monitoring module for fusion processing of the pre-processed multi-dimensional operation data by using a Kalman filtering algorithm to obtain real-time operation state estimation; and an updated model is established for each movement mechanism of the crane to calculate a theoretical operation state; an alarm signal of the corresponding movement mechanism is generated by comparing the real-time operation state estimation with the theoretical operation state, and a historical operation database is constructed; a warning prediction model is trained by using the historical operation data to predict a fault risk probability and generate a warning signal of the corresponding movement mechanism; the system realizes equipment state monitoring, fault warning, safety protection and management by comprehensively monitoring and analyzing the operation parameters, energy consumption data and operation behavior of the crane, and effectively improves the operation safety and operation and maintenance efficiency of the crane.

Description

Technical Field

[0001] This invention relates to the field of crane safety monitoring technology, and in particular to a crane safety monitoring system and method. Background Technology

[0002] As ultra-large special lifting equipment in core sectors of the national economy such as energy and power, port logistics, and heavy manufacturing, the operational safety and reliability of cranes directly affect the construction efficiency and production safety of major projects. Taking offshore wind power installation, large bridge construction, and modern port hubs as examples, the heavy-duty cranes used often have a rated lifting capacity of over a thousand tons, a boom length of over a hundred meters, and a self-weight of several thousand tons. Due to their complex structure and enormous load, it is necessary to monitor the health status of the equipment and manage its operational safety.

[0003] Current crane safety monitoring systems suffer from low monitoring coverage and inconsistent categories, resulting in blind spots in safety control. Furthermore, data processing relies on simple threshold judgments and lacks multi-source data fusion and algorithm analysis, making it impossible to accurately predict fault risks, thereby reducing the safety of crane operation and maintenance efficiency. Summary of the Invention

[0004] In view of this, the present invention proposes a crane safety monitoring system and method, which realizes equipment status monitoring, fault early warning, safety protection and management by comprehensively monitoring and analyzing crane operating parameters, energy consumption data and operating behavior, thereby effectively improving the crane's operating safety and maintenance efficiency.

[0005] The technical solution of the present invention is implemented as follows: In a first aspect, the present invention provides a crane safety monitoring system, the system comprising: The airborne monitoring terminal is deployed on the crane to collect multi-dimensional operating data, lubricating oil status data and driver facial image data of the crane in real time. The operation and maintenance monitoring platform is communicatively connected to the airborne monitoring terminal. The platform includes an operation monitoring module, a lubricating oil monitoring module, and a driver behavior monitoring module. The operation monitoring module is used to fuse the preprocessed multidimensional operation data using the Kalman filter algorithm to obtain a real-time operation status estimate; and to establish update models for the motion mechanism of the crane to calculate the theoretical operation status. Based on the comparison between the real-time operating status estimate and the theoretical operating status, alarm signals for the corresponding motion mechanisms are generated, and a historical operating database is constructed. An improved LSTM neural network is used to construct an alarm prediction model, and historical operating data is used to train the alarm prediction model to predict the probability of failure risk and generate early warning signals for the corresponding moving mechanism. The lubricating oil monitoring module is used to establish an oil deterioration model based on the number of times the equipment operates and the lubricating oil status data, calculate the remaining life percentage of the lubricating oil, compare it with the preset life threshold, and generate an oil change warning reminder. The driver behavior monitoring module uses an improved target detection algorithm to identify the driver's facial features and combines them with driver operation behavior data to obtain a driver status assessment result. The module then provides graded warnings based on the driver status assessment result and generates corresponding warning prompts.

[0006] Based on the above technical solutions, preferably, the airborne monitoring terminal collects multi-dimensional operating data, lubricating oil status data and driver facial image data of the crane in real time. The multi-dimensional operating data includes the crane's trolley travel, operating speed, acceleration, lifting height, lifting weight and motor current harmonic parameters, and preprocesses the collected multi-dimensional operating data.

[0007] Based on the above technical solutions, preferably, the operation monitoring module includes: Based on real-time collected data, a state vector is constructed that includes the crane's operating speed, acceleration, and motor current harmonic components. The Kalman filter algorithm is used to fuse the state vector data to obtain the real-time operating state estimate; Establish corresponding updated models for the hoisting mechanism, luffing mechanism and slewing mechanism of the crane respectively, and calculate the theoretical operating state of each mechanism; The real-time operating status estimate is compared with the theoretical operating status in real time. When the real-time operating status estimate is greater than the theoretical operating status, a fault alarm signal for the corresponding mechanism is generated. A historical operation database is constructed based on equipment operating parameters, alarm records, and maintenance history; An alarm prediction model is constructed using an improved LSTM neural network; An alarm prediction model is trained using data from a historical operation database. By analyzing the time series characteristics of historical alarm data, the probability of crane failure in future periods is predicted. When the probability of failure exceeds the preset risk threshold, a warning signal is generated for the corresponding motion mechanism.

[0008] Based on the above technical solutions, preferably, the step of constructing the alarm prediction model using an improved LSTM neural network includes: After the input layer of the LSTM neural network, a multi-scale feature fusion layer is added to dynamically and weightedly fuse the fine-grained temporal features and coarse-grained temporal features to obtain fused features. The fine-grained temporal features are the collected high-dimensional operating parameter vectors; the coarse-grained temporal features are the aggregated alarm status vectors of each motion mechanism in the historical period. An adaptive attention layer is added before the output layer of the LSTM neural network to calculate the attention weights of the hidden states of each motion mechanism of the crane to the global prediction target; and the hidden states of all mechanisms are weighted and summed according to the attention weights to obtain the global context hidden state. Based on the global context hidden state, the output layer of the LSTM neural network is used to generate a prediction of the failure risk probability of the crane in the future time period. An improved focusing loss function is adopted as the optimization objective for model training, and its expression is: ; In the formula, oh t For time decay weight, N For the sample size, y t+τ for t + t The true label of time, where γ is the focusing parameter. for t + t Predicted labels for each moment; Using time-series data from the historical operational database, the model parameters are iteratively optimized with the goal of minimizing the focusing loss function until the model converges, resulting in a well-trained alarm prediction model.

[0009] Based on the above technical solutions, preferably, the lubricating oil monitoring module includes: Real-time data collection of crane lubricating oil's kinematic viscosity, total acid value, and metal abrasive concentration; Based on the monitored values ​​of the kinematic viscosity, total acid value, and metal abrasive concentration of the lubricating oil, and their corresponding initial values ​​and scrap thresholds, the current deterioration rate of each indicator is calculated. The current degradation rate of each indicator is weighted and summed, and then added to the cumulative number of times the equipment has been used to establish an oil degradation model. The overall degradation degree is calculated and converted into the percentage of remaining life of the lubricating oil. The remaining life percentage of the lubricating oil is compared with a preset life threshold, and an oil change warning is generated when the remaining life percentage falls below the preset life threshold.

[0010] Based on the above technical solutions, preferably, the driver behavior monitoring module includes: An improved YOLOv11 target detection algorithm is used to construct a driver state recognition model. The improvement of the driver state recognition model is that its neck network adopts a bidirectional feature pyramid network to replace the original feature pyramid network and path aggregation network structure, so as to achieve efficient fusion of cross-scale features; and an attention module is integrated into the output layer of the bidirectional feature pyramid network to enhance the model's attention to key facial features of the driver by adaptively recalibrating the feature channels. The driver's facial images collected in real time are input into the driver state recognition model, which identifies and outputs the coordinate information of the face, eyes and mouth areas; Based on coordinate information, temporal image sequences of the eye and mouth regions are extracted, and the closing frequency, closing duration, pupil diameter change rate, and yawning frequency of the eyes are calculated to obtain visual information features. The real-time operating handle frequency and operation response time data of the crane control system are acquired and fused with visual information features for weighted calculation to generate the final driver status assessment score. The driver's condition assessment score is compared with multiple preset fatigue thresholds, and corresponding warning prompts are generated based on the level of the threshold exceeded. When the driver's condition assessment score exceeds the first fatigue threshold, it is judged as a low-level warning and a voice reminder is executed; When the driver's condition assessment score exceeds the second fatigue threshold, it is determined to be a medium-level warning, and an audible and visual alarm and seat vibration reminder are activated. When a driver's condition assessment score exceeds the third fatigue threshold, it is considered a high-level warning, and a mandatory rest instruction is sent to the operation and maintenance monitoring platform.

[0011] Based on the above technical solutions, preferably, the operation and maintenance monitoring platform also includes an energy consumption monitoring module, a personnel management module, a wire rope monitoring module, and a load monitoring module. The energy consumption monitoring module is used to collect crane load data in real time, divide the operating conditions according to the real-time load, establish corresponding operating condition energy consumption models, and perform energy consumption statistics and efficiency analysis. The personnel management module is used to verify the identity of operators through biometric technology and to assign function access permissions and data operation permissions according to preset multi-level permission control rules. The wire rope monitoring module is used to monitor the tension distribution and metal cross-sectional area loss of the wire rope in real time through tension sensors and magnetic flux detection methods, and calculate its remaining service life based on the fatigue cumulative damage theory to generate a fracture risk warning. The load monitoring module calculates the allowable load under the current working condition based on the real-time boom posture and compares it with the actual load in real time. When the limit is exceeded, a graded protection response is triggered.

[0012] Based on the above technical solutions, preferably, the energy consumption monitoring module includes: Based on real-time collected load data, the crane's operating status is dynamically divided into multiple operating conditions such as no-load, light-load, medium-load, and heavy-load according to its percentage relative to the rated load, and the cumulative operating time of each condition is recorded. Energy consumption models are established to correlate energy consumption with load and runtime under different operating conditions. Based on each model and the runtime of the corresponding operating conditions, the comprehensive energy consumption of the equipment is calculated. The time utilization rate is calculated based on the actual operation time and the planned operation time; the performance utilization rate is calculated based on the total handling volume and the actual operation time, as well as the rated handling volume and the theoretical optimal cycle time; the quality pass rate is calculated based on the number of safe operation cycles and the total number of operation cycles; and the overall efficiency is calculated based on the product of the time utilization rate, the performance utilization rate, and the quality pass rate. Based on historical data on overall energy consumption and overall efficiency, trend analysis is performed to generate an energy consumption and efficiency analysis report to guide energy-saving operation and maintenance.

[0013] Based on the above technical solutions, preferably, the load monitoring module includes: Real-time acquisition of crane boom attitude parameters and real-time load; Based on the boom attitude parameters, the allowable load under the current working condition is calculated using the tilting moment model, and the expression is: ; In the formula, Q 允许 For the calculated allowable load, K 安全 To preset a safety factor, M 稳定 The maximum stabilizing moment is calculated based on the equipment's own weight and center of gravity position. G 臂架 For the weight of the boom, L 臂重心 The distance from the boom's center of gravity to the center of rotation. L For boom length, i The angle between the boom and the horizontal plane. cosθ A coefficient for converting boom length to horizontal lever arm; The real-time load is compared with the allowable load in real time. When the real-time load reaches the first preset threshold ratio of the allowable load, an audible and visual warning is triggered. When the real-time load reaches the second preset threshold ratio of the allowable load, the crane's dangerous directional movement is automatically restricted. When the real-time load reaches the third preset threshold ratio of the allowable load, the crane's power source is cut off. The third preset threshold ratio is greater than the second preset threshold ratio, which is greater than the first preset threshold ratio.

[0014] Secondly, the present invention also provides a crane safety monitoring method, implemented using a crane safety monitoring system, comprising the following steps: S1 collects multi-dimensional operating data of the crane, lubricating oil status data, and driver's facial image data in real time; S2. The Kalman filter algorithm is used to fuse the preprocessed multidimensional operating data to obtain a real-time operating state estimate; and an update model is established for the motion mechanism of the crane to calculate the theoretical operating state. S3, based on the comparison between the real-time operating status estimate and the theoretical operating status, generates alarm signals for the corresponding motion mechanism and constructs a historical operating database; S4. An improved LSTM neural network is used to build an alarm prediction model, and historical operating data is used to train the alarm prediction model to predict the probability of failure risk and generate early warning signals for the corresponding moving mechanism. S5: Based on the number of times the equipment operates and the lubricating oil status data, establish an oil deterioration model, calculate the remaining life percentage of the lubricating oil, compare it with the preset life threshold, and generate an oil change warning reminder. S6 uses an improved target detection algorithm to identify the driver's facial features and combines them with the driver's operational behavior data to calculate the driver's state assessment result. The driver's state assessment result is then used to issue graded warnings and generate corresponding warning prompts.

[0015] The crane safety monitoring system and method of the present invention have the following advantages over the prior art: (1) By comprehensively monitoring and analyzing the crane's operating parameters, energy consumption data and operating behavior, the equipment status monitoring, fault warning, safety protection and management can be realized, and potential risks can be warned in advance, which can reduce the rate of unplanned downtime and effectively improve the crane's operating safety and maintenance efficiency. (2) By integrating the Kalman filter algorithm, the update model and the improved long short-term memory neural network through the set operation monitoring module, a comprehensive monitoring system for future fault risk prediction was constructed. The LSTM model improved the prediction accuracy of cross-mechanism related faults under complex working conditions and the ability to identify early weak faults through multi-scale feature fusion, adaptive attention mechanism and focus loss function, thereby improving the safety, reliability and operation and maintenance efficiency of the crane and effectively reducing unplanned downtime and maintenance costs. (3) By integrating kinematic viscosity, total acid value and metal abrasive concentration parameters with equipment usage intensity through the set lubricating oil monitoring module, the remaining life of the lubricating oil can be accurately quantified and early warninged, effectively avoiding equipment failure and unplanned downtime caused by oil failure, reducing maintenance costs and ensuring the safety and reliability of crane operation. (4) By integrating the improved YOLOv11 target detection algorithm with visual information features and operational behavior data through the driver behavior monitoring module, real-time assessment and graded intervention of the driver's status were realized, thereby improving the safety and reliability of crane operation. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a diagram of the crane safety monitoring system architecture of the present invention; Figure 2 This is a schematic diagram of the bidirectional feature pyramid network and attention module of the improved YOLOv11 target detection algorithm of the crane safety monitoring system of the present invention. Figure 3 This is a schematic diagram of the network structure of the attention module in the crane safety monitoring system of the present invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] like Figure 1-3 As shown, in a first aspect, the present invention provides a crane safety monitoring system, the system comprising: The airborne monitoring terminal is deployed on the crane to collect multi-dimensional operating data, lubricating oil status data, and driver facial image data of the crane in real time.

[0020] The multi-dimensional operational data includes the crane's trolley and crane travel, operating speed, acceleration, lifting height, lifting weight, and motor current harmonic parameters. The collected multi-dimensional operational data is preprocessed.

[0021] It should be noted that laser rangefinders and encoders are installed along the tracks of the trolley and crane to monitor their travel and speed in real time; weight and angle sensors are deployed at the hook of the hoisting mechanism to simultaneously collect data on lifting height, lifting weight, and boom angle; the luffing mechanism records amplitude changes via encoders, and the slewing mechanism uses gyroscopes to monitor angular velocity; current transformers and harmonic analysis modules are installed at the motor drive end to collect three-phase current, voltage, and harmonic components in real time; power factor and torque fluctuations are calculated based on motor speed; strain gauge stress sensors are attached to key nodes such as the boom root and main beam to monitor structural stress distribution and fatigue damage; and brake clearance is measured in real time using a constant force opening meter to ensure braking safety.

[0022] The oil monitoring module integrates multi-parameter sensors to collect lubricating oil viscosity, acid value, and metal abrasive concentration in real time. Combined with equipment operating time and load rate, it predicts the remaining life through an oil deterioration model. When the predicted remaining life is less than 20%, the system automatically generates an oil change suggestion and pushes it to the operation and maintenance platform.

[0023] A driver status monitoring unit consisting of a binocular camera and an infrared sensor is deployed above the driver's cab control panel. The binocular camera captures the driver's facial images in real time, while the infrared sensor assists in eye status recognition in low-light environments. The monitoring host has a built-in edge computing box that runs an improved YOLOv11 target detection algorithm to achieve efficient extraction of facial features and simultaneously complete physiological feature analysis of blinking frequency, pupil diameter, and yawning frequency.

[0024] The preprocessing process for the collected multi-dimensional operational data is as follows: The signal acquisition unit uses a 24-bit ADC chip (ADS1256) to perform second-order Butterworth low-pass filtering (cutoff frequency 50Hz) and analog-to-digital conversion on the sensor analog signal, and the sampling frequency is dynamically adjusted (100Hz in steady state and 500Hz in start-stop phase); The monitoring host runs an edge computing program based on an ARM Cortex-A9 processor, removes outliers through a sliding window mean algorithm, and extracts feature parameters to upload to the cloud, reducing the amount of invalid data transmission.

[0025] The operation and maintenance monitoring platform communicates with the airborne monitoring terminal.

[0026] In this embodiment, the operation and maintenance monitoring platform and the airborne monitoring terminal achieve real-time data interaction through a hybrid network transmission architecture. This architecture combines wireless private network and industrial-grade communication technology to achieve efficient and reliable data transmission in complex industrial environments. Within line-of-sight range, communication utilizes a dedicated wireless network comprised of a 5G-CPE and an edge base station to enable real-time transmission of high-definition video streams and high-frequency sampling data. For non-line-of-sight scenarios, communication automatically switches to an industrial-grade LTE-V2X module, ensuring data transmission reliability by controlling the transmission power. This mode is suitable for complex environments such as workshop obstructions and long-distance outdoor locations, ensuring timely delivery of critical alarm information.

[0027] The data transmission process uses a custom binary format, and the data frame structure includes a frame header (device ID), timestamp, data field, and check bit (CRC32) to reduce protocol overhead. The data field is processed by the AES-256 encryption algorithm, and the key is dynamically negotiated between the device's unique identifier and the cloud to prevent data tampering. It also supports breakpoint resume and flow control functions to adapt to complex electromagnetic environments such as ports.

[0028] The data interaction process is as follows: The monitoring terminal collects multi-dimensional data through the sensor module. After filtering and analog-to-digital conversion by the signal acquisition unit, the monitoring host runs the edge computing algorithm to extract feature parameters and remove invalid data. The pre-processed data is transmitted to the operation and maintenance monitoring platform through the 5G / LTE-V2X network. The platform's data middleware uses ETL tools to achieve multi-source data fusion and build a historical operation database.

[0029] The operation and maintenance platform monitors the status of operating equipment. When an anomaly is detected, it triggers an early warning through audible and visual alarms, terminal push notifications, and other means, and sends control commands to the onboard terminal to form a closed-loop management system.

[0030] The operation and maintenance monitoring platform includes an operation monitoring module, a lubricant monitoring module, and a driver behavior monitoring module. In this embodiment, the operation monitoring module is used to fuse the preprocessed multidimensional operation data using the Kalman filter algorithm to obtain a real-time operation status estimate; and to establish update models for the motion mechanism of the crane to calculate the theoretical operation status. Based on the comparison between the real-time operating status estimate and the theoretical operating status, alarm signals for the corresponding motion mechanisms are generated, and a historical operating database is constructed. An improved LSTM neural network is used to construct an alarm prediction model, and historical operating data is used to train the alarm prediction model to predict the probability of failure risk and generate early warning signals for the corresponding moving mechanism.

[0031] Specifically, the operation monitoring module includes: Based on real-time collected data, a state vector is constructed that includes the crane's operating speed, acceleration, and motor current harmonic components. The expression is as follows: ; In the formula, x ( k ) is the state vector.v car For the speed of the car, v crane For the speed of the large vehicle, For lifting acceleration, i harm1 and i harm2 This refers to the harmonic components of the motor current.

[0032] It should be noted that the trolley speed and the crane speed reflect the macroscopic motion performance of the operating mechanism; the lifting acceleration reflects the dynamic load and acceleration characteristics of the lifting mechanism in the vertical direction; abnormal changes in the motor current harmonic components often indicate hidden faults such as broken motor rotor bars, bearing wear, or abnormal load; through this vector, the system can comprehensively perceive the equipment status from both mechanical and electrical dimensions.

[0033] The Kalman filter algorithm is used to fuse the state vector data to obtain the real-time operating state estimate; Specifically, the state equation: ; Observation equation: ; in, F ( k ) is the state transition matrix, B ( k () is the control input matrix. u ( k- 1) For control input vector, H ( k ) represents the observation matrix. w ( k- 1) This is process noise. v ( k ) represents observation noise; Predict the current state based on the state of the previous time step: ; Predicting covariance: ; In the formula, P - ( k ) represents the prior covariance, Q( k- 1) is the covariance matrix of the process noise. F ( k ) T for F ( k The transpose of ) By combining observed values ​​to correct predictions, uncertainty can be reduced. ; In the formula, K ( k ) is the Kalman gain matrix. R ( k ) represents the covariance matrix of the observation noise; Corrected state estimate: ; In the formula, for k Posterior state estimation at time 10:00. To observe residuals; Corrected covariance estimation: ; In the formula, P ( k ) represents the posterior covariance. I It is an identity matrix.

[0034] Establish corresponding updated models for the hoisting mechanism, luffing mechanism and slewing mechanism of the crane respectively, and calculate the theoretical operating state of each mechanism; For the hoisting mechanism, an acceleration calculation model is established, with the following expression: ; The velocity update model expression is: ; In the formula, For the acceleration of the lifting mechanism, F m For the motor output force, m l g For the weight of the goods, c · v l For frictional resistance, m l The weight of the object. c The drag coefficient, v l ( k ) represents the lifting speed, T( k () represents the sampling interval; For the variable amplitude mechanism, an amplitude update model is established, with the following expression: ; ; In the formula, r l ( k () represents the current operating radius. v r ( k () represents the variable speed.v r,min To minimize the variable amplitude speed constraint, v r,max This represents the maximum value of the variable amplitude speed constraint. For the rotary mechanism, an angular velocity update model is established, with the following expression: ; The torque constraint expression is: ; In the formula, ω s α is the current angular velocity. s Angular acceleration, M s To output torque for the rotary mechanism, M s,rated This is the rated torque.

[0035] The real-time operating status estimate is compared with the theoretical operating status in real time. When the real-time operating status estimate is greater than the theoretical operating status, a fault alarm signal for the corresponding mechanism is generated. A historical operation database is constructed based on equipment operating parameters, alarm records, and maintenance history; An alarm prediction model is constructed using an improved LSTM neural network; Specifically, an improved LSTM neural network is used to construct an alarm prediction model, including: After the input layer of the LSTM neural network, a multi-scale feature fusion layer is added to dynamically and weightedly fuse the fine-grained temporal features and coarse-grained temporal features to obtain fused features. The fine-grained temporal features are the collected high-dimensional operating parameter vectors; the coarse-grained temporal features are the aggregated alarm status vectors of each motion mechanism in the historical period. The expression for the multi-scale fusion feature input is as follows: ; In the formula, For dynamic weights, It is a fine-grained feature. It is a coarse-grained feature.

[0036] An adaptive attention layer is added before the output layer of the LSTM neural network to calculate the attention weights of the hidden states of each motion mechanism of the crane to the global prediction target; and the hidden states of all mechanisms are weighted and summed according to the attention weights to obtain the global context hidden state. The institutional association attention mechanism calculates the influence weights of each institution on the prediction target, expressed as: ; The expression for the global context hidden state is: ; In the formula, For the first j The attention weights of each institution at time t. For the first j The hidden state of an organization at time t. h t Hides the state for the global context.

[0037] Based on the global context hidden state, the output layer of the LSTM neural network is used to generate a prediction of the failure risk probability of the crane in the future time period. An improved focusing loss function is adopted as the optimization objective for model training, and its expression is: ; In the formula, oh t For time decay weight, N For the sample size, y t+τ for t + t The true label of time, where γ is the focusing parameter. for t + t Predicted labels for each moment; Using time-series data from the historical operational database, the model parameters are iteratively optimized with the goal of minimizing the focusing loss function until the model converges, resulting in a well-trained alarm prediction model.

[0038] An alarm prediction model is trained using data from a historical operation database. By analyzing the time series characteristics of historical alarm data, the probability of crane failure in future periods is predicted. When the probability of failure exceeds the preset risk threshold, a warning signal is generated for the corresponding motion mechanism.

[0039] This embodiment constructs a comprehensive monitoring system by deeply integrating the Kalman filter algorithm, updating the model, and improving the long short-term memory neural network. This system encompasses real-time accurate state estimation, immediate anomaly detection based on physical deviations, and prediction of future fault risks. It not only effectively suppresses noise interference in industrial settings using Kalman filtering to obtain high-fidelity operational status data, but also achieves keen detection of early, latent faults that deviate from physical laws by establishing theoretical operational models for each mechanism. Furthermore, the LSTM model, through multi-scale feature fusion, adaptive attention mechanism, and focusing loss function, improves the prediction accuracy of cross-mechanism related faults under complex operating conditions and the ability to identify early, subtle faults. This enhances the safety, reliability, and operational efficiency of the crane, effectively reducing unplanned downtime and maintenance costs.

[0040] In this embodiment, the lubricating oil monitoring module is used to establish an oil deterioration model based on the number of times the equipment operates and the lubricating oil status data, calculate the remaining life percentage of the lubricating oil, compare it with a preset life threshold, and generate an oil change warning reminder. Specifically, the lubricating oil monitoring module includes: Real-time data collection of crane lubricating oil's kinematic viscosity, total acid value, and metal abrasive concentration; It should be noted that kinematic viscosity directly reflects the degree of oil aging; during use, lubricating oil viscosity will decrease or increase due to oxidation, shearing, and contamination. Real-time monitoring of its changes is a direct basis for judging whether the lubricating performance of the oil has failed; total acid value is a key chemical parameter for measuring the degree of oxidation of lubricating oil; oil reacts with oxygen at high temperatures to produce acidic substances, leading to an increase in total acid value; a continuously increasing acid value will corrode metal parts and accelerate the deterioration of the oil itself, thus predicting the inherent chemical stability of the oil; metal abrasive concentration, by analyzing the number and composition of metal particles in the oil, can directly obtain wear information of internal mechanical parts of the equipment; abnormal metal abrasive concentration can directly indicate abnormal wear in parts such as gears and bearings, closely linking oil monitoring with equipment mechanical health monitoring.

[0041] Based on the monitored values ​​of the kinematic viscosity, total acid value, and metal abrasive concentration of the lubricating oil, and their corresponding initial values ​​and scrap thresholds, the current deterioration rate of each indicator is calculated. The expression for the current degradation rate of kinematic viscosity is: ; In the formula, Vc This refers to the current kinematic viscosity of the lubricating oil. V 0 The initial kinematic viscosity of the new oil. V lim This refers to the viscosity threshold for scrapping. The expression for the current rate of degradation of total acid value is: ; In the formula, A c This represents the current total acid value of the lubricating oil. A 0 The initial total acid number of the new oil. A lim Acid value rejection threshold; The expression for the current degradation rate of the metal abrasive concentration is: ; In the formula, M c This represents the current concentration of metal abrasive particles in the lubricating oil. M 0 This refers to the initial concentration of metal abrasive particles in the new oil.M lim This is the scrap threshold for metal abrasive particle concentration.

[0042] The current degradation rate of each indicator is weighted and summed, and then added to the cumulative number of times the equipment has been used to establish an oil degradation model. The overall degradation degree is calculated and converted into the percentage of remaining life of the lubricating oil. The expression for the overall degradation degree is: ; In the formula, oh v , oh A , oh M These are the degradation weights for viscosity, acid value, and metal abrasive concentration, respectively. N c The cumulative number of times the equipment has been used. N total Set a safe number of operating cycles for the equipment.

[0043] The expression for the percentage of remaining lifespan is: ; In the formula, RUL % represents the remaining lifespan percentage.

[0044] The remaining life percentage of the lubricating oil is compared with a preset life threshold, and an oil change warning is generated when the remaining life percentage falls below the preset life threshold.

[0045] This embodiment achieves accurate quantitative assessment and early warning of the remaining life of lubricating oil by integrating multi-dimensional oil condition indicators and equipment usage intensity. Specifically, it integrates kinematic viscosity, total acid value, and metal abrasive concentration parameters, and avoids the one-sidedness of judging by a single indicator through weighted calculation, comprehensively reflecting the physical, chemical, and contamination status of the oil. Furthermore, it incorporates the cumulative number of equipment operations into the degradation model, improving the accuracy of remaining life prediction. By using preset thresholds for early warning, it effectively avoids equipment failure and unplanned downtime caused by oil failure, reduces maintenance costs, and ensures the safety and reliability of crane operation.

[0046] In this embodiment, the driver behavior monitoring module uses an improved target detection algorithm to identify the driver's facial features and combines them with driver operation behavior data to obtain a driver status assessment result. The driver status assessment result is then used to issue graded warnings and generate corresponding warning prompts.

[0047] Specifically, the driver behavior monitoring module includes: An improved YOLOv11 target detection algorithm is used to construct a driver state recognition model. The improvement of the driver state recognition model is that its neck network adopts a bidirectional feature pyramid network to replace the original feature pyramid network and path aggregation network structure, so as to achieve efficient fusion of cross-scale features; and an attention module is integrated into the output layer of the bidirectional feature pyramid network to enhance the model's attention to key facial features of the driver by adaptively recalibrating the feature channels. The driver's facial images collected in real time are input into the driver state recognition model, which identifies and outputs the coordinate information of the face, eyes and mouth areas; Based on coordinate information, temporal image sequences of the eye and mouth regions are extracted, and the closing frequency, closing duration, pupil diameter change rate, and yawning frequency of the eyes are calculated to obtain visual information features. It should be noted that the blinking frequency is calculated by counting the number of times the driver's eyes closed within a specific observation period and dividing by the duration to obtain the blinking frequency per unit time; the blinking duration is calculated by averaging the duration of each blink; a longer average blinking duration is an important indicator of deep fatigue; the pupil diameter variation rate is calculated by continuously measuring pupil size and calculating the standard deviation of its diameter value; a larger standard deviation indicates greater pupil fluctuation, reflecting the driver's distraction and drowsiness; the yawning frequency is calculated by counting the number of times the driver yawned within a specific observation period and dividing by the duration to obtain the yawning frequency per unit time.

[0048] The real-time operating handle frequency and operation response time data of the crane control system are acquired and fused with visual information features for weighted calculation to generate the final driver status assessment score.

[0049] It should be noted that the frequency of operation of the control handle is the number of times the driver operates the control handle effectively within a specific observation period, reflecting the driver's operational activity; the operation response time is the average time from when the system issues a command to when the driver reacts.

[0050] The weighted calculation steps include: first, normalizing all indicators by converting them to the same numerical range to eliminate the influence of different units and ensure fairness; then, assigning a weight coefficient to each indicator, which reflects the importance of different indicators in judging fatigue status; finally, multiplying each normalized indicator value by its corresponding weight and adding all products to obtain the final driver status assessment score; this score integrates the driver's facial expressions, physiological reactions, and operational behavior, and can comprehensively and quantitatively reflect their fatigue and attention levels.

[0051] The driver's condition assessment score is compared with multiple preset fatigue thresholds, and corresponding warning prompts are generated based on the level of the threshold exceeded. When the driver's condition assessment score exceeds the first fatigue threshold, it is judged as a low-level warning and a voice reminder is executed; When the driver's condition assessment score exceeds the second fatigue threshold, it is determined to be a medium-level warning, and an audible and visual alarm and seat vibration reminder are activated. When a driver's condition assessment score exceeds the third fatigue threshold, it is considered a high-level warning, and a mandatory rest instruction is sent to the operation and maintenance monitoring platform.

[0052] This embodiment achieves real-time assessment and graded intervention of the driver's state by integrating an improved YOLOv11 target detection algorithm with visual information features and operational behavior data. Specifically, the improved YOLOv11 algorithm, integrating a bidirectional feature pyramid network and an attention module, enhances the detection accuracy and robustness of key facial features under complex lighting and angle variations. Furthermore, the weighted fusion of visual information features and operational behavior data constructs a multi-dimensional assessment model, effectively avoiding the limitations of single visual indicators being easily spoofed or misjudged, making the assessment results more accurately reflect the driver's fatigue and level of inattention. By setting multi-level fatigue thresholds and matching differentiated warning measures, the system effectively alerts the driver and fundamentally prevents safety accidents caused by severe fatigue through mandatory rest commands, thereby improving the safety and reliability of crane operations.

[0053] In this embodiment, the operation and maintenance monitoring platform also includes an energy consumption monitoring module, a personnel management module, a wire rope monitoring module, and a load monitoring module. The energy consumption monitoring module is used to collect crane load data in real time, divide the operating conditions according to the real-time load, establish corresponding operating condition energy consumption models, and perform energy consumption statistics and efficiency analysis.

[0054] Specifically, the energy consumption monitoring module includes: Based on real-time collected load data, the crane's operating status is dynamically divided into multiple operating conditions such as no-load, light-load, medium-load, and heavy-load according to its percentage relative to the rated load, and the cumulative operating time of each condition is recorded. Wherein, if the real-time load L real ≤0.1*P rated If it is 0.1*P, then it is the no-load condition; rated <L real ≤0.4*P rated If it is 0.4*P, then it is a light load condition; rated <L real ≤0.7*P rated If it is 0.7*P, then it is a medium load condition; rated <Lreal ≤1.0*P rated This indicates a heavy-load operating condition; L real For real-time load; P rated The system is the rated load of the crane; it dynamically determines the current operating condition based on the percentage of real-time load to rated load, and accumulates the operating time under each condition.

[0055] Energy consumption models are established to correlate energy consumption with load and runtime under different operating conditions. Based on each model and the runtime of the corresponding operating conditions, the comprehensive energy consumption of the equipment is calculated. The expression for the energy consumption model under no-load conditions is: ; The expression for the energy consumption model under light load conditions is: ; The expression for the energy consumption model under medium load conditions is: ; The expression for the energy consumption model under heavy load conditions is: ; The expression for the total power consumption model is: ; In the formula, P 空载 This is the base power under no-load conditions. t 空载 This refers to the idle running time. C 空载 This is the fixed energy consumption compensation item under no-load conditions. b1 and b1 are the energy consumption coefficients under light load conditions. L For actual load, t 轻载 For light-load operation time, b2 and c2 are the energy consumption coefficients under medium load conditions. t 中载 For medium load operation time, b3 and k 3 represents the energy consumption coefficient under heavy load conditions. t 重载 This refers to the heavy load running time.

[0056] The time utilization rate is calculated based on the actual operation time and the planned operation time; the performance utilization rate is calculated based on the total handling volume and the actual operation time, as well as the rated handling volume and the theoretical optimal cycle time; the quality pass rate is calculated based on the number of safe operation cycles and the total number of operation cycles; and the overall efficiency is calculated based on the product of the time utilization rate, the performance utilization rate, and the quality pass rate. Based on historical data on overall energy consumption and overall efficiency, trend analysis is performed to generate an energy consumption and efficiency analysis report to guide energy-saving operation and maintenance.

[0057] It should be noted that the system acquires the crane's load data in real time and compares it with the equipment's rated load. It then categorizes the current operating status as no-load, light-load, medium-load, or heavy-load using percentage calculations. The system also calculates the cumulative operating time for each condition. For each of the four conditions—no-load, light-load, medium-load, and heavy-load—the system establishes a unique energy consumption calculation model. This model considers the efficiency differences of components such as the motor and transmission system under different loads. Based on the real-time load and operating time, the system calls the corresponding energy consumption model to calculate the energy consumption under each condition and sums them up to obtain the total comprehensive energy consumption over a period of time. The system evaluates energy consumption from three dimensions. The system measures the actual output efficiency of the equipment; time utilization rate, which measures the proportion of actual operating time within the planned working hours; performance utilization rate, which measures the difference between the actual output speed and the theoretical optimal speed; and quality pass rate, which measures the proportion of safe and compliant operations to the total number of operations. Finally, these three ratios are multiplied to obtain the comprehensive efficiency value, which intuitively reflects the overall production efficiency of the equipment. The system continuously records and analyzes historical data on energy consumption and comprehensive efficiency. By identifying the trends in these data, it can proactively discover problems and energy-saving potential in operation and maintenance, and automatically generate analysis reports to achieve the goals of energy saving and efficiency improvement.

[0058] The personnel management module is used to verify the identity of operators through biometric technology and to assign function access permissions and data operation permissions according to preset multi-level permission control rules.

[0059] It should be noted that this system divides users into three levels: administrator, engineer, and operator. Different roles have different access to functional modules and data scopes, and permission configuration supports fine-grained control. For example, operators can only view real-time data and cannot modify system parameters. Furthermore, operators must authenticate their identity through facial recognition when logging into the system or performing sensitive operations to ensure that the operator's identity is uniquely matched with their identity. Once logged in, the system will activate the screen recording function to record all operations performed by the operator on the interface in real time. To save storage space, the recordings will be compressed using H.265 high-efficiency encoding. All records will be accompanied by a precise timestamp and the operator's identity information, forming a complete operation log. Administrators can quickly retrieve and replay any operation process based on conditions such as time, operator name, or operation type, providing a solid data foundation for accident analysis and liability determination. At the same time, after maintenance personnel are verified through facial recognition, the system automatically pushes recent equipment fault records and maintenance suggestions, and sends maintenance reminders to the driver.

[0060] The wire rope monitoring module is used to monitor the tension distribution and metal cross-sectional area loss of the wire rope in real time through tension sensors and magnetic flux detection methods, and calculate its remaining service life based on the fatigue cumulative damage theory to generate a fracture risk warning. It should be noted that the wire rope monitoring module includes tension sensors and magnetic flux sensors. Tension sensors are positioned at the transmission parts or pulley blocks of the wire rope to accurately measure the real-time tension value and distribution of each wire rope during operation. Simultaneously, magnetic flux sensors are installed on the surface of the wire rope or along its path, measuring the change in magnetic field caused by the change in the overall metal cross-sectional area of ​​the wire rope in a non-contact manner. Secondly, the module calculates the tension imbalance between the individual wires to determine if the stress is uniform, thereby detecting potential structural deformation or localized wire breakage. Based on the magnetic flux measurements and a pre-calibrated model, the module calculates the metal cross-sectional area loss rate of the wire rope, directly reflecting the overall strength reduction caused by wear, corrosion, and wire breakage. Next, based on the fatigue cumulative damage theory, the system treats the tension borne by the wire rope in each operation as a fatigue cycle, which continuously accumulates the minor damage caused to the wire rope by each cycle. When the accumulated damage reaches a critical value, it indicates that the wire rope is nearing the end of its design life. This model can dynamically estimate the remaining safe service life of the wire rope. Finally, the three indicators of tension imbalance, metal cross-sectional area loss rate, and cumulative fatigue damage are weighted and integrated to calculate a comprehensive fracture risk index. Based on the fracture risk index, different levels of early warning signals are triggered, similar to the early warning judgment of other modules, and pushed to relevant management personnel and operators to effectively prevent serious accidents such as rope breakage.

[0061] In this embodiment, the load monitoring module calculates the allowable load under the current working condition based on the real-time boom posture and compares it with the actual load in real time. When the limit is exceeded, a graded protection response is triggered.

[0062] Specifically, the load monitoring module includes: Real-time acquisition of crane boom attitude parameters and real-time load; Based on the boom attitude parameters, the allowable load under the current working condition is calculated using the tilting moment model, and the expression is: ; In the formula, Q 允许 For the calculated allowable load, K 安全 To preset a safety factor, M 稳定 The maximum stabilizing moment is calculated based on the equipment's own weight and center of gravity position. G 臂架 For the weight of the boom, L 臂重心The distance from the boom's center of gravity to the center of rotation. L For boom length, i The angle between the boom and the horizontal plane. cosθ A coefficient for converting boom length to horizontal lever arm; The expression for the maximum stabilizing moment is: ; In the formula, G 设备 For the equipment's own weight, d 重心 The center of gravity; The real-time load is compared with the allowable load in real time. When the real-time load reaches the first preset threshold ratio of the allowable load, an audible and visual warning is triggered. When the real-time load reaches the second preset threshold ratio of the allowable load, the crane's dangerous directional movement is automatically restricted. When the real-time load reaches the third preset threshold ratio of the allowable load, the crane's power source is cut off. The third preset threshold ratio is greater than the second preset threshold ratio, which is greater than the first preset threshold ratio.

[0063] In this embodiment, by acquiring boom posture and load data in real time, the precise allowable load under the current working condition is dynamically calculated based on the overturning moment model and compared with the actual load in real time. When the limit is exceeded, a graded protection response is triggered, thereby realizing real-time closed-loop control of the crane's safety status. The safety boundary can be dynamically determined based on the real-time posture such as boom length and angle, effectively preventing overturning accidents caused by overload or lever arm changes. At the same time, the graded response mechanism intervenes in the early stage of risk, which not only ensures operational safety but also avoids unnecessary downtime, achieving a balance between safety and operational efficiency.

[0064] Secondly, the present invention also provides a crane safety monitoring method, implemented using a crane safety monitoring system, comprising the following steps: S1 collects multi-dimensional operating data of the crane, lubricating oil status data, and driver's facial image data in real time; S2. The Kalman filter algorithm is used to fuse the preprocessed multidimensional operating data to obtain a real-time operating state estimate; and an update model is established for the motion mechanism of the crane to calculate the theoretical operating state. S3, based on the comparison between the real-time operating status estimate and the theoretical operating status, generates alarm signals for the corresponding motion mechanism and constructs a historical operating database; S4. An improved LSTM neural network is used to build an alarm prediction model, and historical operating data is used to train the alarm prediction model to predict the probability of failure risk and generate early warning signals for the corresponding moving mechanism. S5: Based on the number of times the equipment operates and the lubricating oil status data, establish an oil deterioration model, calculate the remaining life percentage of the lubricating oil, compare it with the preset life threshold, and generate an oil change warning reminder. S6 uses an improved target detection algorithm to identify the driver's facial features and combines them with the driver's operational behavior data to calculate the driver's state assessment result. The driver's state assessment result is then used to issue graded warnings and generate corresponding warning prompts.

[0065] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A crane safety monitoring system, characterized in that, The system includes: The airborne monitoring terminal is deployed on the crane to collect multi-dimensional operating data, lubricating oil status data and driver facial image data of the crane in real time. The operation and maintenance monitoring platform is communicatively connected to the airborne monitoring terminal. The platform includes an operation monitoring module, a lubricating oil monitoring module, and a driver behavior monitoring module. The operation monitoring module is used to fuse the preprocessed multidimensional operation data using the Kalman filter algorithm to obtain a real-time operation status estimate; and to establish update models for the motion mechanism of the crane to calculate the theoretical operation status. Based on the comparison between the real-time operating status estimate and the theoretical operating status, alarm signals for the corresponding motion mechanisms are generated, and a historical operating database is constructed. An improved LSTM neural network is used to construct an alarm prediction model, and historical operating data is used to train the alarm prediction model to predict the probability of failure risk and generate early warning signals for the corresponding moving mechanism. The lubricating oil monitoring module is used to establish an oil deterioration model based on the number of times the equipment operates and the lubricating oil status data, calculate the remaining life percentage of the lubricating oil, compare it with the preset life threshold, and generate an oil change warning reminder. The driver behavior monitoring module uses an improved target detection algorithm to identify the driver's facial features and combines them with driver operation behavior data to obtain a driver status assessment result. The module then provides graded warnings based on the driver status assessment result and generates corresponding warning prompts.

2. The crane safety monitoring system as described in claim 1, characterized in that, The airborne monitoring terminal collects multi-dimensional operating data, lubricating oil status data, and driver facial image data of the crane in real time. The multi-dimensional operating data includes the crane's trolley travel, operating speed, acceleration, lifting height, lifting weight, and motor current harmonic parameters. The collected multi-dimensional operating data is preprocessed.

3. The crane safety monitoring system as described in claim 2, characterized in that, The operation monitoring module includes: Based on real-time collected data, a state vector is constructed that includes the crane's operating speed, acceleration, and motor current harmonic components. The Kalman filter algorithm is used to fuse the state vector data to obtain the real-time operating state estimate; Establish corresponding updated models for the hoisting mechanism, luffing mechanism and slewing mechanism of the crane respectively, and calculate the theoretical operating state of each mechanism; The real-time operating status estimate is compared with the theoretical operating status in real time. When the real-time operating status estimate is greater than the theoretical operating status, a fault alarm signal for the corresponding mechanism is generated. A historical operation database is constructed based on equipment operating parameters, alarm records, and maintenance history; An alarm prediction model is constructed using an improved LSTM neural network; An alarm prediction model is trained using data from a historical operation database. By analyzing the time series characteristics of historical alarm data, the probability of crane failure in future periods is predicted. When the probability of failure exceeds the preset risk threshold, a warning signal is generated for the corresponding motion mechanism.

4. The crane safety monitoring system as described in claim 3, characterized in that: The alarm prediction model constructed using an improved LSTM neural network includes: After the input layer of the LSTM neural network, a multi-scale feature fusion layer is added to dynamically and weightedly fuse the fine-grained temporal features and coarse-grained temporal features to obtain fused features. The fine-grained temporal features are the collected high-dimensional operating parameter vectors; the coarse-grained temporal features are the aggregated alarm status vectors of each motion mechanism in the historical period. An adaptive attention layer is added before the output layer of the LSTM neural network to calculate the attention weights of the hidden states of each motion mechanism of the crane to the global prediction target; and the hidden states of all mechanisms are weighted and summed according to the attention weights to obtain the global context hidden state. Based on the global context hidden state, the output layer of the LSTM neural network is used to generate a prediction of the failure risk probability of the crane in the future time period. An improved focusing loss function is adopted as the optimization objective for model training, and its expression is: ； In the formula, ω t For time decay weight, N For the sample size, y t+τ for t + τ The true label of time, where γ is the focusing parameter. for t + τ Predicted labels for each moment; Using time-series data from the historical operational database, the model parameters are iteratively optimized with the goal of minimizing the focusing loss function until the model converges, resulting in a well-trained alarm prediction model.

5. The crane safety monitoring system as described in claim 1, characterized in that, The lubricating oil monitoring module includes: Real-time data collection of crane lubricating oil's kinematic viscosity, total acid value, and metal abrasive concentration; Based on the monitored values ​​of the kinematic viscosity, total acid value, and metal abrasive concentration of the lubricating oil, and their corresponding initial values ​​and scrap thresholds, the current deterioration rate of each indicator is calculated. The current degradation rate of each indicator is weighted and summed, and then added to the cumulative number of times the equipment has been used to establish an oil degradation model. The overall degradation degree is calculated and converted into the percentage of remaining life of the lubricating oil. The remaining life percentage of the lubricating oil is compared with a preset life threshold, and an oil change warning is generated when the remaining life percentage falls below the preset life threshold.

6. The crane safety monitoring system as described in claim 1, characterized in that, The driver behavior monitoring module includes: An improved YOLOv11 target detection algorithm is used to construct a driver state recognition model. The improvement of the driver state recognition model is that its neck network adopts a bidirectional feature pyramid network to replace the original feature pyramid network and path aggregation network structure, so as to achieve efficient fusion of cross-scale features; and an attention module is integrated into the output layer of the bidirectional feature pyramid network to enhance the model's attention to key facial features of the driver by adaptively recalibrating the feature channels. The driver's facial images collected in real time are input into the driver state recognition model, which identifies and outputs the coordinate information of the face, eyes and mouth areas; Based on coordinate information, temporal image sequences of the eye and mouth regions are extracted, and the closing frequency, closing duration, pupil diameter change rate, and yawning frequency of the eyes are calculated to obtain visual information features. The real-time operating handle frequency and operation response time data of the crane control system are acquired and fused with visual information features for weighted calculation to generate the final driver status assessment score. The driver's condition assessment score is compared with multiple preset fatigue thresholds, and corresponding warning prompts are generated based on the level of the threshold exceeded. When the driver's condition assessment score exceeds the first fatigue threshold, it is judged as a low-level warning and a voice reminder is executed; When the driver's condition assessment score exceeds the second fatigue threshold, it is determined to be a medium-level warning, and an audible and visual alarm and seat vibration reminder are activated. When a driver's condition assessment score exceeds the third fatigue threshold, it is considered a high-level warning, and a mandatory rest instruction is sent to the operation and maintenance monitoring platform.

7. The crane safety monitoring system as described in claim 2, characterized in that, The operation and maintenance monitoring platform also includes an energy consumption monitoring module, a personnel management module, a wire rope monitoring module, and a load monitoring module. The energy consumption monitoring module is used to collect crane load data in real time, divide the operating conditions according to the real-time load, establish corresponding operating condition energy consumption models, and perform energy consumption statistics and efficiency analysis. The personnel management module is used to verify the identity of operators through biometric technology and to assign function access permissions and data operation permissions according to preset multi-level permission control rules. The wire rope monitoring module is used to monitor the tension distribution and metal cross-sectional area loss of the wire rope in real time through tension sensors and magnetic flux detection methods, and calculate its remaining service life based on the fatigue cumulative damage theory to generate a fracture risk warning. The load monitoring module calculates the allowable load under the current working condition based on the real-time boom posture and compares it with the actual load in real time. When the limit is exceeded, a graded protection response is triggered.

8. The crane safety monitoring system as described in claim 7, characterized in that: The energy consumption monitoring module includes: Based on real-time collected load data, the crane's operating status is dynamically divided into multiple operating conditions such as no-load, light-load, medium-load, and heavy-load according to its percentage relative to the rated load, and the cumulative operating time of each condition is recorded. Energy consumption models are established to correlate energy consumption with load and runtime under different operating conditions. Based on each model and the runtime of the corresponding operating conditions, the comprehensive energy consumption of the equipment is calculated. The time utilization rate is calculated based on the actual operation time and the planned operation time; the performance utilization rate is calculated based on the total handling volume and the actual operation time, as well as the rated handling volume and the theoretical optimal cycle time; the quality pass rate is calculated based on the number of safe operation cycles and the total number of operation cycles; and the overall efficiency is calculated based on the product of the time utilization rate, the performance utilization rate, and the quality pass rate. Based on historical data on overall energy consumption and overall efficiency, trend analysis is performed to generate an energy consumption and efficiency analysis report to guide energy-saving operation and maintenance.

9. The crane safety monitoring system as described in claim 7, characterized in that: The load monitoring module includes: Real-time acquisition of crane boom attitude parameters and real-time load; Based on the boom attitude parameters, the allowable load under the current working condition is calculated using the tilting moment model, and the expression is: ； In the formula, Q 允许 For the calculated allowable load, K 安全 To preset a safety factor, M 稳定 The maximum stabilizing moment is calculated based on the equipment's own weight and center of gravity position. G 臂架 For the weight of the boom, L 臂重心 The distance from the boom's center of gravity to the center of rotation. L For boom length, θ The angle between the boom and the horizontal plane. cosθ A coefficient for converting boom length to horizontal lever arm; The real-time load is compared with the allowable load in real time. When the real-time load reaches the first preset threshold ratio of the allowable load, an audible and visual warning is triggered. When the real-time load reaches the second preset threshold ratio of the allowable load, the crane's dangerous directional movement is automatically restricted. When the real-time load reaches the third preset threshold ratio of the allowable load, the crane's power source is cut off. The third preset threshold ratio is greater than the second preset threshold ratio, which is greater than the first preset threshold ratio.

10. A crane safety monitoring method, implemented using the crane safety monitoring system as described in any one of claims 1-9, characterized in that: Includes the following steps: S1 collects multi-dimensional operating data of the crane, lubricating oil status data, and driver's facial image data in real time; S2. The Kalman filter algorithm is used to fuse the preprocessed multidimensional operating data to obtain a real-time operating state estimate; and an update model is established for the motion mechanism of the crane to calculate the theoretical operating state. S3, based on the comparison between the real-time operating status estimate and the theoretical operating status, generates alarm signals for the corresponding motion mechanism and constructs a historical operating database; S4. An improved LSTM neural network is used to build an alarm prediction model, and historical operating data is used to train the alarm prediction model to predict the probability of failure risk and generate early warning signals for the corresponding moving mechanism. S5: Based on the number of times the equipment operates and the lubricating oil status data, establish an oil deterioration model, calculate the remaining life percentage of the lubricating oil, compare it with the preset life threshold, and generate an oil change warning reminder. S6 uses an improved target detection algorithm to identify the driver's facial features and combines them with the driver's operational behavior data to calculate the driver's state assessment result. The driver's state assessment result is then used to issue graded warnings and generate corresponding warning prompts.

Images