A method and system for monitoring dynamic stress of a metal structure of a bridge crane

Abstract

The application relates to the technical field of hoisting equipment safety monitoring, and discloses a bridge crane metal structure dynamic stress monitoring method and system, wherein the method comprises the following steps: a three-dimensional model of a bridge crane metal structure is established, a typical working condition load is applied for finite element analysis, a high stress area is identified, and a distributed optical fiber sensing network is arranged; wavelength change data of each sensor in the distributed optical fiber sensing network is collected, the wavelength change data is converted into strain values and is subjected to denoising pretreatment, equivalent stress data is calculated based on the strain values; the equivalent stress data of each sensor is input into a preset space-time convolutional neural network model, stress anomaly probability is output, and a graded early warning is triggered according to a preset threshold; stress amplitude and mean value are extracted based on the equivalent stress data, cumulative fatigue damage is calculated, the remaining life of the bridge crane metal structure is predicted, and a maintenance decision is generated. The application can realize stress anomaly positioning, fatigue damage evaluation and remaining life prediction, and provides intelligent decision support for safe operation of the crane.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring technology for lifting equipment, and in particular to a method and system for dynamic stress monitoring of the metal structure of a bridge crane. Background Technology

[0002] Bridge cranes, widely used in industrial production, port logistics, and equipment manufacturing, rely on their metal structures as the core components bearing the entire load, directly determining the equipment's operational safety and service life. During long-term service, the metal structure continuously endures alternating loads, impact loads, and eccentric loads, making it prone to developing microscopic fatigue cracks in stress concentration areas. The gradual propagation of these cracks can lead to structural fractures, collapses, and other major safety accidents, causing not only equipment damage and production stoppages but also potential casualties, resulting in severe economic losses and social impacts. Therefore, real-time and accurate monitoring of the stress state of the bridge crane's metal structure, timely detection of potential safety hazards, and preventative maintenance are of significant engineering importance and application value.

[0003] Currently, stress detection and health monitoring technologies for bridge crane metal structures still have many shortcomings. Traditional strain gauge technology uses a single-point measurement method, which cannot obtain the overall stress distribution of the structure and is prone to missing hidden stress concentration areas. Its installation relies on welding or adhesive bonding, which can damage the protective coating on the metal structure's surface. Furthermore, under long-term vibration and harsh environments, the sensors are prone to detachment and failure, resulting in high maintenance costs. While resistance strain gauges are widely used, they have extremely poor resistance to electromagnetic interference. The crane's working environment contains numerous motors, frequency converters, and other strong electromagnetic devices, leading to severe signal distortion. Simultaneously, resistance strain gauges suffer from significant temperature drift, requiring frequent on-site calibration and failing to meet the needs of long-term online monitoring.

[0004] Visual inspection methods identify defects such as cracks by capturing images of structural surfaces with high-resolution cameras. However, the results are greatly affected by factors such as ambient light, dust, and vibration, resulting in low reliability in complex industrial environments. Furthermore, this method can only detect macroscopic surface defects and cannot detect internal stress changes or early fatigue damage, and its real-time performance is insufficient for dynamic monitoring requirements. The application of existing fiber optic sensing technology in crane monitoring is still immature. Sensor deployment schemes do not consider the dynamic load characteristics and stress distribution patterns of cranes, failing to effectively capture transient stress changes caused by impact loads. Simultaneously, there is a lack of targeted data demodulation and processing algorithms, resulting in low signal-to-noise ratios, large measurement errors, and most methods only provide simple stress value displays, lacking intelligent stress anomaly identification, fatigue damage assessment, and remaining life prediction capabilities, thus failing to provide comprehensive decision support for equipment safety management. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes a dynamic stress monitoring method and system for the metal structure of a bridge crane. Based on a distributed optical fiber sensor network and machine learning algorithms, a high-density optical fiber sensor array captures the dynamic stress distribution of key components such as the main beam and end beams in real time. Combined with finite element simulation and deep learning models, it can achieve stress anomaly location, fatigue damage assessment, and remaining life prediction, providing intelligent decision support for the safe operation of the crane.

[0006] The technical solution adopted in this invention is as follows: A method for monitoring dynamic stress in the metal structure of a bridge crane includes: A three-dimensional model of the metal structure of the bridge crane was established, and finite element analysis was performed by applying typical working loads to identify high stress areas and deploy a distributed optical fiber sensor network. Wavelength change data from each sensor in a distributed optical fiber sensor network are collected, converted into strain values, and preprocessed for noise reduction. Equivalent stress data are then calculated based on the strain values. The equivalent stress data from each sensor is input into a preset spatiotemporal convolutional neural network model, which outputs the stress anomaly probability and triggers graded early warnings according to preset thresholds. Stress amplitude and mean values ​​are extracted based on equivalent stress data, cumulative fatigue damage is calculated, the remaining life of the bridge crane metal structure is predicted, and maintenance decisions are generated.

[0007] Furthermore, the process of establishing a three-dimensional model of the bridge crane's metal structure, applying typical working condition loads for finite element analysis, identifying high-stress areas, and deploying a distributed fiber optic sensor network includes: A three-dimensional model of the metal structure of the bridge crane, including the main beam, end beams and legs, was constructed. Typical working conditions, including rated lifting capacity, impact load and off-center load, were applied for finite element analysis to obtain the principal stress direction and the maximum equivalent stress distribution. High-stress areas are identified at the mid-span lower flange of the main beam, the connection between the end beam and the support leg, and the track support point. A distributed optical fiber sensing network composed of optical fiber Bragg grating sensors is deployed in the high-stress areas. The optical fibers are laid out in a spiral staggered manner to make the sensitive axis directions of adjacent sensors orthogonal.

[0008] Furthermore, the deployment of a distributed optical fiber sensing network composed of optical fiber Bragg grating sensors in the high-stress region includes: A bend-resistant fiber array integrating multiple fiber Bragg grating sensors is used. The fiber Bragg grating sensors are encapsulated in a metal matrix composite material, and the encapsulated fiber Bragg grating sensors are fixed to the surface of the metal structure by laser micro-welding process.

[0009] Furthermore, the process of acquiring wavelength change data from each sensor in the distributed fiber optic sensor network, converting it into strain values ​​and performing noise reduction preprocessing, and calculating equivalent stress data based on the strain values ​​includes: Wavelength offset and ambient temperature data of each sensor are acquired synchronously using a multi-channel dynamic demodulator; based on wavelength offset, strain sensitivity coefficient and temperature sensitivity coefficient, strain value at the corresponding position of each sensor is calculated. Wavelet thresholding denoising method is used to filter the original strain data to remove vibration noise; strain values ​​of orthogonally arranged sensor groups are decoupled to obtain each stress component, and equivalent von Mises stress is calculated as equivalent stress data.

[0010] Furthermore, the calculation of strain values ​​at corresponding locations of each sensor based on wavelength offset, strain sensitivity coefficient, and temperature sensitivity coefficient includes:

[0011] in, Let be the strain value at the location corresponding to the i-th sensor. Let be the wavelength offset of the i-th sensor. The strain sensitivity coefficient, For temperature sensitivity coefficient, This refers to the change in ambient temperature. The strain values ​​based on the orthogonally arranged sensor group are decoupled to obtain each stress component, and the equivalent von Mises stress is calculated as equivalent stress data, including:

[0012] in, For equivalent von Mises stress, The stress component is obtained by decoupling the strain values ​​from the orthogonally arranged sensor group.

[0013] Furthermore, the step of inputting the equivalent stress data from each sensor into a preset spatiotemporal convolutional neural network model, outputting the stress anomaly probability, and triggering a graded early warning according to a preset threshold includes: The equivalent stress data at continuous time points of each sensor are used as input to the spatiotemporal convolutional neural network model. The spatial correlation features between multiple sensors are extracted through the spatial convolutional layer of the spatiotemporal convolutional neural network model, and the temporal features of stress change are extracted through the temporal convolutional layer of the spatiotemporal convolutional neural network model. The classifier of the spatiotemporal convolutional neural network model outputs the stress anomaly probability of each monitoring location, and triggers the corresponding level of early warning signal according to the preset anomaly probability threshold.

[0014] Furthermore, the training method for the spatiotemporal convolutional neural network model includes: Equivalent stress data of bridge cranes under normal working conditions, overload conditions, and structural damage conditions were collected to construct a training dataset. The original spatiotemporal convolutional neural network model was trained using the Adam optimizer and cross-entropy loss function. The model parameters were adjusted using the validation set to obtain the trained spatiotemporal convolutional neural network model.

[0015] Furthermore, the step of extracting stress amplitude and mean values ​​based on equivalent stress data and calculating cumulative fatigue damage includes: The equivalent stress data were statistically analyzed using the rainflow counting method to extract the amplitude of stress cycles. with the mean Based on the modified Goodman fatigue damage model, and combining the material's ultimate strength and material constants, the cumulative fatigue damage D at corresponding locations of the bridge crane's metal structure is calculated.

[0016] in, Let be the actual number of cycles under the i-th stress level. Let be the theoretical fatigue life cycle number under the i-th stress level; i This is the stress level number, representing the i-th stress level grade divided according to stress amplitude and mean. Let i be the stress amplitude of level i. The average stress of level i is... C represents the ultimate tensile strength of the material; m is a material constant.

[0017] Furthermore, the prediction of the remaining life of the bridge crane's metal structure and the generation of maintenance decisions include: Calculate the remaining life of the metal structure of the bridge crane :

[0018] in, This represents the running time. When the cumulative fatigue damage D reaches the preset threshold, preventive maintenance recommendations for the corresponding monitoring location are generated; based on the remaining lifespan and the warning level, a maintenance work order containing the inspection location and maintenance priority is generated.

[0019] A dynamic stress monitoring system for the metal structure of a bridge crane, comprising: A distributed optical fiber sensing network is deployed in the high-stress area of ​​the metal structure of a bridge crane, comprising an optical fiber array composed of multiple fiber Bragg grating sensors. The multi-channel dynamic demodulator communicates with a distributed optical fiber sensor network to collect wavelength change data from each sensor and convert it into corresponding strain values. The edge computing unit communicates with the multi-channel dynamic demodulator to perform noise reduction preprocessing on strain data and calculate equivalent stress data. The cloud-based analysis unit communicates with the edge computing unit and uses a spatiotemporal convolutional neural network to detect stress anomalies based on equivalent stress time-series data, calculate cumulative fatigue damage, and predict the remaining life of the structure. The visualization monitoring unit communicates and connects with the cloud analysis unit to display stress monitoring results, graded early warning information, and maintenance decisions.

[0020] The beneficial effects of this invention are as follows: 1. This invention utilizes a sensor optimization layout scheme driven by finite element simulation, combined with a spiral interlaced fiber optic layout, to accurately cover the high-stress areas of the metal structure of a bridge crane. Simultaneously, it achieves synchronous measurement of principal stress and shear stress components, comprehensively reflecting the complex stress state of the structure. The use of metal-based composite material encapsulation and laser micro-welding fixation effectively protects the fiber optic sensors from external environmental corrosion and mechanical damage, while ensuring efficient strain transmission. This significantly improves the long-term stability and reliability of the sensing system, solving the problems of traditional sensors being easily detached and damaged.

[0021] 2. This invention synchronously acquires wavelength and temperature data using a multi-channel dynamic demodulator, and combined with a built-in temperature compensation algorithm, effectively eliminates wavelength drift caused by changes in ambient temperature, thus improving the accuracy of strain measurement. It employs wavelet threshold denoising and adaptive Kalman filtering techniques to effectively suppress vibration noise during crane operation, significantly improving the signal-to-noise ratio. Furthermore, it utilizes a stress decoupling method based on orthogonal sensor groups and equivalent von Neumann filters. The Mises stress calculation provides an accurate and reliable data foundation for subsequent anomaly detection and fatigue analysis.

[0022] 3. The spatiotemporal convolutional neural network model constructed in this invention can simultaneously extract the spatial correlation features and temporal variation features of stress data from multiple sensors. Compared with traditional single feature extraction methods, it significantly improves the accuracy and robustness of stress anomaly detection. The hierarchical early warning mechanism can trigger corresponding level early warning signals according to different anomaly probabilities, which not only avoids unnecessary false alarms and shutdowns, but also ensures that serious safety hazards are dealt with in a timely manner, effectively improving the safety of equipment operation.

[0023] 4. This invention combines the rainflow counting method and the modified Goodman fatigue damage model to accurately calculate the cumulative fatigue damage of metal structures and make accurate predictions of remaining life based on the cumulative damage value. The system can automatically generate targeted preventive maintenance suggestions and graded maintenance work orders, realizing intelligent management of the entire life cycle of crane metal structures from real-time monitoring and anomaly warning to life prediction and maintenance decision-making. This not only reduces the operation and maintenance costs of the equipment, but also effectively extends the service life of the equipment, providing comprehensive technical support for the safe and efficient operation of bridge cranes. Attached Figure Description

[0024] Figure 1 This is a diagram of the architecture of a dynamic stress monitoring system for the metal structure of a bridge crane according to Embodiment 2 of the present invention.

[0025] Figure 2 This is a flowchart of a dynamic stress monitoring method for the metal structure of a bridge crane according to Embodiment 2 of the present invention. Detailed Implementation

[0026] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments are now described. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention; that is, the described embodiments are only a part of the embodiments of the invention, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0027] Example 1 This embodiment provides a method for monitoring the dynamic stress of the metal structure of a bridge crane, including: A three-dimensional model of the metal structure of the bridge crane was established, and finite element analysis was performed by applying typical working loads to identify high stress areas and deploy a distributed optical fiber sensor network. Wavelength change data from each sensor in a distributed optical fiber sensor network are collected, converted into strain values, and preprocessed for noise reduction. Equivalent stress data are then calculated based on the strain values. The equivalent stress data from each sensor is input into a preset spatiotemporal convolutional neural network model, which outputs the stress anomaly probability and triggers graded early warnings according to preset thresholds. Stress amplitude and mean values ​​are extracted based on equivalent stress data, cumulative fatigue damage is calculated, the remaining life of the bridge crane metal structure is predicted, and maintenance decisions are generated.

[0028] It should be noted that this method guides sensor deployment through finite element simulation, enabling precise coverage of high-risk areas of the structure and avoiding ineffective sensor deployment. By combining distributed fiber optic sensing technology with deep learning algorithms, it achieves real-time and comprehensive monitoring and intelligent analysis of stress in metal structures, enabling timely detection of stress anomalies and fatigue life assessment, thus providing reliable technical support for the safe operation of cranes.

[0029] Preferably, a three-dimensional model of the bridge crane's metal structure, including the main beam, end beams, and legs, is constructed. Finite element analysis is performed by applying typical working condition loads, including rated lifting capacity, impact load, and eccentric load, to obtain the principal stress direction and the maximum equivalent stress distribution. High-stress areas are identified at the lower flange of the main beam mid-span, the connection between the end beams and the legs, and the track support points. A distributed optical fiber sensing network composed of optical fiber Bragg grating sensors is deployed in the high-stress areas, and the optical fibers are arranged in a spiral staggered manner to make the sensitive axis directions of adjacent sensors orthogonal.

[0030] Specifically, when constructing the 3D model, key structural details such as the main beam web, flanges, end beam welds, and outrigger connections are preserved to ensure that the geometric accuracy of the model is consistent with the actual structure. In the finite element analysis, the normal operation of the crane under rated lifting capacity, the impact load during lifting and lowering, and the stress state when the cargo is unevenly loaded are simulated to calculate the direction, magnitude, and distribution range of the principal stress under different working conditions. Based on the analysis results, the lower flange area with the greatest bending at mid-span of the main beam, the multi-directional stress superposition area at the connection between the end beam and the outrigger, and the local stress concentration area where the track contacts the main beam are marked. When deploying fiber optic sensors in these high-stress areas, the optical fibers are fixed to the structural surface in a spiral staggered manner so that the sensitive axes of two adjacent sensors are perpendicular to each other, enabling simultaneous measurement of stress components in different directions.

[0031] It should be noted that this method, through finite element analysis simulating various typical working conditions, can comprehensively identify high-stress areas that may occur in the structure during actual operation; the spiral staggered arrangement allows the sensors to simultaneously capture principal stress and shear stress components, improving the accuracy and comprehensiveness of stress measurement and avoiding the loss of stress information caused by measurement in a single direction.

[0032] More preferably, a bend-resistant fiber array integrating multiple fiber Bragg grating sensors is used, the fiber Bragg grating sensors are encapsulated in a metal matrix composite material, and the encapsulated fiber Bragg grating sensors are fixed to the surface of the metal structure by laser micro-welding process.

[0033] Specifically, fiber arrays are fabricated using fiber substrates with excellent bending resistance, and multiple fiber Bragg grating sensors are integrated on a single fiber to form a distributed sensing unit. The sensors are encapsulated using metal-based composite materials, ensuring a tight bond between the sensor and the encapsulation material without gaps or air bubbles during the encapsulation process. The encapsulated sensor units are then welded onto the surface of a metal structure using laser micro-welding equipment. During welding, the laser output parameters and welding time are controlled to avoid damage to the sensor performance and the metal structure substrate.

[0034] It should be noted that the integrated anti-bending fiber array can reduce the number of optical fibers and wiring complexity, and improve the reliability of the system; the metal matrix composite encapsulation can effectively protect the sensor from external environmental corrosion and mechanical damage, while ensuring that strain can be efficiently transferred from the structure to the sensor; the laser micro-welding process fixes the sensor firmly and reliably without damaging the surface coating and mechanical properties of the metal structure, making it suitable for the long-term vibration working environment of cranes.

[0035] Preferably, the wavelength offset and ambient temperature data of each sensor are synchronously acquired by a multi-channel dynamic demodulator; based on the wavelength offset, strain sensitivity coefficient and temperature sensitivity coefficient, the strain value at the corresponding position of each sensor is calculated; the original strain data is filtered by wavelet threshold denoising method to remove vibration noise; based on the strain value decoupled from the orthogonally arranged sensor group, each stress component is obtained, and the equivalent von Mises stress is calculated as equivalent stress data.

[0036] Specifically, each fiber of the distributed optical fiber sensor network is connected to the corresponding channel of the multi-channel dynamic demodulator. The sampling parameters of the demodulator are set so that it can synchronously acquire wavelength signals and ambient temperature signals from all sensors. Based on the pre-calibrated sensor strain sensitivity coefficient and temperature sensitivity coefficient, combined with the acquired wavelength offset and temperature change, the actual strain value at each sensor location is calculated. A suitable wavelet basis function and decomposition level are selected to perform wavelet threshold denoising on the raw strain data to filter out high-frequency vibration noise generated during crane operation. Using the strain values ​​measured in different directions by the orthogonally arranged sensor group, the normal stress component and shear stress component at that location are calculated using the mechanical decoupling formula. Then, the equivalent von Mises stress formula is substituted to calculate the equivalent stress that comprehensively reflects the stress state of the structure.

[0037] It should be noted that synchronous acquisition of wavelength and temperature data enables real-time temperature compensation, eliminating the influence of ambient temperature changes on strain measurement results; wavelet threshold denoising method can effectively suppress vibration noise and improve the signal-to-noise ratio of strain data; by decoupling the orthogonal sensor group to obtain each stress component and calculating the equivalent von Mises stress, the complex stress state of the structure can be accurately assessed, providing a reliable data foundation for subsequent anomaly detection and fatigue analysis.

[0038] More preferably, based on the wavelength shift, strain sensitivity coefficient, and temperature sensitivity coefficient, the strain value at the corresponding location of each sensor is calculated, including:

[0039] in, Let be the strain value at the location corresponding to the i-th sensor. Let be the wavelength offset of the i-th sensor. The strain sensitivity coefficient, For temperature sensitivity coefficient, This represents the change in ambient temperature.

[0040] More preferably, the stress components are obtained by decoupling the strain values ​​of the orthogonally arranged sensor group, and the equivalent von Mises stress is calculated as equivalent stress data, including:

[0041] in, For equivalent von Mises stress, The stress component is obtained by decoupling the strain values ​​from the orthogonally arranged sensor group.

[0042] It should be noted that the sensitivity coefficient of the sensor was obtained through experimental calibration, which ensured the accuracy of strain calculation; the strain calculation formula that takes into account the influence of temperature can effectively eliminate the measurement error caused by temperature drift; the stress decoupling method based on the generalized Hooke's law and the equivalent von Mises stress calculation can accurately reflect the stress state of the structure under complex stress, and provide a scientific basis for the safety assessment of the structure.

[0043] Preferably, the equivalent stress data of each sensor at continuous time points are used as input to the spatiotemporal convolutional neural network model. The spatial correlation features between multiple sensors are extracted through the spatial convolutional layer of the spatiotemporal convolutional neural network model, and the temporal features of stress change are extracted through the temporal convolutional layer of the spatiotemporal convolutional neural network model. The classifier of the output layer of the spatiotemporal convolutional neural network model outputs the stress anomaly probability of each monitoring location, and triggers the corresponding level of early warning signal according to the preset anomaly probability threshold.

[0044] Specifically, the equivalent von Mises stress data from all sensors collected at multiple consecutive time points are organized into a two-dimensional tensor and used as input to a spatiotemporal convolutional neural network model. The spatial convolutional layer of the model performs convolution operations on the stress data from different sensors at the same time point through convolution kernels to extract the spatial correlation features between different monitoring locations, reflecting the distribution law of stress on the structure. The temporal convolutional layer of the model performs convolution operations on the stress data from the same sensor at different time points through convolution kernels to extract the temporal features of stress changes over time, reflecting the dynamic trend of stress change. The extracted spatiotemporal features are input into a fully connected layer and a classifier to output the probability of stress anomalies at each monitoring location. Multiple different anomaly probability thresholds are preset. When the anomaly probability at a certain location reaches the corresponding threshold, an early warning signal of the corresponding level is automatically triggered, and the abnormal location is marked on the monitoring interface.

[0045] It should be noted that the spatiotemporal convolutional neural network model can simultaneously extract the spatial distribution features and temporal variation features of stress data. Compared with traditional single feature extraction methods, it can more accurately identify stress anomaly patterns. The hierarchical early warning mechanism can take different countermeasures according to the severity of the anomaly, avoid unnecessary downtime, and ensure that serious anomalies can be dealt with in a timely manner.

[0046] More preferably, equivalent stress data of bridge cranes under normal working conditions, overload conditions, and structural damage conditions are collected to construct a training dataset; the original spatiotemporal convolutional neural network model is trained using the Adam optimizer and cross-entropy loss function, and the model parameters are adjusted through the validation set to obtain the trained spatiotemporal convolutional neural network model.

[0047] Specifically, monitoring systems were installed on multiple bridge cranes of different models and service lives to collect equivalent stress data under normal operation, overload tests, and artificially simulated structural damage conditions. The collected data underwent preprocessing, including data cleaning, labeling, and partitioning. The labeled dataset was divided into training, validation, and test sets according to a certain ratio. An initial model of a spatiotemporal convolutional neural network was constructed, setting hyperparameters such as the number of network layers, kernel size, and activation function. The model was trained using the training set data, with the Adam optimizer used to update the model parameters, employing the cross-entropy loss function as the optimization objective. During training, the model's performance was evaluated using the validation set data, and the learning rate and hyperparameters were adjusted based on the validation set accuracy until the model converged and achieved the preset accuracy requirement on the test set, resulting in a fully trained model.

[0048] It should be noted that by collecting stress data from multiple working conditions and multiple devices to construct a training dataset, the generalization ability of the model can be improved, making it applicable to different types of bridge cranes; using the Adam optimizer and cross-entropy loss function can accelerate the convergence speed of the model and improve the classification accuracy of the model; adjusting the model parameters through the validation set can effectively prevent the model from overfitting and ensure the reliability of the model in practical applications.

[0049] Preferably, the equivalent stress data are statistically analyzed using the rainflow counting method to extract the amplitude of stress cycles. with the mean Based on the modified Goodman fatigue damage model, and combining the material's ultimate strength and material constants, the cumulative fatigue damage D at corresponding locations of the bridge crane's metal structure is calculated.

[0050] in, Let be the actual number of cycles under the i-th stress level. Let be the theoretical fatigue life cycle number under the i-th stress level; i This is the stress level number, representing the i-th stress level grade divided according to stress amplitude and mean. Let i be the stress amplitude of level i. The average stress of level i is... C represents the ultimate tensile strength of the material; m is a material constant.

[0051] Specifically, the continuously collected equivalent von Mises stress time series data is input into the rainflow counting program. The program automatically identifies the start and end points of stress cycles and calculates the amplitude and mean of each stress cycle. Based on the mechanical properties of the materials used in the crane's metal structure, the ultimate strength and fatigue characteristic constants of the materials are determined. The amplitude and mean of each stress cycle are substituted into the modified Goodman fatigue damage model to calculate the fatigue damage caused to the structure by that stress cycle. According to the Miner linear cumulative damage criterion, the fatigue damage of all stress cycles is accumulated to obtain the cumulative fatigue damage D at the monitoring location.

[0052] It should be noted that the rainflow counting method can accurately count the number of cycles and amplitude distribution of alternating stress, and is a commonly used and effective method in fatigue analysis; the modified Goodman fatigue damage model considers the influence of mean stress on fatigue life, and the calculation results are more accurate compared with the traditional fatigue damage model; the cumulative fatigue damage calculated based on the Miner linear cumulative damage criterion can intuitively reflect the degree of fatigue damage of the structure and provide a basis for predicting the remaining life.

[0053] Preferably, the remaining lifespan of the bridge crane's metal structure The calculation method is as follows:

[0054] in, This represents the running time.

[0055] When the cumulative fatigue damage D reaches the preset threshold, preventive maintenance recommendations for the corresponding monitoring location are generated; based on the remaining lifespan and the warning level, a maintenance work order containing the inspection location and maintenance priority is generated.

[0056] Specifically, the cumulative operating time T of the bridge crane from its commissioning to the current moment is recorded. Combined with the calculated cumulative fatigue damage D, the remaining service life (RUL) of the structure is obtained by substituting it into the remaining service life calculation formula. A warning threshold for cumulative fatigue damage is preset. When the cumulative fatigue damage at a certain monitoring location reaches the threshold, a preventive maintenance suggestion for that location is automatically generated, including inspection content, inspection methods, and precautions. Taking into account the length of the remaining service life and the warning level of stress anomalies, the maintenance tasks are divided into different priorities, and a maintenance work order containing the inspection location, maintenance content, priority, and suggested completion time is generated and pushed to relevant management personnel.

[0057] It should be noted that the remaining life prediction method based on cumulative fatigue damage and operating time is simple to calculate and the results are reliable, providing a scientific basis for equipment maintenance planning; the hierarchical maintenance work order system can rationally allocate maintenance resources, prioritize maintenance tasks for high-risk parts, improve maintenance efficiency, and reduce the risk of equipment failure.

[0058] Accordingly, this embodiment also provides a dynamic stress monitoring system for the metal structure of a bridge crane, including: A distributed optical fiber sensing network is deployed in the high-stress area of ​​the metal structure of a bridge crane, comprising an optical fiber array composed of multiple fiber Bragg grating sensors. The multi-channel dynamic demodulator communicates with a distributed optical fiber sensor network to collect wavelength change data from each sensor and convert it into corresponding strain values. The edge computing unit communicates with the multi-channel dynamic demodulator to perform noise reduction preprocessing on strain data and calculate equivalent stress data. The cloud-based analysis unit communicates with the edge computing unit and uses a spatiotemporal convolutional neural network to detect stress anomalies based on equivalent stress time-series data, calculate cumulative fatigue damage, and predict the remaining life of the structure. The visualization monitoring unit communicates and connects with the cloud analysis unit to display stress monitoring results, graded early warning information, and maintenance decisions.

[0059] Specifically, the distributed optical fiber sensor network is installed on key parts of the bridge crane, such as the main beam, end beams, and outriggers, according to the deployment scheme determined by finite element simulation, to sense the strain changes of the structure in real time. The multi-channel dynamic demodulator is installed in the crane's electrical control cabinet and connected to the distributed optical fiber sensor network via optical fiber patch cords. It converts the collected optical signals into electrical signals and demodulates the wavelength data. The edge computing unit is connected to the multi-channel dynamic demodulator via industrial Ethernet, receives the wavelength data in real time, performs preprocessing, calculates the equivalent stress data, and uploads it to the cloud. The cloud analysis unit is deployed on a cloud server, runs deep learning models and fatigue analysis algorithms, and completes stress anomaly detection and remaining life prediction. The visualization monitoring unit is installed on a computer in the monitoring room and connected to the cloud analysis unit via the network. It displays the monitoring results in the form of three-dimensional stress cloud maps, data curves, and early warning pop-ups, and supports historical data query and maintenance work order management.

[0060] It should be noted that this system adopts a layered architecture design, which pushes data preprocessing tasks down to the edge computing unit, thereby reducing the amount of data transmission and computing pressure in the cloud and improving the real-time performance of the system. The cloud analysis unit can realize centralized management and data analysis of multiple devices, which facilitates unified monitoring and maintenance. The visualization monitoring unit provides an intuitive human-machine interface, enabling managers to keep abreast of the crane's operating status and safety conditions.

[0061] Example 2 This embodiment provides a method and system for dynamic stress monitoring of the metal structure of a bridge crane. By optimizing the fiber optic sensor deployment scheme, improving data demodulation accuracy, and constructing an intelligent early warning model, it achieves comprehensive monitoring and safety assessment of the stress in the crane's metal structure. The specific details are as follows.

[0062] I. System Architecture like Figure 1 As shown, the dynamic stress monitoring system for the metal structure of the bridge crane in this embodiment includes: 1. Distributed Fiber Optic Sensing Network: Composed of multiple fiber optic sensors, deployed along the surfaces of key structures such as the crane's main beam and end beams. Wavelength-modulated fiber Bragg grating (FBG) sensors are used to achieve distributed stress measurement. A bend-resistant FBG array is employed, integrating 50-100 FBG sensors (reference value) per fiber, spaced 5-10cm (reference value), and deployed along high-stress areas such as the crane's main beam web, end beam welds, and outrigger connections. The sensors are encapsulated in a metal-based composite material and directly fixed to the steel structure surface using laser micro-welding technology, ensuring a strain transfer efficiency ≥95%.

[0063] 2. Data Acquisition and Multi-Channel Dynamic Demodulator: Utilizing a high-speed data acquisition card and fiber optic demodulator, it acquires wavelength variation data from the fiber optic sensor in real time and converts it into stress values. The dynamic demodulator supports 32-channel parallel demodulation, with a wavelength resolution of ±0.1 pm, a sampling frequency of 1 kHz, and a dynamic range of ±5000 με. A built-in temperature compensation algorithm eliminates the influence of ambient temperature on wavelength drift (compensation accuracy ±0.5℃).

[0064] 3. Edge Computing Unit: Equipped with an FPGA chip, it processes raw data in real time and calculates parameters such as stress gradient and spectral characteristics. It integrates an adaptive Kalman filter to suppress noise caused by crane vibration (signal-to-noise ratio improvement ≥20dB).

[0065] 4. Cloud-based analytics platform: A spatiotemporal convolutional neural network (ST-CNN) model is built based on the PyTorch framework. It takes multi-sensor time-series data as input and outputs a stress anomaly probability map. The Miner-Palmgren correction model is used to calculate cumulative fatigue damage and predict remaining life (error ≤ 5%).

[0066] 5. Visual Monitoring Platform: Displays stress distribution, anomaly warnings, and fatigue life assessment results in real time via a GUI interface, supporting data export and report generation. It uses a WebGL 3D visualization engine to display stress cloud maps in real time, highlighting high-risk areas (e.g., areas with stress values ​​≥ 80% of the material's yield strength are marked in red). It supports historical data review, warning log export, and maintenance work order generation.

[0067] II. Method and Flow like Figure 2 As shown, the dynamic stress monitoring method for the metal structure of a bridge crane in this embodiment includes: Step 1: Sensor Optimization and Deployment Driven by Finite Element Simulation 1. Establish a three-dimensional model of the crane's metal structure, apply typical working loads (rated lifting capacity, impact load, off-center load) for finite element analysis, and identify the principal stress direction and the maximum equivalent stress region.

[0068] 2. Based on the simulation results, sensors should be preferentially deployed in the following areas: The lower flange at mid-span of the main beam bears the maximum bending stress.

[0069] Connection between end beam and support leg: a region of superimposed multi-directional stress.

[0070] Track support point: Localized area of ​​stress concentration.

[0071] 3. Use a spiral staggered arrangement method: The optical fiber is spirally wound along the length of the main beam at a 30° inclination angle (reference value) to ensure that the sensitive axes of adjacent sensors are orthogonal and to cover the principal stress and shear stress components.

[0072] Step 2: Dynamic stress data acquisition and preprocessing 1. The wavelength change data of the fiber optic sensor is acquired in real time by the fiber optic demodulator, with a sampling frequency ≥100Hz (reference value).

[0073] 2. Obtain fiber wavelength offset in real time using a demodulator. (i is the sensor number), converted to strain :

[0074] in, The strain sensitivity coefficient, For temperature sensitivity coefficient, Temperature is measured synchronously by a temperature sensor.

[0075] 3. Perform wavelet thresholding denoising on the original data (sym8 wavelet basis can be used, decomposition level 5) to remove high-frequency vibration noise.

[0076] 4. Calculate the equivalent Von Mises stress :

[0077] in, The values ​​are obtained by decoupling the measurements from orthogonally arranged sensor groups.

[0078] Step 3: Stress Anomaly Detection Based on Deep Learning 1. Construct a spatiotemporal convolutional neural network (ST-CNN) model: Input layer: Stress sequence of 1000 consecutive time points (reference value) for each sensor (time window can be set to 1s).

[0079] Spatial convolutional layer: 3×3 convolutional kernels extract spatial correlation features from the sensor.

[0080] Temporal convolutional layer: 1×5 convolutional kernels capture temporal patterns of stress changes.

[0081] Output layer: Softmax classifier outputs the probability of normal / abnormal conditions.

[0082] 2. Model Training: Dataset: Stress data of 10 cranes under normal operating conditions, overload, and structural damage conditions were collected, totaling 100,000 samples.

[0083] Optimizer: Adam algorithm, initial learning rate 0.001, cross-entropy loss function.

[0084] 3. Online testing: Input stress data in real time, and trigger a three-level warning when the probability of an anomaly is greater than 0.9 (reference value) (e.g., yellow warning: probability 0.9 - 0.95; orange warning: 0.95 - 0.98; red warning: >0.98).

[0085] Step 4: Fatigue Life Prediction and Maintenance Decision 1. Extract stress amplitude using the rainflow counting method. with the mean Calculate the corrected Goodman fatigue damage D:

[0086] in, Let be the actual number of cycles under the i-th stress level. Let be the theoretical fatigue life cycle number under the i-th stress level; i This is the stress level number, representing the i-th stress level grade divided according to stress amplitude and mean. Let i be the stress amplitude of level i. The average stress of level i is... C represents the ultimate tensile strength of the material; m This is a material constant. Specifically, , .

[0087] 2. When cumulative damage When the reference value is reached, preventative maintenance recommendations are generated, prompting for focused inspection of the corresponding areas.

[0088] 3. Remaining lifespan Estimate:

[0089] in, This represents the running time.

[0090] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

[0091] It should be noted that, for the sake of simplicity, the foregoing method embodiments are described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

Claims

1. A method for monitoring dynamic stress in the metal structure of a bridge crane, characterized in that, include: A three-dimensional model of the metal structure of the bridge crane was established, and finite element analysis was performed by applying typical working loads to identify high stress areas and deploy a distributed optical fiber sensor network. Wavelength change data from each sensor in a distributed optical fiber sensor network are collected, converted into strain values, and preprocessed for noise reduction. Equivalent stress data are then calculated based on the strain values. The equivalent stress data from each sensor is input into a preset spatiotemporal convolutional neural network model, which outputs the stress anomaly probability and triggers graded early warnings according to preset thresholds. Stress amplitude and mean values ​​are extracted based on equivalent stress data, cumulative fatigue damage is calculated, the remaining life of the bridge crane metal structure is predicted, and maintenance decisions are generated.

2. The method for dynamic stress monitoring of bridge crane metal structures according to claim 1, characterized in that, The process of establishing a three-dimensional model of the bridge crane's metal structure, applying typical working loads for finite element analysis, identifying high-stress areas, and deploying a distributed fiber optic sensor network includes: A three-dimensional model of the metal structure of the bridge crane, including the main beam, end beams and legs, was constructed. Typical working conditions, including rated lifting capacity, impact load and off-center load, were applied for finite element analysis to obtain the principal stress direction and the maximum equivalent stress distribution. High-stress areas are identified at the mid-span lower flange of the main beam, the connection between the end beam and the support leg, and the track support point. A distributed optical fiber sensing network composed of optical fiber Bragg grating sensors is deployed in the high-stress areas. The optical fibers are laid out in a spiral staggered manner to make the sensitive axis directions of adjacent sensors orthogonal.

3. The method for dynamic stress monitoring of the metal structure of a bridge crane according to claim 2, characterized in that, The deployment of a distributed fiber optic sensing network composed of fiber Bragg grating sensors in a high-stress region includes: A bend-resistant fiber array integrating multiple fiber Bragg grating sensors is used. The fiber Bragg grating sensors are encapsulated in a metal matrix composite material, and the encapsulated fiber Bragg grating sensors are fixed to the surface of the metal structure by laser micro-welding process.

4. The method for dynamic stress monitoring of the metal structure of a bridge crane according to claim 1, characterized in that, The process involves collecting wavelength change data from each sensor in the distributed fiber optic sensor network, converting it into strain values, performing noise reduction preprocessing, and calculating equivalent stress data based on the strain values, including: Wavelength offset and ambient temperature data of each sensor are acquired synchronously using a multi-channel dynamic demodulator; based on wavelength offset, strain sensitivity coefficient and temperature sensitivity coefficient, strain value at the corresponding position of each sensor is calculated. Wavelet thresholding denoising method is used to filter the original strain data to remove vibration noise; strain values ​​of orthogonally arranged sensor groups are decoupled to obtain each stress component, and equivalent von Mises stress is calculated as equivalent stress data.

5. The method for dynamic stress monitoring of the metal structure of a bridge crane according to claim 4, characterized in that, The calculation of strain values ​​at corresponding locations of each sensor based on wavelength offset, strain sensitivity coefficient, and temperature sensitivity coefficient includes: in, Let be the strain value at the location corresponding to the i-th sensor. Let be the wavelength offset of the i-th sensor. The strain sensitivity coefficient, For temperature sensitivity coefficient, This refers to the change in ambient temperature. The strain values ​​based on the orthogonally arranged sensor group are decoupled to obtain each stress component, and the equivalent von Mises stress is calculated as equivalent stress data, including: in, For equivalent von Mises stress, The stress component is obtained by decoupling the strain values ​​from the orthogonally arranged sensor group.

6. The method for dynamic stress monitoring of the metal structure of a bridge crane according to claim 1, characterized in that, The process of inputting equivalent stress data from each sensor into a preset spatiotemporal convolutional neural network model, outputting stress anomaly probability, and triggering graded early warnings according to preset thresholds includes: The equivalent stress data at continuous time points of each sensor are used as input to the spatiotemporal convolutional neural network model. The spatial correlation features between multiple sensors are extracted through the spatial convolutional layer of the spatiotemporal convolutional neural network model, and the temporal features of stress change are extracted through the temporal convolutional layer of the spatiotemporal convolutional neural network model. The classifier of the spatiotemporal convolutional neural network model outputs the stress anomaly probability of each monitoring location, and triggers the corresponding level of early warning signal according to the preset anomaly probability threshold.

7. The method for dynamic stress monitoring of the metal structure of a bridge crane according to claim 6, characterized in that, The training method for the spatiotemporal convolutional neural network model includes: Equivalent stress data of bridge cranes under normal working conditions, overload conditions, and structural damage conditions were collected to construct a training dataset. The original spatiotemporal convolutional neural network model was trained using the Adam optimizer and cross-entropy loss function. The model parameters were adjusted using the validation set to obtain the trained spatiotemporal convolutional neural network model.

8. The method for dynamic stress monitoring of the metal structure of a bridge crane according to claim 1, characterized in that, The step of extracting stress amplitude and mean values ​​based on equivalent stress data and calculating cumulative fatigue damage includes: The equivalent stress data were statistically analyzed using the rainflow counting method to extract the amplitude of stress cycles. with the mean Based on the modified Goodman fatigue damage model, and combining the material's ultimate strength and material constants, the cumulative fatigue damage D at corresponding locations of the bridge crane's metal structure is calculated. in, Let be the actual number of cycles under the i-th stress level. Let be the theoretical fatigue life cycle number under the i-th stress level; i This is the stress level number, representing the i-th stress level grade divided according to stress amplitude and mean. Let i be the stress amplitude of level i. The average stress of level i is... C represents the ultimate tensile strength of the material; m is a material constant.

9. The method for dynamic stress monitoring of the metal structure of a bridge crane according to claim 8, characterized in that, The method of predicting the remaining lifespan of the bridge crane's metal structure and generating maintenance decisions includes: Calculate the remaining life of the metal structure of the bridge crane : in, This represents the running time. When the cumulative fatigue damage D reaches the preset threshold, preventive maintenance recommendations for the corresponding monitoring location are generated; based on the remaining lifespan and the warning level, a maintenance work order containing the inspection location and maintenance priority is generated.

10. A dynamic stress monitoring system for the metal structure of a bridge crane, employing the dynamic stress monitoring method for the metal structure of a bridge crane as described in claim 1, characterized in that, The system includes: A distributed optical fiber sensing network is deployed in the high-stress area of ​​the metal structure of a bridge crane, comprising an optical fiber array composed of multiple fiber Bragg grating sensors. The multi-channel dynamic demodulator communicates with a distributed optical fiber sensor network to collect wavelength change data from each sensor and convert it into corresponding strain values. The edge computing unit communicates with the multi-channel dynamic demodulator to perform noise reduction preprocessing on strain data and calculate equivalent stress data. The cloud-based analysis unit communicates with the edge computing unit and uses a spatiotemporal convolutional neural network to detect stress anomalies based on equivalent stress time-series data, calculate cumulative fatigue damage, and predict the remaining life of the structure. The visualization monitoring unit communicates and connects with the cloud analysis unit to display stress monitoring results, graded early warning information, and maintenance decisions.

Images