Method and system for intelligently controlling crane

Abstract

The invention relates to a method and system for intelligently controlling a crane, and belongs to the technical field of crane control. The method comprises the following steps: acquiring working state data of a crane through a sensor, analyzing and modeling, calculating a dynamic response weight and a static characteristic weight, and quickly capturing key factors influencing the state of the crane by the dynamic response weight to trigger quick response; and the static feature weight maintains basic data classification and reduces redundant calculation. In combination with dynamic and static risks, risk indexes are calculated in real time, risk levels are adaptively adjusted, in addition, a risk assessment model is expanded through multi-source data acquisition, operation difficulty assessment, scene sensitive feature extraction and the like, sensor noise is eliminated through adaptive Kalman filtering, weight distribution is dynamically adjusted, and a risk assessment model is established. According to the method, the data accuracy and the evaluation flexibility are remarkably improved, intelligent control over the working state of the crane is achieved, the operation safety and efficiency are enhanced, and meanwhile the maintenance cost and resource configuration are optimized.

Description

Technical Field

[0001] The present invention relates to a method and system for intelligent control of a crane, belonging to the technical field of crane control. Background Art

[0002] With the continuous development of the existing industry, cranes have become a very widely used mechanical device in industrial and mining enterprises, such as: mining, metallurgy, coal carbonization, cement, and electric power industries. Specifically, a crane is a mechanical device used for operations such as lifting, transporting, loading and unloading, and installing materials.

[0003] However, in the prior art, crane control mainly relies on sensor data acquisition and static model analysis, and cannot dynamically adapt to the time-varying nature of sensor noise, resulting in data distortion and affecting the accuracy of subsequent analysis. The sensor weights in traditional models are mostly preset fixed values and cannot be dynamically adjusted according to real-time working conditions. For example, when the load suddenly changes or the environmental wind speed suddenly increases, static weights are difficult to capture key risk indicators, resulting in a lag in risk assessment. Traditional maintenance strategies lack a refined classification mechanism, often increasing costs due to over-maintenance, or causing safety hazards due to insufficient maintenance. Summary of the Invention

[0004] The present invention provides a method and system for intelligent control of a crane to solve the problems of insufficient sensor noise processing, rigid weight allocation, and imbalance between maintenance cost and safety in the prior art.

[0005] The present invention provides a method and system for intelligent control of a crane, which includes 8) Collecting the working state data of the crane through sensors and parsing the collected data; 9) Building a working state model based on the processed collected data and calculating the weights corresponding to the sensor data; 10) Parsing the sensor weights includes dynamic response weights and static feature weights. Among them, the dynamic response weights comprehensively sort out the factors affecting the working state of the tower crane and trigger a rapid response. The static feature weights analyze the motion parameters, maintain the basic data classification, and reduce redundant calculations; 11) Calculating the real-time risk index by combining dynamic risk and static risk, classifying the risk levels, adaptively adjusting according to the degree of deviation of the real-time risk from the safety threshold, controlling the adjustment rate, balancing the weights of sudden risks and long-term trends, and preventing a single indicator from dominating the decision-making; 12) Conducting multi-source data acquisition, evaluating the operation difficulty, calculating the action combination entropy value, expanding the static weight Bayesian network, and integrating the evaluation results into the risk assessment model; 13) Extracting scene-sensitive features, expanding the static weight Bayesian network, comparing the predicted life and actual damage data, and dynamically updating the model parameters; 14) Send the parsed working status data of the crane to the operator for data interaction.

[0006] Preferably, in step 1), the collected data is subjected to adaptive Kalman filtering to eliminate sensor noise, time synchronization is performed through multi-source data spatio-temporal registration, and filtering iteration is carried out.

[0007] Preferably, in step 2), a state equation and an observation equation are constructed, a crane coordinate system is established, installation errors are eliminated by using multi-sensor data fusion, the absolute coordinates of the origin of the sensor coordinate system are obtained, and a working status model is built based on the processed collected data.

[0008] Preferably, in step 3), initial weight values are assigned to the indicators that affect the operation safety of the crane according to historical accident data. By collecting the data of each monitoring indicator in real time, the dynamic response weights are calculated and updated in real time, static characteristic reference weights are assigned to different sensor data. When major repairs, component replacements or major changes occur in the working environment of the crane, the weights are re-evaluated and adjusted, and a relationship model is constructed based on the dynamic response weights and static characteristic weights.

[0009] Preferably, in step 4), a Bayesian network causal relationship diagram is established according to the static weights. When major maintenance or environmental changes occur, the update of Bayesian network parameters is triggered, and the maintenance cost is minimized under the constraint of the particle swarm optimization algorithm, so as to balance cost and safety.

[0010] Preferably, in step 5), the complexity score of the calculation model is calculated by quantization, the hierarchical clustering method is used to divide into 3 levels, and a work shift grouping is established; the grouping is integrated into the dynamic weight calculation, and grouping nodes of work complexity and work shifts are added to the Bayesian network to optimize the risk assessment model, and environmental interference is misidentified as equipment failure, resulting in mis-shutdown.

[0011] Preferably, in step 6), a scenario-specific risk index is introduced into the real-time risk assessment model, the static weight Bayesian network is extended, new root nodes of construction stage and environmental complexity are added, a physical-data fusion model and a vibration fatigue damage model are established to enhance the life prediction model, and simulations are carried out under extreme working conditions to compare the predicted life with the actual damage data and dynamically update the model parameters.

[0012] Preferably, in step 7), when the working data of the crane is greater than the first threshold, it is determined that the current tower crane working mode is normal. When the working data of the crane is less than the first threshold, it is determined that the current crane working mode is abnormal, and the data of working status control and maintenance are sent to the background server.

[0013] Preferably, the extracted features are standardized, a time window is set to divide the continuous data stream into time series segments, and the segments are input into a bidirectional LSTM neural network. The forward layer captures the historical risk accumulation, the backward layer captures the future abnormal trends, weights are dynamically assigned to the features at each time step, and a real-time risk index is output.

[0014] Preferably, the load weight and volume are monitored in real time, an inclinometer records the path curvature radius and the operation action sequence, an environmental sensor collects the wind speed and temperature, the continuous variables are standardized by Z-score, and the discrete variables are normalized to 0-1.

[0015] The present application also provides a system for intelligent control of a crane, including sensors and an inclinometer record. The working state data is collected through the sensors, analyzed according to the collected data, a working state model is built, the dynamic response weights and static feature weights of the corresponding data are calculated, the real-time risk index is calculated by combining the dynamic risk and the static risk, the risk levels are divided. The sensors and the inclinometer perform multi-source data collection, expand the static weight Bayesian network, extract the scene-sensitive features, dynamically update the model parameters, and send the working state data to the operator for data interaction.

[0016] Advantages of the present invention: The present invention provides a method and system for intelligent control of a crane. Through adaptive Kalman filtering, it can dynamically adjust the noise parameters to adapt to data changes, effectively eliminate sensor noise, and significantly improve the accuracy of data. The introduction of the fading factor method or square root adaptive filtering further enhances the filtering effect, ensuring that the denoised sensor data is more reliable. By using linear interpolation or cubic spline interpolation based on timestamps, combined with a timestamp error compensation model and the least squares method, it can efficiently align asynchronous sensor data to a unified time series, significantly reducing the synchronization error. A crane coordinate system is established, and a coordinate transformation matrix is used to achieve spatial alignment of multi-sensor data, which not only simplifies the data processing flow but also improves the accuracy and efficiency of data fusion. The dynamic response weight is calculated and updated in real time according to the factors affecting the working state of the crane, ensuring the timeliness and accuracy of weight allocation. The static feature weight is assigned based on key data of motion parameters and re-evaluated and adjusted under specific circumstances, avoiding the limitations of fixed traditional weight allocation and improving the accuracy and flexibility of evaluation. The combination of dynamic weight and static weight balances real-time response and basic logic, reduces redundant calculations, and improves the overall efficiency. The application of a bidirectional LSTM neural network can process forward and backward time series information simultaneously, capture historical risk accumulation and future abnormal trends, dynamically assign weights to features at each time step, and output a real-time risk index, improving the flexibility and accuracy of evaluation. By dynamically calculating the risk level and dynamically adjusting the coefficients, the refinement of risk level classification is achieved. According to different risk levels, corresponding maintenance measures are taken, which not only ensures the safe operation of the crane but also reduces the maintenance cost. According to the comparison result of the crane working data and the preset threshold, it can timely determine whether the working mode of the crane is normal, providing an effective fault warning for the operator. The data is interacted with the operator and the background server, realizing real-time sharing and remote monitoring of data, facilitating timely maintenance measures and fault repair, and improving the safety and reliability of the crane. Based on the particle swarm optimization algorithm, the maintenance cost is minimized under the constraint that the risk level is lower than a certain threshold, realizing the refined allocation of resources. The introduction of a scenario-specific risk index can more accurately evaluate the risk level in the current scenario, providing strong support for decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic flowchart of a method and system for intelligent control of a crane according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] Embodiment 1 Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.

[0019] The present invention provides a method and system for intelligent control of a crane, which includes: Collect the working state data of the crane through sensors and analyze the collected data; Specific sensors collect the load changes, motion parameters, environmental parameters, and structural stresses of the crane, and perform adaptive Kalman filtering on the collected data to eliminate sensor noise. Time synchronization is carried out through multi-source data spatio-temporal registration. Specifically, based on the collected data of each sensor, the measurement noise covariance matrix and the process noise covariance matrix are estimated online, and the fading factor method or square root adaptive filtering is used to dynamically adjust the noise parameters to adapt to data changes, and the state equation and observation equation are constructed as follows:

[0020] Among them

[0021] : System state vector (lifting capacity, motion parameters); : Observation vector; : Process noise; : Measurement noise; Perform filtering iteration. By executing the prediction and update steps, output the denoised sensor data and dynamically optimize the measurement noise covariance matrix and the process noise covariance matrix through residual analysis; Based on linear interpolation or cubic spline interpolation of timestamps, align the asynchronous sensor data to a unified time series. For asynchronous sampling sensors, introduce a timestamp error compensation model and optimize the synchronization accuracy through the least squares method; Establish a crane coordinate system, convert the data of each sensor into physical quantities in a unified coordinate system, and use the coordinate transformation matrix To achieve spatial alignment of multi-sensor data. When the crane is in a stationary state, directly use the external measurement device as the translation vector And use multi-sensor data fusion to optimize Parameters to eliminate installation errors and obtain the absolute coordinates of the origin of the sensor coordinate system; Build a working state model based on the processed collected data, specifically:

[0022] Among them: : Comprehensive evaluation value of working state; : Number of sensors; : The th data collected and processed by the sensor; : the weight corresponding to the sensor data of the th sensor; Among them, the weight corresponding to the sensor data includes a dynamic response weight and a static feature weight. The dynamic response weight comprehensively sorts out the factors affecting the working state of the tower crane and triggers a quick response, including load changes, environmental parameters, and structural stress. For each factor, the corresponding monitoring indicators are clarified. Among them, the load changes include the lifting moment and the lifting weight. The environmental parameter is the environmental wind speed, which is directly used as a monitoring indicator. The structural stress includes the stress values of the boom and the tower standard section. According to the historical accident data, the initial weight value is assigned to each indicator affecting the safe operation of the crane. By collecting the data of each monitoring indicator in real time, the dynamic response weight is calculated and updated in real time. After each collection, the weight is updated to ensure the timeliness of the weight; The static feature weight analyzes the motion parameters, maintains the basic data classification, and reduces redundant calculations. The motion parameters include three working modes of the crane: stretching, rotating, and supporting. The key data types under each mode are determined. When stretching, the stretching displacement and telescopic speed of the component are the key data. When rotating, the rotation angle and rotation speed are the key. When supporting, the pressure and verticality of the supporting component are the key. According to the key data, the reference weight is assigned to different sensor data. When the crane undergoes major repairs, component replacements, or major changes in the working environment, the weight is re-evaluated and adjusted; Construct a relationship model based on the dynamic response weight and the static feature weight. The dynamic response weight ( ) changes according to the real-time working state, and the static feature weight ( ) is preset based on the fixed working mode. Introduce the dynamic weight coefficient ( ) and the static weight coefficient , and the two satisfy , and . The dynamic response weight and the static feature weight are combined through the following relationship model:

[0023] The dynamic weight quickly captures anomalies, and the static weight maintains the basic logic, accelerating the overall state determination while ensuring the depth of key data; Send the parsed working state data of the crane to the operator for data interaction. Specifically, when the working data of the crane is greater than the first threshold, it is determined that the current working mode of the tower crane is normal, and the working state of the crane in the current environment is within the normal range. When the working data of the crane is less than the first threshold, it is determined that the current working mode of the crane is abnormal. The operator prohibits the operation of the crane's working condition and performs fault repair, and sends the data of working state control and maintenance to the background server.

[0024] Specifically, during use, working condition data such as the load changes of the crane (such as lifting moment, lifting weight), motion parameters (extension displacement, rotation angle, etc.), environmental parameters (ambient wind speed), and structural stress (stress values of the boom and tower standard sections) are collected by sensors. The sensor noise is eliminated through adaptive Kalman filtering to adapt to real-time data changes. The measurement noise covariance matrix and process noise covariance matrix are estimated online. The fading factor method or square root adaptive filtering is used to dynamically adjust the noise parameters. During the filtering iteration, the noise parameters are dynamically optimized through prediction and update steps to improve the denoising accuracy. Based on linear interpolation or cubic spline interpolation of timestamps, combined with the timestamp error compensation model and the least squares method, the asynchronous sensor data is aligned to a unified time series to ensure that the data is in a unified time series, reducing synchronization errors and avoiding time differences during data fusion caused by different sensor sampling rates or asynchronous timestamps, which affect real-time performance and decision-making accuracy. A crane coordinate system is established, and the coordinate transformation matrix is used to convert the data of each sensor into physical quantities in a unified coordinate system. The parameters are optimized through external measurement devices and multi-sensor data fusion to obtain the absolute coordinates of the origin of the sensor coordinate system, and the dynamic response weights and static feature weights corresponding to the sensor data are determined. The dynamic response weights are calculated and updated in real time according to the factors affecting the working condition of the crane, and the static feature weights are allocated based on the key data of the motion parameters and re-evaluated and adjusted under specific circumstances to avoid the fixed traditional weight allocation that cannot be adjusted according to real-time working conditions, resulting in inaccurate evaluation in a dynamic environment. The combination of dynamic weights and static weights balances real-time response and basic logic, reduces redundant calculations, and improves efficiency. The two are combined through a relationship model, and a working condition model is built based on the processed collected data, and the parsed working condition data is sent to the operator. When the working data of the crane is greater than the first threshold, the working mode is considered normal; when it is less than the first threshold, the working mode is considered abnormal. The operator prohibits the crane from working and performs fault maintenance, and at the same time sends the working condition control and maintenance data to the background server.

[0025] Compared with the prior art, through adaptive Kalman filtering, the noise parameters can be dynamically adjusted to adapt to data changes, effectively eliminating sensor noise, improving the accuracy of data. The introduction of the fading factor method or square root adaptive filtering further enhances the filtering effect, ensuring that the denoised sensor data is more reliable. By using linear interpolation or cubic spline interpolation based on timestamps, combined with the timestamp error compensation model and the least squares method, the asynchronous sensor data can be efficiently aligned to a unified time series, significantly reducing the synchronization error. A crane coordinate system is established, and a coordinate transformation matrix is used to achieve the spatial alignment of multi-sensor data, which not only simplifies the data processing flow but also improves the accuracy and efficiency of data fusion. The dynamic response weight is calculated and updated in real time according to the factors affecting the working state of the crane, ensuring the timeliness and accuracy of weight allocation. The static feature weight is allocated based on the key data of motion parameters and re-evaluated and adjusted under specific circumstances, avoiding the limitations of fixed traditional weight allocation and improving the accuracy and flexibility of evaluation. The combination of dynamic weight and static weight balances real-time response and basic logic, reduces redundant calculations, and improves the overall efficiency. According to the comparison result of the crane working data with the preset threshold, it is timely determined whether the working mode of the crane is normal, providing an effective fault warning for the operator. By interacting the data with the operator and the background server, real-time data sharing and remote monitoring are achieved, facilitating the timely adoption of maintenance measures and fault repair, and improving the safety and reliability of the crane.

[0026] Embodiment 2 In the above embodiment, the dynamic response weight is calculated and updated in real time according to the factors affecting the working state of the crane, ensuring the timeliness and accuracy of weight allocation. The static feature weight is allocated based on the key data of motion parameters and re-evaluated and adjusted under specific circumstances, avoiding the limitations of fixed traditional weight allocation and improving the accuracy and flexibility of evaluation. The embodiment of the present application is optimized on the basis of the above embodiment.

[0027] In this embodiment, the dynamic response weight extracts the load mutation rate, wind speed gradient, and stress growth rate. The extracted features are Z-score standardized. A time window is set to divide the continuous data stream into time series segments and input into a bidirectional LSTM neural network. The bidirectional LSTM neural network processes the forward and backward time series information simultaneously. The forward layer captures the historical risk accumulation, and the backward layer captures the future abnormal trend, dynamically allocating weights to the features at each time step and outputting a real-time risk index; Perform offline modeling of static weights based on static weights, collect historical maintenance records, fault data, and component replacement logs, and establish a Bayesian network causal relationship diagram. The root nodes are the component aging stage and working mode, the intermediate nodes are sensor data, and the leaf nodes are fault types. Use the maximum likelihood estimation method to learn the conditional probabilities between nodes from historical data. When major maintenance or environmental changes occur, trigger the update of Bayesian network parameters; Calculate the real-time risk index by combining dynamic risk and static risk:

[0028] Among them: : Real-time risk index calculation; : Dynamic risk; : Static risk Use the Sigmoid function to calculate dynamically , and perform dynamic adjustment of coefficients: Among them: k = 0.1; : Safety threshold; = 1 - ; Classify the risk levels, : Extend the maintenance cycle by 30%; 30% ≤ < 70% to maintain the original cycle and increase the online monitoring frequency; ≥ 70% to immediately shut down for maintenance; Based on the particle swarm optimization algorithm, minimize the maintenance cost under the constraint of < 70%. After each maintenance, compare the actual fault data with the prediction results, update the LSTM and Bayesian network parameters for data feedback loop. When the dynamic weight detects an anomaly but the static weight shows that the remaining life rate of the component > 80%, trigger the "observation period" and make a decision after verification by multi-source data, to avoid frequent maintenance when the dynamic weight detects an anomaly, increase the cost, and the historical data of the static weight can help determine whether immediate maintenance is required or it can be postponed, so as to balance cost and safety.

[0029] Dynamic data acquisition during use: Real-time monitoring of the working state parameters of the crane, including dynamic characteristics such as load mutation rate, wind speed gradient, stress growth rate, etc. Static data collection: Offline collation of historical maintenance records, fault data, component replacement logs and other static information. Perform Z-score standardization on the dynamic characteristics to eliminate the dimension difference. Divide the continuously collected dynamic data into time series segments of fixed length, so as to adapt to the timing characteristics of the equipment operation state, improve the sensitivity to instantaneous anomalies, and process through a bidirectional LSTM neural network. The forward layer analyzes historical data to capture the trend of risk accumulation (such as continuous stress growth), and the reverse layer predicts future abnormal signals (such as the possibility of wind speed mutation). Assign weights to the features at each time step and output the real-time dynamic risk index. Through the maximum likelihood estimation method, learn the conditional probability between nodes from historical data. When major maintenance or environmental changes occur, retrain the Bayesian network to output the static risk index, which is dynamically calculated using the Sigmoid function, adaptively adjusted according to the degree of real-time risk deviation from the safety threshold, control the adjustment rate, balance the weights of sudden risks and long-term trends, and prevent a single indicator from dominating the decision-making. At Generate a maintenance plan using the particle swarm optimization algorithm under the <70% constraint to minimize the total cost. After each maintenance is completed, compare the actual fault data with the prediction results and update the parameters of the bidirectional LSTM and Bayesian networks. If the dynamic weight detects an anomaly, but the static weight shows that the remaining component life rate > 80%, then trigger the "observation period" to comprehensively verify the authenticity of the risk with multi-sensor data, avoid blind maintenance, and achieve refined resource allocation.

[0030] Compared with the prior art, by dynamically responding to weights and calculating and updating weights in real time according to the factors affecting the working state of the crane, the timeliness and accuracy of weight allocation are ensured. The static feature weights are re-evaluated and adjusted under specific circumstances, avoiding the limitations of fixed traditional weight allocation. The application of the bidirectional LSTM neural network can process forward and backward time series information simultaneously, capture historical risk accumulation and future abnormal trends, dynamically allocate weights to features at each time step, output a real-time risk index, improve the flexibility and accuracy of evaluation, dynamically calculate the risk level through the Sigmoid function, and perform coefficient dynamic adjustment, achieving refined risk level division. According to different risk levels, corresponding maintenance measures are taken, such as extending the maintenance cycle, maintaining the original cycle and increasing the online monitoring frequency, immediately shutting down for maintenance, etc., which not only ensures the safe operation of the crane but also reduces the maintenance cost. When the dynamic weight detects an anomaly but the static weight shows that the remaining component life rate > 80%, a "observation period" is triggered, and a decision is made after verification by multi-source data, avoiding the increased cost caused by frequent maintenance. The introduction of static weights can help determine whether immediate maintenance is required or it can be postponed, thus achieving a balance between cost and safety. Based on the particle swarm optimization algorithm, the maintenance cost is minimized under the constraint that the risk level is lower than 70%, realizing refined resource allocation. After each maintenance is completed, the actual failure data is compared with the prediction results, and the parameters of the bidirectional LSTM and Bayesian network are updated, improving the accuracy of prediction and maintenance.

[0031] Embodiment 3 In the above embodiments, corresponding maintenance measures are taken according to different risk levels, and a decision is made after verification by multi-source data, avoiding the increased cost caused by frequent maintenance. The embodiments of the present application are optimized on the basis of the above embodiments.

[0032] In this embodiment, multi-source data is collected to obtain the density of obstacles in the lifting path, the load weight and volume are monitored in real time, the inclinometer records the path curvature radius and the operation action sequence, and the environmental sensor collects the wind speed and temperature. The continuous variables such as weight, volume, and curvature radius are standardized by Z-score, and the discrete variables (such as the number of obstacles) are normalized to 0-1; For the evaluation of operation difficulty, calculate the action combination entropy value:

[0033] Where: : is the occurrence probability of the th action sequence; Calculate the complexity score through the quantization calculation model: ; Where: : Weight coefficient; : Volume coefficient; : Obstacle density; : Number of curvature points; : Environment factor; Use hierarchical clustering method to divide into 3 levels: where low complexity: score < 0.4, medium complexity: 0.4 ≤ score < 0.7, high complexity: score ≥ 0.7; Extract spatio-temporal features and establish job shift groups; Integrate the groups into the dynamic weight calculation, and the real-time weight update formula:

[0034] Where: : Group sensitivity coefficient matrix; : 0.1 - 0.3 (dynamically adjusted); Group sensitivity coefficients include high-complexity operations and night-shift operations, where high-complexity operations include the weight of stress sensors, and night-shift operations include the weight of visibility sensors; Expand the static weight Bayesian network, add grouping nodes of job complexity and job shift to the Bayesian network, where job complexity includes low, medium, and high, and job shift includes day, middle, and night, and calculate the joint probability of grouping and failure type:

[0035] Optimize the risk assessment model, and the real-time risk index is specifically:

[0036] Where; : Complexity correction coefficient; : Shift correction coefficient; Group correction coefficient for high-complexity operations: dynamic risk R × 1.2; Group correction coefficient for night-shift operations: static risk S × 1.1; When in use, by obtaining the obstacle density, collecting external environment parameters, the inclinometer records the path curvature radius, and real-time monitoring of the load weight and volume, perform Z-score normalization on the weight, volume, and curvature radius, and perform 0-1 normalization on the number of obstacles to achieve data preprocessing; Statistical historical operation action sequences, and calculate the occurrence probability of each action sequence , perform the calculation of the entropy value of the action combination. The higher the entropy value, the more complex the operation. Calculate the complexity score, classify the complexity level, divide it into three levels of complexity through hierarchical clustering, implement the scenario adaptive strategy, extract spatio-temporal features, combine shifts with the complexity level to achieve job shift grouping. Factors such as the reduced visibility and personnel fatigue during night shifts are not included in the model, and environmental interference is misidentified as equipment failure, leading to false shutdowns. Establish job shift grouping, capture the day-night environmental differences, update the dynamic weights, expand the Bayesian network and static risk modeling, add nodes for job complexity and job shifts, calculate the joint probability, and recalculate the conditional probability table when new job data is added to optimize the fault prediction.

[0037] Compared with the prior art, through multi-source data collection, including information such as the density of obstacles in the lifting path, load weight and volume, path curvature radius, operation action sequence, wind speed and temperature, etc., comprehensively evaluate the working state of the crane, which can more comprehensively reflect the operating conditions of the crane, improve the comprehensiveness and accuracy of the evaluation. By calculating the entropy value of the action combination and the complexity score, evaluate the operation difficulty, and integrate the evaluation results into the risk assessment model, which can more accurately reflect the impact of the operation difficulty on the working state of the crane and improve the accuracy of the evaluation.

[0038] Embodiment 4 In the above embodiment, the dynamic weight adjustment is macro-grouped, and the complexity of the operation plan and the job shift are macro-grouped to optimize the result of the safety balance and ultimately achieve the reduction of the maintenance cost. The embodiment of the present application is optimized to a certain extent on the basis of the above embodiment.

[0039] In this embodiment, scene-sensitive features are extracted, including water conservancy project feature extraction, traffic engineering feature extraction, and cross-scene feature extraction. Among them, water conservancy project feature extraction includes the water level change rate, water flow impact force, and scour depth; traffic engineering feature extraction includes the traffic flow of adjacent lanes, vibration interference, and dynamic load; cross-scene feature extraction includes the complexity of multi-device collaboration, specifically:

[0040] Among them : Calculate the distance between devices; S: The action synchronization rate between devices; For continuous variables (such as water level, traffic flow), use Z-score standardization; for discrete variables (such as scour depth level), use 0-1 normalization; Expand the dynamic weights:

[0041] Among them: : Scenario For the sensor Sensitivity coefficient; : Risk response coefficient (0.2 - 0.4, dynamically adjusted); The real - time risk assessment model introduces a scenario - specific risk index, specifically:

[0042] Where: : Basic risk index (output of the original model); : Scenario risk amplification coefficient; Expand the static - weighted Bayesian network, add root nodes of construction stage and environmental complexity, and update the conditional probability table, specifically:

[0043] Establish a physical - data fusion model and a vibration fatigue damage model to enhance the life prediction model. The specific physical - data fusion model is:

[0044] Where: : Theoretical life based on fracture mechanics; : Remaining life prediction based on LSTM; The vibration fatigue damage model is specifically:

[0045] Where: : Actual number of vibrations; : Number of vibrations for fatigue life; Simulate extreme working conditions in the simulation model, compare the predicted life with the actual fatigue damage, and update the model parameters; During use, collect the characteristics of water conservancy projects, transportation projects, and cross - scenario characteristics, and for continuous variables (such as water level, traffic flow), use Z - score standardization, and for discrete variables (such as scour depth level), use 0 - 1 normalization to achieve multi - scenario feature data collection and pre - processing. For the dynamic weight expansion and scenario - specific risk assessment, as well as the static weight Bayesian network expansion and fault prediction optimization, physical - data fusion life prediction and vibration fatigue modeling, simulate under extreme working conditions, compare the predicted life with the actual damage data, and dynamically update the model parameters.

[0046] Compared with the prior art, by extracting hydraulic engineering characteristics (such as water level change rate, water flow impact force, scour depth), traffic engineering characteristics (such as traffic flow in adjacent lanes, vibration interference, dynamic load), and cross-scene characteristics (such as multi-device collaboration complexity), it is possible to more accurately reflect the operating environment and requirements in different scenarios, which helps subsequent risk assessment and weight allocation, thereby improving the safety and efficiency of operations. By introducing the sensitivity coefficient of the scene to the sensor and the risk response coefficient, the expansion of the dynamic weight is realized, and it is possible to more flexibly respond to the risk changes in different scenarios. The introduction of the real-time risk assessment model, combined with the scene-specific risk index, can more accurately evaluate the risk level in the current scene and provide strong support for decision-making. By expanding the static weight Bayesian network, adding root nodes such as the construction stage and environmental complexity, and updating the conditional probability table, the model is made more perfect and can more accurately reflect the actual situation. The introduction of the physical-data fusion model and the vibration fatigue damage model further enhances the accuracy of the life prediction model. This fusion model combines the theoretical life of fracture mechanics and the remaining life prediction of LSTM, and can more comprehensively consider the influence of various factors on the equipment life.

[0047] Embodiment 5 In this embodiment, a system for intelligent control of a crane is provided, including sensors and inclinometer records. The sensors collect the load changes, motion parameters, environmental parameters, and structural stresses of the crane, analyze the collected data, and build a working state model to calculate the dynamic response weight and static characteristic weight of the corresponding data. The dynamic response weight comprehensively sorts out the factors affecting the working state of the tower crane and triggers a rapid response, including load changes, environmental parameters, and structural stresses. For each factor, the corresponding monitoring indicators are clarified. The static characteristic weight analyzes the motion parameters, maintains the basic data classification, and reduces redundant calculations. The motion parameters include three working modes of the crane: extension, rotation, and support, and the key data types in each mode are determined. The real-time risk index is calculated by combining the dynamic risk and the static risk, and the risk level is divided. The sensors and inclinometers perform multi-source data collection, collect the density of obstacles in the hoisting path, real-time monitor the load weight and volume, the inclinometer records the path curvature radius and the operation action sequence, and the environmental sensors collect the wind speed and temperature. The static weight Bayesian network is expanded, the scene-sensitive features are extracted, the model parameters are dynamically updated, and the working state data is sent to the operator for data interaction. When the crane working data is greater than the first threshold, it is determined that the current tower crane working mode is normal, and the working state of the crane in the current environment is within the normal range. When the crane working data is less than the first threshold, it is determined that the current crane working mode is abnormal, and the operator prohibits the operation of the crane's working condition and performs fault repair, and sends the data of the working state control and maintenance to the background server.

[0048] The present invention and its embodiments have been described above. Such description is not restrictive. What is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. In general, if those of ordinary skill in the art are inspired by it and, without departing from the gist of the present invention, design similar structural modes and embodiments to this technical solution without creative efforts, they shall fall within the protection scope of the present invention.

Claims

1. A method for intelligent control of a crane, characterized in that, Including: 1) Collect the working state data of the crane through sensors and analyze the collected data; 2) Build a working state model based on the processed collected data and calculate the weights corresponding to the sensor data; 3) The analysis of the sensor weights includes dynamic response weights and static feature weights. Among them, the dynamic response weights comprehensively sort out the factors affecting the working state of the tower crane and trigger rapid responses. The static feature weights analyze the motion parameters, maintain the classification of basic data, and reduce redundant calculations; 4) Calculate the real-time risk index by combining dynamic risks and static risks, classify the risk levels, adaptively adjust according to the degree of deviation of the real-time risk from the safety threshold, control the adjustment rate, balance the weights of sudden risks and long-term trends, and prevent a single indicator from dominating the decision-making; 5) Conduct multi-source data collection, evaluate the operation difficulty, calculate the action combination entropy value, expand the static weight Bayesian network, and integrate the evaluation results into the risk assessment model; 6) Extract the scene-sensitive features, expand the static weight Bayesian network, compare the predicted life and the actual damage data, and dynamically update the model parameters; 7) Send the analyzed working state data of the crane to the operator for data interaction.

2. The method for intelligently controlling a crane according to claim 1, wherein: In step 1), adaptive Kalman filtering is performed on the collected data to eliminate sensor noise, time synchronization is performed through multi-source data spatio-temporal registration, and filtering iteration is carried out.

3. A method for intelligently controlling a crane according to claim 1, characterized in that: In step 2), a state equation and an observation equation are constructed, a crane coordinate system is established, multi-sensor data fusion is used to eliminate installation errors, the absolute coordinates of the origin of the sensor coordinate system are obtained, and a working state model is built based on the processed collected data.

4. A method for intelligent control of a crane according to claim 1, characterized in that: In step 3), initial weight values are assigned to the indicators that affect the operation safety of the crane according to the historical accident data. By collecting the data of each monitoring indicator in real time, the dynamic response weights are calculated and updated in real time, and static feature reference weights are assigned to different sensor data. When major repairs, component replacements or major changes in the working environment occur to the crane, the weights are re-evaluated and adjusted, and a relationship model is constructed based on the dynamic response weights and static feature weights.

5. A method for intelligent control of a crane according to claim 1, characterized in that: In step 4), a Bayesian network causal relationship diagram is established based on the static weights. When major maintenance or environmental changes occur, the update of the Bayesian network parameters is triggered, and the maintenance cost is minimized under the constraint of the particle swarm optimization algorithm, so as to balance cost and safety.

6. A method for intelligently controlling a crane according to claim 1, characterized in that: In step 5), the model complexity score is calculated through a quantitative calculation model, and the hierarchical clustering method is used to divide it into 3 levels to establish a work shift grouping; the grouping is integrated into the dynamic weight calculation, and grouping nodes of operation complexity and work shifts are added to the Bayesian network to optimize the risk assessment model, misidentifying environmental interference as equipment failure and causing mis-shutdown.

7. A method for intelligent control of a crane according to claim 1, characterized in that: In step 6), the real-time risk assessment model introduces a scene-specific risk index, expands the static weight Bayesian network, adds root nodes of construction stage and environmental complexity, establishes a physical-data fusion model and a vibration fatigue damage model to enhance the life prediction model, simulates under extreme working conditions, compares the predicted life and the actual damage data, and dynamically updates the model parameters; In step 7), when the crane working data is greater than the first threshold, it is determined that the current tower crane working mode is normal; when the crane working data is less than the first threshold, it is determined that the current crane working mode is abnormal, and the data of working state control and maintenance is sent to the background server.

8. A method for intelligent control of a crane according to claim 5, characterized in that: Standardize the extracted features, set a time window to divide the continuous data stream into time series segments, and input them into a bidirectional LSTM neural network. The forward layer captures the historical risk accumulation, the backward layer captures the future abnormal trend, dynamically assigns weights to the features at each time step, and outputs the real-time risk index.

9. A method for intelligent control of a crane according to claim 6, characterized in that: Monitor the load weight and volume in real time. The inclinometer records the path curvature radius and the operation action sequence. The environmental sensor collects the wind speed and temperature. Perform Z-score standardization on the continuous variables and 0-1 normalization on the discrete variables.

10. A system for intelligent control of a crane, characterized in that, It includes sensors and inclinometer records. The working state data is collected through the sensors, parsed according to the collected data, and a working state model is built. Calculate the dynamic response weight and static feature weight of the corresponding data, calculate the real-time risk index by combining the dynamic risk and static risk, and classify the risk level. The sensors and inclinometer perform multi-source data collection, expand the static weight Bayesian network, extract the scene-sensitive features, dynamically update the model parameters, and send the working state data to the operator for data interaction.

Images