Crane abnormity monitoring method and application thereof in crane work control

Abstract

The invention discloses a crane abnormity monitoring method and application thereof in crane work control, and relates to the technical field of crane abnormity monitoring. Comprising the following steps: arranging target sensing equipment to obtain a sensing equipment set, creating a data acquisition layer based on the sensing equipment set, and performing multi-source data acquisition based on the data acquisition layer to acquire crane associated monitoring information so as to obtain an acquired data set; and performing data division based on the acquired data set to obtain first division data and second division data. According to the method, causal relationship modeling is carried out by utilizing the obtained multi-modal data, dynamic matching is carried out according to the modeling result and the real-time working scene of the crane, important parts in the real-time scene are output, the classified data are optimized and updated, and data monitoring is carried out again according to the corresponding monitoring method for the optimized data, so that the data monitoring accuracy is improved. The key data can be monitored under the condition that the data load is reduced.

Description

Technical Field

[0001] This invention relates to the field of crane anomaly monitoring technology, specifically to crane anomaly monitoring methods and their application in crane operation control. Background Technology

[0002] A crane is a special lifting device that uses mechanical power or human power to vertically lift and horizontally move heavy objects through a specific combination of mechanisms. Crane anomaly monitoring refers to the use of sensor technology, data acquisition systems, communication technology, and intelligent analysis algorithms to collect, transmit, process, and analyze key parameters during crane operation in real time or periodically. This is to identify potential deviations from normal operating conditions, thereby enabling fault warnings, condition assessments, predictive maintenance, and safety assurance. Most existing anomaly monitoring methods involve real-time sensor monitoring, such as vibration sensors that monitor stress changes in key structures like the main beam and outriggers, and sensors that monitor abnormal vibrations in slewing bearings, gearboxes, and motor bearings.

[0003] The crane anomaly monitoring method, system, and application disclosed in patent publication number CN119461072A integrates environmental monitoring data with crane-related parameter monitoring data and status monitoring data. When a crane malfunctions, the system analyzes this data to roughly determine whether the cause of the abnormality is due to internal factors or sudden environmental changes. This provides assistance to back-end management personnel or on-site operators in anomaly detection and offers targeted troubleshooting directions for maintenance personnel. Furthermore, by establishing a database containing anomaly data sets, this solution can match real-time environmental and operational monitoring data with these anomaly data sets during subsequent crane operation monitoring. This provides advance warnings for potential crane malfunctions and offers operators a reference for risk avoidance during operation.

[0004] When the above-mentioned and similar technical solutions are used to monitor the anomalies of cranes, the anomaly monitoring of cranes involves multiple mechanical structures and components. When all components and mechanical structures are monitored in an all-round way, a large amount of useless data will be generated, which will affect the effective analysis of the data. When only key components are monitored, the sensors are prone to "false diagnosis" due to the failure of monitoring equipment such as sensors or environmental influences, resulting in the neglect of the coupling relationship of multiple signals and poor anomaly monitoring effect. Summary of the Invention

[0005] The purpose of this invention is to provide a method for monitoring abnormalities in cranes and its application in crane operation control, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a crane anomaly monitoring method, comprising: The target sensing devices are deployed to obtain a set of sensing devices. A data acquisition layer is created based on the set of sensing devices. Multi-source data is collected based on the data acquisition layer to obtain crane-related monitoring information, thereby obtaining the collected dataset. Data is divided based on the collected dataset to obtain first-division data and second-division data. A first monitoring method is set based on the first-division data, and a second monitoring method is set based on the second-division data. Data monitoring of the crane is performed based on the first monitoring method and the second monitoring method. Causal relationship modeling is performed based on the collected dataset. Using structured causal model and time-series causal inference technology, the causal relationship and coupling degree between signals in the collected dataset are modeled as a dynamic causal network. The dynamic causal network is then encoded and corrected to obtain causal model terms. The dynamic results are output based on the causal model terms to obtain output data terms. The first and second monitoring methods are then adaptively adjusted accordingly to obtain the first and second monitoring adjustment terms. Anomaly monitoring is then performed based on the first and second monitoring adjustment terms, thereby achieving dynamic monitoring and adjustment based on causal relationships without comprehensive monitoring.

[0007] Furthermore, the crane-related monitoring information includes sensor data and operating condition data, and the method for acquiring the dataset includes: Acquire the working data of the target crane, including process information and machine information; obtain dynamic correlation location information based on the working data; obtain comparative change information based on the dynamic correlation location information; and obtain an information change set, which includes at least one information change item corresponding to the dynamic correlation location information. Based on the information change set, the adapted sensors are obtained to obtain the target sensor set, which includes at least one target sensor item. Based on the target sensor set, a placement threshold is set, which corresponds to each target sensor item. The placement threshold is the placement distance value. Based on the placement threshold, the target sensor set is installed to acquire the crane's sensing data and operating condition data, thus obtaining the collected dataset.

[0008] Furthermore, the method for acquiring the target sensor set includes: Set at least two information stability levels to obtain stability level items, and set reference level data based on the stability level items, including the number of sensing devices; Based on the stability level items, set up corresponding stability scores, including working hours score and maintenance frequency score, to obtain at least two stability score ranges; Based on the stable scoring range, the stability score is determined for the dynamic associated location corresponding to the information change set. The stability score values ​​of different dynamic associated locations are obtained, and the corresponding stability level items are obtained. At the same time, the comparison level data is adapted to obtain different numbers of sensing devices for different associated locations, thus obtaining the target sensor set.

[0009] Furthermore, the crane-related monitoring information also includes environmental data, and the methods for acquiring the dataset include: Based on the information change set, environmental impact data associated with the information change set is obtained through data acquisition to obtain an environmental impact set, which includes at least one environmental impact item corresponding to the information change item. Based on the environmental impact items, obtain the environmental impact coefficient, which is a percentage value. Then, classify and sort the environmental impact items based on the environmental impact coefficient. A judgment threshold is set, which is a fixed percentage value. The classification and sorting results are filtered and judged based on the judgment threshold to obtain the retained sorting results and obtain the retained environment items. Based on the retained environment items, the control environment acquisition device is obtained, and environmental data is obtained based on the control environment acquisition device, thus obtaining the collected dataset.

[0010] Furthermore, the first segmentation data is key data, and the second segmentation data is non-key data. The methods for obtaining the first segmentation data and the second segmentation data include: Obtain the correlation data between the collected dataset and the crane-related monitoring information, establish a risk-oriented priority matrix including the severity of failure, create at least two failure levels based on the severity of failure, and obtain the failure level item; The collected dataset is classified into key failures and non-key failures based on the failure level item. Key failures are used as key data and non-key failures are used as non-key data, resulting in the first and second classification data.

[0011] Furthermore, the first monitoring method and the second monitoring method include: Based on the first and second partition data, the corresponding warning thresholds are obtained, resulting in the first warning threshold item and the second warning threshold item. Based on the first warning threshold item and the second warning threshold item, at least two first warning values ​​and second warning values ​​are set. The first warning values ​​and the second warning values ​​are combined with the first warning threshold item and the second warning threshold item respectively to perform first warning threshold division and second warning threshold division. Based on the divided second warning threshold, segmented warnings are performed to obtain the second monitoring method. Based on the first warning threshold, a partition value is set, which is a fixed time value. The partition increase information is obtained based on the partition value, and an increase warning is issued. Based on the first warning threshold after division, a segmented warning is issued, resulting in a segmented warning. The first monitoring method is obtained by combining the increase warning and the segmented warning.

[0012] Furthermore, the method for constructing the causal model terms includes: Based on the acquired dataset, time alignment and sampling rate unification are performed to determine the reference time granularity. High-frequency signals are downsampled, low-frequency signals are upsampled and preserved as is, and time lag variables are introduced into the model. Alignment strategies are used to align multi-source data and retain the measurement of alignment error to obtain the processed dataset. The dataset is grouped into component group, operating condition group, environment group, and sensor health group. Semantic labels are assigned to the signals in each group. Key components and core signals are selected as manifest variables, and other signals are selected as candidate variables. The causal relationship between the manifest variables and candidate variables is obtained to obtain a relational connection graph. The relational connection graph is used as training data to train the model and obtain causal model terms.

[0013] Furthermore, the method for obtaining the output data item includes: Obtain real-time scene information of the target crane, including real-time process information and real-time environment information, to obtain real-time scene items; Based on the real-time scene item, the key component list in the current scene is output through the causal model item, thereby obtaining the output data item.

[0014] Furthermore, the methods for obtaining the first monitoring adjustment term and the second monitoring adjustment term include: Based on the output data items, data is removed from the first data division and data is entered into the second data division to obtain the first data modification item and the second data modification item. The first monitoring method and the second monitoring method are adjusted by comparison based on the first data modification item and the second data modification item, thereby obtaining the first monitoring adjustment item and the second monitoring adjustment item.

[0015] The application of crane anomaly monitoring methods in crane operation control utilizes the aforementioned crane anomaly monitoring methods, including: The system acquires crane-related monitoring information to obtain a dataset. It then sets up a first monitoring method and a second monitoring method to monitor the crane data. The system models the causal relationships and coupling degree between signals in the dataset as a dynamic causal network to obtain causal model terms. Based on the causal model terms, it outputs dynamic results and adaptively adjusts the first and second monitoring methods to monitor anomalies.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This crane anomaly monitoring method and its application in crane operation control acquires sensor data and operational data of the crane by arranging target sensing devices on the crane, while also acquiring environmental data. Utilizing multimodal data as a basis, the method classifies and categorizes the data, setting corresponding monitoring methods for each category. This avoids data interference and data storage burden caused by comprehensive monitoring. Furthermore, it uses the acquired multimodal data to model causal relationships, dynamically matching the modeling results with the real-time operating scenario of the crane to output key components in the real-time scenario. The categorized data is then optimized and updated, and the optimized data is monitored again using the corresponding monitoring methods. This achieves dynamic monitoring and adjustment based on causal relationships without comprehensive monitoring, enabling the monitoring of key data while reducing data load. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall process of the present invention; Figure 2 This is a schematic diagram of the data acquisition process of the present invention; Figure 3 This is a schematic diagram illustrating the relationship between the information stability level and the control level data, and the control stability score of the present invention. Figure 4 This is a schematic diagram of the output data item acquisition process of the present invention; Figure 5 This is a schematic diagram of the acquisition process for the first and second monitoring and adjustment items of the present invention. Detailed Implementation

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

[0019] Cranes are indispensable pieces of equipment in industrial production, and their safe and stable operation is of paramount importance. Effective anomaly monitoring of cranes can detect potential faults in advance and prevent major accidents. However, in actual operation, cranes have complex structures involving various mechanical components and parts, such as hoisting mechanisms, traveling mechanisms, luffing mechanisms, slewing mechanisms, as well as wire ropes, pulleys, and hooks. While comprehensive monitoring of all components and mechanical structures can undoubtedly cover potential fault points to the greatest extent, this "large and comprehensive" monitoring strategy often generates a large amount of redundant data. This useless data not only increases the burden of data storage and processing but also interferes with the effective analysis of key data, reduces monitoring efficiency, and may even mask true anomaly signals. The technical solution provided in this application, through... Target sensing devices are deployed on the crane to acquire sensor data and operational data, as well as environmental data. Multimodal data is used as a basis for classification, and corresponding monitoring methods are set for each category. This avoids data interference and data storage burden caused by comprehensive monitoring. Furthermore, causal relationship modeling is performed using the acquired multimodal data. The modeling results are dynamically matched with the real-time operating scene of the crane to output key components in the real-time scene. The classified data is optimized and updated, and data monitoring is performed again based on the corresponding monitoring methods after optimization. This achieves the effect of dynamic monitoring and adjustment based on causal relationships without comprehensive monitoring, enabling monitoring of key data while reducing data load. Simultaneously, ... Figure 1 As shown, it includes steps S100-S600.

[0020] Step S100: Arrange the target sensing devices to obtain a set of sensing devices, create a data acquisition layer based on the set of sensing devices, and collect multi-source data based on the data acquisition layer to obtain crane-related monitoring information, thereby obtaining a collection dataset.

[0021] It is important to note that, such as Figure 2 As shown, the crane-related monitoring information includes sensor data and operating condition data. The method for acquiring the dataset includes: acquiring the operating data of the target crane, including process information and machine information; acquiring dynamic correlation location information based on the operating data; acquiring comparison change information based on the dynamic correlation location information to obtain an information change set, which includes at least one information change item corresponding to the dynamic correlation location information; acquiring compatible sensors based on the information change set to obtain a target sensor set, which includes at least one target sensor item; setting a placement threshold based on the target sensor set, where each placement threshold corresponds to a target sensor item and is a placement distance value; installing the target sensor set based on the placement threshold; acquiring the crane's sensor data and operating condition data to obtain the dataset.

[0022] Specifically, crane operating data includes process information and equipment information. For example, in the process of container loading and unloading, the crane's process information includes: ship positioning and alignment, precise positioning of the spreader, container lifting and moving, and container landing and unlocking. The corresponding equipment information includes: wire rope, hoisting motor, brake, height encoder, traction drum, etc. Therefore, position information can be dynamically correlated through operating data. For example, the wire rope is correlated with the end of the boom, hook, winch, and brake, while the hoisting motor is correlated with the crane's power source, usually located below or to the side of the hoisting unit and directly driven by the winch, etc., and various dynamic relationships can be obtained. The system obtains comparative change information from the location of the crane, such as load information and deformation information of the boom end and hook, and temperature and vibration information of the winch and brake. Based on this information, it acquires suitable sensors, including temperature sensors, deformation sensors, and vibration sensors. Then, it sets placement thresholds, which are the optimal placement distances for each sensor. Therefore, the placement thresholds correspond to the target sensor items. For example, the optimal placement distance for temperature sensors is 1-1.5m, and for load sensors it is 1-1.2m. Based on the placement thresholds, the target sensor set is installed to acquire the crane's sensing data and operating condition data, resulting in a collected dataset.

[0023] It should be noted that the method for obtaining the target sensor set includes: setting at least two information stability levels to obtain stability level items; setting reference level data based on the stability level items, including the number of sensor devices; setting reference stability scores based on the stability level items, including working time scores and maintenance frequency scores, to obtain at least two stability score ranges; determining the stability score of the dynamically associated locations corresponding to the information change set based on the stability score ranges, obtaining the stability score values ​​of different dynamically associated locations, and thus obtaining the corresponding stability level items; simultaneously adapting the reference level data to obtain different numbers of sensor devices for different associated locations, thereby obtaining the target sensor set.

[0024] Specifically, such as Figure 3As shown, different components of the crane are affected by factors such as working time and maintenance frequency. When the working time is too long or the maintenance frequency is high, key monitoring is required, and multiple sensors need to monitor simultaneously. Therefore, three information stability levels are set: normal, moderate, and severe, with reference level data of 1, 2, and 3 respectively. Based on the stability level item, reference stability scores are set, where the score ranges for normal, moderate, and severe are 0-4, 5-8, and >8 respectively, resulting in three stability score ranges. For the corresponding working time and maintenance frequency, one year of working time is worth one point, and one maintenance frequency is worth one point. Based on the stability score range, the stability score of the dynamic associated position corresponding to the information change set is determined, and the stability score value of different dynamic associated positions is obtained, thus obtaining the corresponding stability level item. At the same time, reference level data is adapted to obtain different numbers of sensing devices for different associated positions, thus obtaining the target sensor set.

[0025] In the specific implementation process, it is now necessary to monitor the abnormality of a certain crane. The crane has been in service for two years, with the wire rope having undergone three maintenances and the chassis having undergone one repair. When it is necessary to obtain the tension, deformation, temperature, and vibration data of the wire rope and the chassis through load sensors, deformation sensors, temperature sensors, and vibration sensors, the wire rope is rated as five points and the chassis as three points according to the set scoring range. At this time, the information stability levels of the wire rope and the chassis are medium and normal, respectively, and the corresponding control level data are 2 and 1, respectively. That is, two temperature sensors and a vibration sensor are set to monitor the temperature and vibration data of the chassis, and one load sensor and a deformation sensor are set to obtain the tension and deformation data of the wire rope.

[0026] It is important to note that the crane-related monitoring information also includes environmental data. The methods for acquiring the dataset include: based on the information change set, acquiring environmental impact data associated with the information change set through data acquisition to obtain an environmental impact set, which includes at least one environmental impact item corresponding to the information change item; based on the environmental impact item, obtaining the environmental impact coefficient, which is a percentage value, and classifying and sorting the environmental impact items based on the environmental impact coefficient; setting a judgment threshold, which is a fixed percentage value, and filtering and judging the classification and sorting results based on the judgment threshold to obtain the retained sorting results, obtaining the retained environmental items; acquiring the control environment acquisition equipment based on the retained environmental items; acquiring environmental data based on the control environment acquisition equipment, and thus obtaining the collected dataset.

[0027] Specifically, environmental factors also influence the information change set. For example, the load information and deformation information of the boom end and hook, as well as the temperature and vibration information of the winch and brake, are affected by environmental factors such as temperature, humidity, and vibration. Furthermore, the magnitude of the impact varies among different environmental factors. For instance, environmental vibration has a relatively small impact on the temperature information of the winch and brake, while environmental temperature and humidity have a significant impact. Therefore, using big data acquisition, the first step is to obtain the influence coefficient of each environmental factor on the information change factor, and then classify and sort them according to the coefficient. For example, big data analysis reveals that the load information of the boom end and hook is affected by environmental temperature, humidity, and vibration. The influence coefficients of environmental vibration are 35%, 15%, and 50%, respectively, while the influence coefficients of hook deformation on environmental temperature, environmental humidity, and environmental vibration are 45%, 35%, and 20%, respectively. Therefore, environmental impact items are classified and sorted based on the environmental influence coefficients. According to the set judgment threshold of 40%, the classification and sorting results are filtered and judged based on the judgment threshold to obtain the retained sorting results and obtain the retained environmental items. For example, the retained sorting result for the load information of the boom end and hook is environmental vibration, while the retained result for hook deformation is environmental temperature. Based on the retained environmental items, the control environment acquisition equipment, such as vibration sensors and temperature sensors, is used to acquire environmental data, thereby obtaining the collected dataset.

[0028] Step S200: Divide the collected dataset into first and second partitioned data.

[0029] It should be noted that the first set of data is key data, and the second set of data is non-key data. The methods for obtaining the first and second sets of data include: obtaining the correlation data between the collected dataset and the crane-related monitoring information; establishing a risk-oriented priority matrix, including the severity of failures; creating at least two failure levels based on the severity of failures to obtain failure level items; classifying the collected dataset into key failures and non-key failures based on the failure level items; using key failures as key data and non-key failures as non-key data to obtain the first and second sets of data.

[0030] Specifically, the risk-oriented priority matrix obtains the severity of failure based on data category and creates two failure levels: minor and catastrophic. Minor represents non-critical data, while catastrophic represents critical data. Catastrophic means that it leads to the destruction of the whole machine or casualties, such as main beam breakage or wire rope breakage. Minor means high-frequency failure, which occurs ≥1 time per month on average, such as brake overheating or bearing wear.

[0031] Step S300: Set a first monitoring method based on the first division data, set a second monitoring method based on the second division data, and perform data monitoring on the crane based on the first monitoring method and the second monitoring method.

[0032] It should be noted that the first and second monitoring methods include: obtaining corresponding warning thresholds based on the first and second segmentation data to obtain a first warning threshold item and a second warning threshold item; setting at least two first and second warning values ​​based on the first and second warning threshold items, combining the first and second warning values ​​with the first and second warning threshold items respectively to perform first and second warning threshold segmentation, and performing segmented warnings based on the segmented second warning thresholds to obtain the second monitoring method; setting partition values ​​based on the first warning threshold items, with partition values ​​being fixed time values, obtaining partition increase information based on the partition values, performing increase warnings, performing segmented warnings based on the segmented first warning thresholds to obtain segmented warnings, and combining increase warnings and segmented warnings to obtain the first monitoring method.

[0033] Specifically, different datasets have different warning thresholds. For example, the warning threshold for load information at the boom end and hook is the tensile limit value, while the warning threshold for deformation information is the deformation threshold, and the warning threshold for temperature information of the winch and brake is the temperature threshold, resulting in a first warning threshold item and a second warning threshold item. Based on the first and second warning threshold items, two first warning values ​​and a second warning value are set, namely 90% and 80%, respectively. The first and second warning values ​​are then combined with the first and second warning threshold items to perform first and second warning threshold division, respectively. Based on the divided second warning threshold, The second monitoring method is derived from the segmented warning. For example, the warning threshold for the temperature information of the winch and brake is the temperature threshold, which is 100℃. At this time, according to the set first warning value and second warning value of 90% and 80%, the segmented warnings are 80℃ and 90℃ respectively. Based on the first warning threshold, a partition value is set, and the partition value is a fixed time value of 5s. Based on the partition value, the partition increase information is obtained, and the increase warning is issued. Based on the divided first warning threshold, the segmented warning is issued, resulting in a segmented warning. Based on the combination of the increase warning and the segmented warning, the first monitoring method is obtained. That is, the second monitoring method is a segmented warning, while the first monitoring method is a combination of segmented warning and increase warning.

[0034] Step S400: Based on the collected dataset, perform causal relationship modeling, model the causal relationship and coupling degree between each signal in the collected dataset as a dynamic causal network, and obtain the causal model terms.

[0035] It is important to note that the method for constructing causal model terms includes: performing time alignment and sampling rate unification based on the collected dataset, determining the baseline time granularity, downsampling high-frequency signals, upsampling low-frequency signals, preserving the original data, introducing time lag variables into the model, using alignment strategies to align multi-source data, and retaining the measurement of alignment error to obtain a processed dataset; grouping the processed dataset into component groups, operating condition groups, environmental groups, and sensor health groups, assigning semantic labels to each group of signals, selecting key components and core signals as manifest variables, and other signals as candidate variables, obtaining the causal relationship between the manifest and candidate variables, obtaining a relational connection graph, and using the relational connection graph as training data to train the model to obtain causal model terms.

[0036] In the specific implementation process, it is now necessary to monitor the abnormality of the hoisting mechanism of the port container crane. The first step is to collect data, as shown in Table 1: Table 1

[0037] Based on Table 1, time alignment and sampling rate unification were performed. The selection of the reference time granularity was based on the minimum cycle of the lifting action, approximately 40 seconds for a single hoisting operation, with a reference frequency of 10Hz, balancing high-frequency dynamics and storage efficiency. The resampling strategy included downsampling of high-frequency signals (motor current 10kHz-10Hz, requiring preservation of peak impact characteristics) and upsampling of low-frequency signals (wind speed 0.2Hz-10Hz, requiring cubic spline interpolation). Lag variables were then introduced, including brake thermal inertia and tension response delay; a 5-second lag corresponds to heat conduction. Alignment error was then measured, with a maximum time deviation of ±180ms. Due to GPS clock synchronization accuracy, residual vectors were recorded, and data was grouped and semantically labeled. The grouping structure definition included: graph TD A [Process Dataset] --> B [Component Group] A --> C [Working Condition Group] A --> D [Environment Group] A --> E [Sensor Health Group] B --> B1 [Wire Rope System] B1 --> B11(“Tension Signal_F_wire”) B1 --> B12(“Vibration signal_V_wire”) C --> C1 [Load / Unload Mode] C1 -->C11(“Load status_W_load:Unload / 20t / 40t”) C1 --> C12(“Action Phase_Phase: Acceleration / Constant Speed / Deceleration”) D -->D1("Ambient Wind_Wind: Wind Speed ​​+ Wind Direction") D -->D2(“Basic Vibration_Vib_ground: 0-80Hz spectrum”) E --> E1(“Signal Integrity_Health: Packet Loss Rate < 0.1%”) The selection logic for explicit variables directly reflects the core failure mode, while the selection logic for candidate variables considers potential influencing factors or auxiliary diagnostics. Finally, causal relationship discovery is performed using a causal skeleton extraction algorithm. This algorithm takes the sets of explicit and candidate variables as input, outputs an undirected causal skeleton, and determines the causal direction and time delay. Ultimately, a dynamic causal network is constructed. flowchart LR Wind([wind speed]) --τ=1.2s / 0.65-->F_wire Phase([Action Phase]) --τ=0s / 0.91-->I_motor I_motor([motor current]) --τ=8.3s / 0.78-->brake_temp Vib_ground([basic vibration]) --τ=3.5s / 0.52-->brake_temp F_wire([wire rope tension]) --τ=0.5s / 0.43-->brake_temp Then, a causal model is trained. The input layer consists of time series data with lag terms, the loss function is MAE plus a causal regularization term, the model used is the LSTM baseline model, and finally, causal model terms are generated.

[0038] Step S500: Output dynamic results based on the causal model terms to obtain output data terms.

[0039] It is important to note that, such as Figure 4 As shown, the method for obtaining output data items includes: obtaining real-time scene information of the target crane, including real-time process information and real-time environment information, to obtain real-time scene items; based on the real-time scene items, outputting a list of key components in the current scene through causal model items, thereby obtaining output data items.

[0040] Specifically, because the real-time process information and implementation environment information of cranes vary in different working scenarios—for example, the processes are completely different when hoisting and transferring goods—it is necessary to obtain the real-time scenario information of the target crane and output a list of key parts and the contribution of key signals to anomalies in the current scenario through causal model terms. For example, when hoisting goods, the list of key parts includes: wire rope, hoisting motor, brake, height encoder, and traction drum, while when transferring goods, the list of key parts includes: wire rope, brake, position encoder, steering bearing, steering base, etc. Therefore, it is necessary to output the corresponding information according to the real-time working scenario.

[0041] Step S600: Adaptively adjust the first monitoring method and the second monitoring method based on the output data items to obtain the first monitoring adjustment item and the second monitoring adjustment item.

[0042] It should be noted that anomaly monitoring is performed based on the first and second monitoring adjustment items, thereby achieving dynamic monitoring and adjustment based on causal relationships without comprehensive monitoring. The method for obtaining the first and second monitoring adjustment items includes: based on the output data items, data is removed from the first division data and data is entered into the second division data to obtain the first and second data modification items; the first and second monitoring methods are adjusted by comparison based on the first and second data modification items to obtain the first and second monitoring adjustment items.

[0043] Specifically, such as Figure 5 As shown, according to the set first monitoring method and second monitoring method, after the causal model item is output, some non-key data is removed from the first division data, and some key data is entered into the second division data, thus obtaining the first data modification item and the second data modification item; based on the first data modification item and the second data modification item, the first monitoring method and the second monitoring method are continued to be used for comparison and adjustment, thus obtaining the first monitoring adjustment item and the second monitoring adjustment item.

[0044] The application of the crane anomaly monitoring method in crane operation control uses the aforementioned crane anomaly monitoring method, including: acquiring crane-related monitoring information to obtain a data set; setting a first monitoring method and a second monitoring method to monitor the crane data; modeling the causal relationship and coupling degree between signals in the data set as a dynamic causal network to obtain causal model terms; outputting dynamic results based on the causal model terms; and adaptively adjusting the first monitoring method and the second monitoring method to perform anomaly monitoring.

[0045] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended embodiments and their equivalents.

Claims

1. Crane anomaly monitoring methods, including: The target sensing devices are deployed to obtain a set of sensing devices. A data acquisition layer is created based on the set of sensing devices. Multi-source data is collected based on the data acquisition layer to obtain crane-related monitoring information, thereby obtaining the collected dataset. Its characteristic is that it further includes: Data is divided based on the collected dataset to obtain first-division data and second-division data. A first monitoring method is set based on the first-division data, and a second monitoring method is set based on the second-division data. Data monitoring of the crane is performed based on the first monitoring method and the second monitoring method. Causal relationship modeling is performed based on the collected dataset. Using structured causal model and time-series causal inference technology, the causal relationship and coupling degree between signals in the collected dataset are modeled as a dynamic causal network. The dynamic causal network is then encoded and corrected to obtain causal model terms. The dynamic results are output based on the causal model terms to obtain output data terms. The first and second monitoring methods are then adaptively adjusted accordingly to obtain the first and second monitoring adjustment terms. Anomaly monitoring is then performed based on the first and second monitoring adjustment terms, thereby achieving dynamic monitoring and adjustment based on causal relationships without comprehensive monitoring.

2. The crane anomaly monitoring method according to claim 1, characterized in that: The crane-related monitoring information includes sensor data and operating condition data, and the methods for acquiring the dataset include: Acquire the working data of the target crane, including process information and machine information; obtain dynamic correlation location information based on the working data; obtain comparative change information based on the dynamic correlation location information; and obtain an information change set, which includes at least one information change item corresponding to the dynamic correlation location information. Based on the information change set, the adapted sensors are obtained to obtain the target sensor set, which includes at least one target sensor item. Based on the target sensor set, a placement threshold is set, which corresponds to each target sensor item. The placement threshold is the placement distance value. Based on the placement threshold, the target sensor set is installed to acquire the crane's sensing data and operating condition data, thus obtaining the collected dataset.

3. The crane anomaly monitoring method according to claim 2, characterized in that: The method for acquiring the target sensor set includes: Set at least two information stability levels to obtain stability level items, and set reference level data based on the stability level items, including the number of sensing devices; Based on the stability level items, set up corresponding stability scores, including working hours score and maintenance frequency score, to obtain at least two stability score ranges; Based on the stable scoring range, the stability score is determined for the dynamic associated location corresponding to the information change set. The stability score values ​​of different dynamic associated locations are obtained, and the corresponding stability level items are obtained. At the same time, the comparison level data is adapted to obtain different numbers of sensing devices for different associated locations, thus obtaining the target sensor set.

4. The crane anomaly monitoring method according to claim 2, characterized in that: The crane-related monitoring information also includes environmental data, and the methods for acquiring the dataset include: Based on the information change set, environmental impact data associated with the information change set is obtained through data acquisition to obtain an environmental impact set, which includes at least one environmental impact item corresponding to the information change item. Based on the environmental impact items, obtain the environmental impact coefficient, which is a percentage value. Then, classify and sort the environmental impact items based on the environmental impact coefficient. A judgment threshold is set, which is a fixed percentage value. The classification and sorting results are filtered and judged based on the judgment threshold to obtain the retained sorting results and obtain the retained environment items. Based on the retained environment items, the control environment acquisition device is obtained, and environmental data is obtained based on the control environment acquisition device, thus obtaining the collected dataset.

5. The crane anomaly monitoring method according to claim 1, characterized in that: The first data segment is key data, and the second data segment is non-key data. The methods for obtaining the first and second data segments include: Obtain the correlation data between the collected dataset and the crane-related monitoring information, establish a risk-oriented priority matrix including the severity of failure, create at least two failure levels based on the severity of failure, and obtain the failure level item; The collected dataset is classified into key failures and non-key failures based on the failure level item. Key failures are used as key data and non-key failures are used as non-key data, resulting in the first and second classification data.

6. The crane anomaly monitoring method according to claim 1, characterized in that: The first monitoring method and the second monitoring method include: Based on the first and second partition data, the corresponding warning thresholds are obtained, resulting in the first warning threshold item and the second warning threshold item. Based on the first warning threshold item and the second warning threshold item, at least two first warning values ​​and second warning values ​​are set. The first warning values ​​and the second warning values ​​are combined with the first warning threshold item and the second warning threshold item respectively to perform first warning threshold division and second warning threshold division. Based on the divided second warning threshold, segmented warnings are performed to obtain the second monitoring method. Based on the first warning threshold, a partition value is set, which is a fixed time value. The partition increase information is obtained based on the partition value, and an increase warning is issued. Based on the first warning threshold after division, a segmented warning is issued, resulting in a segmented warning. The first monitoring method is obtained by combining the increase warning and the segmented warning.

7. The crane anomaly monitoring method according to claim 1, characterized in that: The method for constructing the causal model terms includes: Based on the acquired dataset, time alignment and sampling rate unification are performed to determine the reference time granularity. High-frequency signals are downsampled, low-frequency signals are upsampled and preserved as is, and time lag variables are introduced into the model. Alignment strategies are used to align multi-source data and retain the measurement of alignment error to obtain the processed dataset. The dataset is grouped into component group, operating condition group, environment group, and sensor health group. Semantic labels are assigned to the signals in each group. Key components and core signals are selected as manifest variables, and other signals are selected as candidate variables. The causal relationship between the manifest variables and candidate variables is obtained to obtain a relational connection graph. The relational connection graph is used as training data to train the model and obtain causal model terms.

8. The crane anomaly monitoring method according to claim 1, characterized in that: The method for obtaining the output data items includes: Obtain real-time scene information of the target crane, including real-time process information and real-time environment information, to obtain real-time scene items; Based on the real-time scene item, the key component list in the current scene is output through the causal model item, thereby obtaining the output data item.

9. The crane anomaly monitoring method according to claim 1, characterized in that: The methods for obtaining the first monitoring adjustment item and the second monitoring adjustment item include: Based on the output data items, data is removed from the first data division and data is entered into the second data division to obtain the first data modification item and the second data modification item. The first monitoring method and the second monitoring method are adjusted by comparison based on the first data modification item and the second data modification item, thereby obtaining the first monitoring adjustment item and the second monitoring adjustment item.

10. The application of crane anomaly monitoring methods in crane operation control, characterized by: The crane anomaly monitoring method according to any one of claims 1-9 includes: The system acquires crane-related monitoring information to obtain a dataset. It then sets up a first monitoring method and a second monitoring method to monitor the crane data. The system models the causal relationships and coupling degree between signals in the dataset as a dynamic causal network to obtain causal model terms. Based on the causal model terms, it outputs dynamic results and adaptively adjusts the first and second monitoring methods to monitor anomalies.

Images