In-service steel crane beam fatigue risk monitoring method and system

Abstract

The invention belongs to the technical field of building component crack identification and monitoring, and provides an in-service steel crane beam fatigue risk monitoring method aiming at the technical problem that fatigue crack monitoring of an in-service steel crane beam is relatively difficult at present, and the method comprises the following steps: S1, obtaining data of a to-be-monitored steel crane beam, a high-risk area classification result of the steel crane beam is obtained; s2, acquiring a crack image; s3, according to the obtained crack image, obtaining a crack length estimation value at the current moment, and obtaining a crack length trend chart; s4, according to the crack length trend chart, obtaining the average crack development rate of each stage, and performing risk classification on the average crack development rate of each stage by using a preset crack rate-risk classification deep neural network to obtain the risk level of the fatigue crack risk part of the steel crane beam; and S5, when the risk level is a high risk level, generating early warning information. According to the invention, accurate positioning of fatigue crack high-risk parts and accurate risk assessment can be realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of building component crack identification and monitoring, and particularly relates to a fatigue risk monitoring method and system for in-service steel crane beams. Background Art

[0002] In the steel structure system of industrial buildings, steel crane beams, as key force-transmitting components, are subjected to dynamic cyclic loads for a long time. Their service performance directly affects the structural safety and production continuity of industrial plants. However, engineering practice has shown that due to the coupling of complex multiple factors (including but not limited to: the eccentric effect caused by asymmetric track layout, the effect of overhead crane braking force, differences in crane beam geometry, etc.), crane beam systems are prone to fatigue cracks in stress concentration areas. Such cracks have significant nonlinear expansion characteristics. When the crack growth rate exceeds the critical threshold, it may lead to catastrophic fracture accidents, seriously threatening industrial production safety. Therefore, fatigue cracks need to be monitored, but the current fatigue monitoring methods have the following technical problems:

[0003] 1. Difficulty in predicting crack initiation locations: Traditional standard formulas are derived based on ideal boundary conditions and cannot accurately represent the complex stress fields formed by the coupling of multiple factors under actual operating conditions. Research has shown that for crane girders of the same structural form, the maximum principal stress region shifts significantly under the coupling of different factors, resulting in significant discreteness in the crack initiation location. Traditional stress verification methods suffer from a mismatch between theoretical models and actual operating conditions.

[0004] 2. Limitations of manual inspection: Dust pollution in industrial environments, restrictions on working at height, and equipment obstruction make manual inspection difficult and costly. Manual inspections are subject to limitations such as strong reliance on subjective experience and low crack quantification accuracy, making them difficult to meet the precision requirements for crack growth rate monitoring.

[0005] 3. Bottlenecks in intelligent monitoring technology: Existing visual monitoring equipment faces three technical obstacles: (1) Dependence on external power supplies leads to low deployment feasibility, and 90% of industrial sites cannot provide stable power; (2) Battery life bottleneck, the average continuous working time of commercially available equipment is less than 6 months, which cannot meet the 5-year monitoring cycle required by ASTM E647; (3) The false alarm rate of image recognition algorithms is difficult to avoid under complex field noise, and is significantly affected by interference such as vibration noise and oil adhesion, making it difficult to meet the use requirements in real service environments.

[0006] 4. Lack of safety assessment system: The current specifications lack a quantitative correlation model between crack growth rate and remaining life of the structure. The warning error rate based on manual experience is over 40%, and it is impossible to establish a graded warning mechanism based on fracture mechanics. Summary of the Invention

[0007] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a fatigue risk monitoring method and system for in-service steel crane beams, which can achieve accurate positioning of high-risk areas of fatigue cracks, long-term monitoring, accurate risk assessment, and noise interference elimination.

[0008] To achieve the above object, the technical solution adopted by the present invention is:

[0009] A fatigue risk monitoring method for an in-service steel crane beam comprises the following steps:

[0010] S1. Obtain data of the steel crane beam to be monitored, and classify the obtained crane beam data using a pre-trained fatigue crack location deep neural network to obtain a classification result of high-risk areas of the steel crane beam;

[0011] S2. Deploy crack image acquisition terminals in high-risk areas based on the high-risk area classification results to acquire crack images;

[0012] S3. Obtain an estimated crack length at the current moment based on the acquired crack image, summarize the estimated length data, and obtain a crack length trend chart;

[0013] S4. Correcting the crack length on the crack length trend chart, using the corrected crack length trend chart to obtain the average crack growth rate at each stage, and using a preset crack rate-risk classification deep neural network to perform risk classification on the average crack growth rate at each stage to obtain the risk level of the fatigue crack risk location of the steel crane beam;

[0014] S5. When the risk level is high, generate warning information.

[0015] Preferably, in step S1, the pre-trained fatigue crack location deep neural network is constructed by the following method:

[0016] S11. Establish a fatigue cracking database for steel crane beams;

[0017] S12. Construct a first deep neural network, train it with data from a steel crane beam fatigue cracking database, obtain a fatigue crack location deep neural network model, and then modify the model to obtain a pre-trained fatigue crack location deep neural network.

[0018] Preferably, in step S12, the model correction method is as follows:

[0019] S121, Acquisition of prior knowledge in the field;

[0020] S122. Modify the model based on prior knowledge in the field to obtain a pre-trained fatigue crack location deep neural network.

[0021] Preferably, in step S2, the crack image acquisition terminal includes:

[0022] The main processor is responsible for scheduling and executing the terminal's core business logic, configuring image sensor parameters, performing crack image capture tasks, performing preliminary processing on raw images, and communicating with the microcontroller unit.

[0023] Communication module, used to upload the collected high-definition crack images and equipment status information to the remote server;

[0024] Image sensor, used to capture high-definition crack images in conjunction with a fill light;

[0025] Fill light, used to provide controllable light source for lighting compensation in dark or night environments;

[0026] FLASH memory, used to temporarily store high-definition crack images taken by the image sensor;

[0027] The microcontroller unit is used to control the power on and off of each module and to perform power-on self-test and task termination protocols.

[0028] Preferably, in step S4, the calculation formula for the average crack growth rate in each stage is:

[0029]

[0030] Among them, L d (t1) is the final displayed length of the crack at a certain stage in the crack length trend diagram, L d (t0) is the initial display length of a certain stage in the crack length trend diagram, Δ s t is the total time interval of this stage.

[0031] Preferably, in step S4, the preset crack rate-risk classification deep neural network is constructed by the following method:

[0032] S41. Establish a fatigue crack rate-risk database;

[0033] S42. Construct a second deep neural network and use the data in the fatigue crack rate-risk database for training to obtain a preset crack rate-risk classification deep neural network.

[0034] Preferably, the first deep neural network and the second deep neural network have the same structure, both including an input layer, multiple hidden layers, corresponding multiple Sigmoid activation layers and an output layer connected in sequence.

[0035] Preferably, in step S4, the displayed length of the crack length trend graph after correction at time t is:

[0036]

[0037] Among them, L d (t) is the length displayed on the crack length trend chart at time t, L c (t) is the recognition length at time t.

[0038] The present invention also provides a fatigue risk monitoring system for in-service steel crane beams, comprising:

[0039] Data acquisition module; used to obtain crack images through the crack image acquisition terminal and send the obtained crack images to the data processing module;

[0040] A data processing module is used to obtain an estimated value of the crack length at the current moment based on the acquired crack image, and to summarize the length estimation data to obtain a crack length trend chart;

[0041] The data post-processing module is used to obtain the average crack growth rate in each stage based on the crack length trend chart, and use the preset crack rate-risk classification deep neural network to classify the average crack growth rate in each stage, obtain the risk level of the fatigue crack risk area of ​​the steel crane beam, and generate early warning information when the risk level is high.

[0042] Preferably, the in-service steel crane beam fatigue risk monitoring system further comprises:

[0043] Data display module, used to display monitoring point distribution information, fatigue crack images, crack length identification results and early warning information;

[0044] The equipment control module is used to adjust the shooting interval and fill light intensity of the crack image acquisition terminal in real time;

[0045] The system management module is used to manage users, roles, organizations, and resource menus.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] (1) Compared with traditional manual experience judgment, the monitoring method proposed in the present invention can quickly and accurately predict high-risk areas for fatigue crack initiation by combining pre-trained fatigue crack location deep neural network with steel crane beam body data and overhead crane data. In addition, the crack rate-risk classification deep neural network is used to calculate the average development rate of each stage in combination with the crack length trend chart, realizing dynamic classification and graded early warning of fatigue crack risks. Therefore, the early warning accuracy of the present invention is greatly improved and the misjudgment rate is significantly reduced.

[0048] (2) The present invention integrates prior knowledge in the field and introduces prior knowledge in the field (the impact of track eccentricity on stress distribution) into model training. By calibrating the prediction probability through the correction function, the applicability and accuracy of the deep neural network in actual working conditions are improved.

[0049] (3) The crack image acquisition terminal proposed in the present invention has dynamic power management and abnormal interrupt processing mechanisms, which solves the problem of existing visual monitoring equipment relying on external power supplies and insufficient battery life. At the same time, through software optimization, the device response time is significantly shortened, and the device operating power consumption is further reduced to meet long-term monitoring needs.

[0050] (4) The present invention corrects the crack length and identifies the crack length data that does not conform to the statistical law as an outlier, which can effectively eliminate the interference of industrial environmental noise on the crack identification algorithm and ensure the accuracy and reliability of the crack length data. Since the correction method is based on statistical laws and physical constraints, the false alarm rate is significantly reduced.

[0051] (5) The fatigue risk monitoring system provided by the present invention integrates data acquisition, processing, post-processing, display, equipment control, and management modules, achieving automated management of the entire process from crack detection to risk assessment. The equipment control module in the present invention can dynamically adjust the shooting interval and fill light intensity according to lighting conditions and risk levels to further optimize the monitoring effect. In summary, the present invention significantly outperforms existing technologies in terms of crack location, risk assessment, monitoring efficiency, anti-interference capability, and system integration, providing an efficient, accurate, and reliable solution for fatigue risk monitoring of in-service steel crane beams. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 The plug-in type steel crane beam model provided by the present invention;

[0053] Figure 2 Finite element model of the plate-type crane beam established for the present invention; (a) is the model without eccentricity; (b) is the eccentric model when the track eccentricity is 30mm;

[0054] Figure 3 is the overall stress cloud diagram of working condition 1;

[0055] Figure 4 This is the stress cloud diagram at the support for working condition 1;

[0056] Figure 5 is the overall stress cloud diagram of working condition 2;

[0057] Figure 6 The stress cloud diagram at the support of working condition 2;

[0058] Figure 7 A framework diagram of the fatigue risk monitoring system for in-service steel crane beams provided by the present invention;

[0059] Figure 8 The crack length curve and crack development rate curve of high-risk areas;

[0060] Figure 9 This is a crack growth trend diagram in the present invention;

[0061] Figure 10 This is a diagram showing the actual application of the crack image acquisition terminal in the present invention. DETAILED DESCRIPTION

[0062] The following is a summary of the embodiments of the present invention. Figures 1 to 10 The technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0063] Example 1

[0064] An embodiment of the present invention provides a method for monitoring fatigue risk of an in-service steel crane beam, comprising the following steps:

[0065] S1. Obtain data of the steel crane beam to be monitored, and classify the obtained crane beam data using a pre-trained fatigue crack location deep neural network to obtain a classification result of high-risk areas of the steel crane beam;

[0066] In this embodiment, the data acquired for monitoring the steel crane beam includes the data of the steel crane beam itself, the data of the overhead crane supported by the steel crane beam, and the eccentricity data of the steel crane beam track. Based on the original design drawings of the steel crane beam, the data of the steel crane beam itself and the data of the overhead crane supported by the steel crane beam can be easily acquired. The eccentricity data of the steel crane beam track needs to be obtained through on-site testing. The testing method is as follows:

[0067] Use a tape measure to measure the distance h from the edge of the upper flange to the edge of the web of the steel crane beam on site f 1. Distance from upper flange edge to overhead crane track edge h f 2. Check the original design drawing to obtain the web width t of the steel crane beam w 1. Track width t w 2, then the orbital eccentricity value is O e for:

[0068]

[0069] If the orbital eccentricity is O e Greater than the web width t of the steel crane beam w 1, then condition A exists: orbital eccentricity.

[0070] In this embodiment, the pre-trained fatigue crack location deep neural network is constructed by the following method:

[0071] S11. Establish a fatigue cracking database for steel crane beams;

[0072] Based on years of experience in crane beam inspection projects, a large amount of fatigue cracking data on steel crane beams was obtained. This data was used as output labels, and the cracked areas of the steel crane beams were specifically divided into: the upper flange area, the support plug-end area, the support cover area, and the lower flange mid-span area. The main body data and overhead crane load data of crane beams with fatigue cracking were obtained by consulting the corresponding design drawings. The input features were the main body data of the steel crane beams and the overhead crane data supported by the steel crane beams, and a fatigue cracking database for steel crane beams was established.

[0073] S12. Construct a first deep neural network. The first deep neural network specifically adopts an existing neural network architecture. This embodiment does not involve improvements to the architecture. The first deep neural network is trained using data from a steel crane beam fatigue cracking database to obtain a fatigue crack location deep neural network model. The fatigue crack location deep neural network model is then modified to obtain a pre-trained fatigue crack location deep neural network. The first deep neural network in the present invention includes an input layer, multiple hidden layers, multiple Sigmoid activation layers, and an output layer connected in sequence. The hidden layers correspond one to one to the Sigmoid activation layers.

[0074] This invention utilizes prior knowledge in the field (the impact of track eccentricity on the support area of ​​plug-in steel crane beams) to modify the fatigue crack location deep neural network model. The stress concentration area in the support area of ​​plug-in steel crane beams is significantly affected by track eccentricity. Under the action of the overhead crane load, when there is no track eccentricity, the support stress concentration area is the sealing plate. When there is eccentricity, the support stress concentration areas occur in the sealing plate and the plug-in plate. Eccentricity can cause stress in the plug-in plate stress concentration area to increase by more than 30%. Therefore, it is necessary to use prior knowledge in the field to modify the model's prediction probability to form a pre-trained fatigue crack location deep neural network.

[0075] In the embodiment of the present invention, the method for model correction is specifically as follows:

[0076] S121. Acquisition of prior knowledge in the domain. The specific acquisition method is:

[0077] Obtain the morphological information of the plug-in steel crane beam and obtain the steel crane beam model, such as Figure 1 As shown, a finite element model is then established based on the steel crane beam model, which is divided into a model without track eccentricity and a model with track eccentricity, as shown in Figure 2As shown, the specific comparison working conditions of the two models are shown in Table 1 below. Then, finite element analysis is performed on the two models to obtain stress results, as shown in Table 2 below. At the same time, the changes in the stress concentration parts of the support area of ​​the plug-in steel crane beam under the two working conditions are compared. The results are as follows Figures 3 to 6 shown.

[0078] Table 1 Specific settings of the comparison conditions of the two models

[0079] Working conditions No track eccentricity Eccentricity 1.5 times the web thickness Working condition 1 √ Working condition 2 √

[0080] Table 2 Changes in stress concentration locations in the support area of ​​the plug-in steel crane beam under two working conditions

[0081] Working conditions Stress at sealing plate MPa Stress at the insert plate Mpa Working condition 1 144.3 86.7 Working condition 2 158.9 135.6

[0082] The results in Table 2 show that the presence of track eccentricity significantly increases the stress levels at the sealing plate and the inserting plate. Compared with the working condition 1 (no track eccentricity), the stress at the sealing plate in working condition 2 (with track eccentricity) increases from 144.3MPa to 158.9MPa, an increase of about 10.1%; the stress at the inserting plate increases from 86.7MPa to 135.6MPa, an increase of 56.4%. Figures 3 to 6 It can be seen that under the action of overhead crane load, when there is no track eccentricity, the stress concentration part of the support is the closing plate, and when there is eccentricity, the stress concentration part of the support appears in the closing plate and the insert plate. The above results show that the track eccentricity has a significant impact on the stress distribution in the support area of ​​the steel crane beam, especially the stress sensitivity of the insert plate is higher, and its stress increase is much greater than that of the closing plate. Eccentricity will cause the stress concentration part of the insert plate to increase by more than 30%.

[0083] S122. Modify the model based on prior knowledge in the field to obtain a pre-trained fatigue crack location deep neural network.

[0084] The embodiment of the present invention modifies the model through prior knowledge in the field and calibrates the data probability of the model, so that when condition A "track eccentricity" exists, the probability of positioning in the support area is increased.

[0085] Specifically, by training a calibration function, inputting the condition A and the original predicted probability p, the output corrected probability is:

[0086]

[0087] Among them, δ is the increment set according to prior knowledge.

[0088] S2. Deploy crack image acquisition terminals in high-risk areas according to the classification results of high-risk areas to obtain crack images, such as Figure 10 As shown;

[0089] In this embodiment, the crack image acquisition terminal includes:

[0090] The main processor is responsible for scheduling and executing the core business logic of the terminal, configuring image sensor parameters, executing crack image capture tasks, performing preliminary processing on the original image, and communicating with the microcontroller unit. Specifically, the main processor in this embodiment:

[0091] (1) System control, running a streamlined Linux system, responsible for scheduling and executing the core business logic of the terminal, including image acquisition, data processing, and communication coordination.

[0092] (2) Image processing: configure image sensor parameters, execute crack image shooting task, and perform preliminary processing on the original image.

[0093] (3) Collaborative management: communicate with the microcontroller unit, receive its instructions, and start or shut down the high-power modules as needed.

[0094] Communication module, used to upload the collected high-definition crack images and equipment status information to the remote server;

[0095] Image sensor, used to capture high-definition crack images in conjunction with a fill light;

[0096] Fill light, used to provide controllable light source for lighting compensation in dark or night environments;

[0097] FLASH memory, used to temporarily store high-definition crack images taken by the image sensor;

[0098] The microcontroller unit is used to control the power supply of each module (such as sequential activation / deactivation of the main processor and communication module), as well as to perform power-on self-test (ADC detection of power supply stability) and task termination protocol (deep sleep to 10μA).

[0099] Furthermore, the crack image acquisition terminal of the present invention has the following innovative design functions:

[0100] 1. Low-power control unit MCU design method

[0101] (1) Independent power management architecture

[0102] Hardware-level power consumption control is achieved by using an STM32F030 microcontroller unit (MCU) that is physically isolated from the main processor (RV1106) and communication module (4G). The MCU controls the working hours of all hardware components to ensure that different functional modules are powered on at the correct time when needed.

[0103] The MCU has:

[0104] a) Ultra-low sleep current: typical value ≤ 5μA (including real-time clock RTC operation power consumption);

[0105] b) Multiple wakeup source management: integrated RTC timer wakeup and GPIO external interrupt wakeup dual-channel mechanism;

[0106] (2) Dynamic power management strategy

[0107] Power-on self-test process:

[0108] a) The MCU starts up first and uses the analog-to-digital converter (ADC) and power management unit (PMU) to detect the power supply status. The ADC and PMU collect external power and battery voltage data and transmit it to the MCU.

[0109] b) After the voltage stability verification is passed, activate the PMU → RV1106 → 4G module power supply in sequence;

[0110] Mission Termination Protocol:

[0111] a) After the service cycle ends, execute the reverse power off sequence (4G → RV1106 → PMU);

[0112] b) Autonomously enters deep sleep mode (standby current ≤ 10μA), starts the RTC timing before sleep, and restarts the MCU after the set wake-up time is reached.

[0113] (3) Abnormal interrupt handling mechanism

[0114] Establish a communication status monitoring channel, when the following abnormal conditions are detected:

[0115] a) The duration of network interruption is ≥ 120 seconds;

[0116] b) Data traffic exceeding limit alarm;

[0117] c) The number of server connection failures is greater than 3 times;

[0118] Execute the emergency response procedures as follows:

[0119] a) Forcefully terminate the current process and turn off the power of high-power modules (RV1106 and 4G modules);

[0120] b) Automatically start intermittent sleep mode (working / sleeping cycle = 15 minutes), wake up every 15 minutes and check whether the abnormal situation has been resolved. If the abnormal situation disappears, the established work process will proceed normally.

[0121] 2. Main processor (RV1106) software optimization method

[0122] In conventional designs, the crack image acquisition terminal's main processor (RV1106), communication module (4G), image sensor, fill light, and FLASH memory consume a lot of power. 95% of the power in a single shooting task is consumed by these components.

[0123] The main processor (RV1106) runs a streamlined Linux system. It takes 30 seconds from booting the kernel to loading the service program, which includes communicating with the MCU, configuring the image sensor, activating the fill light, executing the capture function, and uploading images via 4G. The 4G module takes about 20 seconds to boot, configure AT commands, and complete network registration. This can take longer in areas with weak wireless signals, especially in areas with weak 4G base station signals.

[0124] Based on the above problems, the crack image acquisition terminal in the present invention optimizes its internal software. The software optimization method specifically includes startup process optimization and communication protocol optimization:

[0125] a) Image sensor parameter configuration is performed simultaneously with 4G network registration;

[0126] b) Fill light preheating is completed during the data transmission stage;

[0127] c) Using AT command pre-caching technology, the 4G module network registration time is shortened by 40% (from the original 20 seconds to 12 seconds).

[0128] S3. Obtain an estimated crack length at the current moment based on the acquired crack image, summarize the estimated length data, and obtain a crack length trend chart;

[0129] The estimated value of the crack length at the current moment in this embodiment is specifically obtained by using the steel structure fatigue crack identification method disclosed in the invention patent publication number CN117523470A.

[0130] S4. Correcting the crack length on the crack length trend chart, using the corrected crack length trend chart to obtain the average crack growth rate at each stage, and using a preset crack rate-risk classification deep neural network to perform risk classification on the average crack growth rate at each stage to obtain the risk level of the fatigue crack risk location of the steel crane beam;

[0131] In the embodiment of the present invention, the calculation formula for the average crack growth rate in each stage is:

[0132]

[0133] Among them, L d (t1) is the final displayed length of the crack at a certain stage in the crack length trend diagram, Ld (t0) is the initial display length of a certain stage in the crack length trend diagram, Δ s t is the total time interval of this stage.

[0134] In the embodiment of the present invention, the preset crack rate-risk classification deep neural network is constructed by the following method:

[0135] S41. Establish a fatigue crack rate-risk database;

[0136] Leveraging relevant project experience, we tracked the development of cracks in the upper flange area of ​​the steel crane beam, the support cover area, and the support plug area. Because the lower flange area is a purely tensile zone, once cracks appear, the crane beam will be quickly damaged. Therefore, the crack risk level in the lower flange area is directly classified as high risk, and the development rate is not calculated. From the time the cracks are discovered, we tracked the change in crack length over time. Once cracks are discovered, the risk level is classified as low risk. Using the crack length acquisition technology disclosed in the invention patent with publication number CN117523470A, we recorded the crack length daily and calculated the crack development rate at each stage. Based on the crack location, crack length, and crack development rate, we manually divided the corresponding risk levels for each rate, thereby obtaining a fatigue crack rate-risk database.

[0137] The fatigue crack rate-risk database established is characterized by the location, length, and development rate of each fatigue crack in the steel crane beam, and the label is the manually calibrated risk level;

[0138] S42. Construct a second deep neural network. The second deep neural network specifically uses an existing neural network architecture; this embodiment does not involve any improvement to the architecture. The second deep neural network is trained using data from the fatigue crack rate-risk database to obtain a preset crack rate-risk classification deep neural network. The second deep neural network in this embodiment includes a sequentially connected input layer, multiple hidden layers, corresponding multiple sigmoid activation layers, and an output layer.

[0139] In this embodiment, the crack length on the crack length trend chart is corrected based on statistics, and the crack length data that does not conform to the statistical law is determined as an outlier to avoid generating unreasonable warning information. The specific correction method is as follows:

[0140] (1) Basic Definition

[0141] L c (t)——identification length at time t;

[0142] Ave c (t-1)——The latest average value at time t, the length data L of the previous stage cDoes not participate in the calculation of the latest average value in this stage;

[0143] L d (t)——the length displayed on the crack length trend chart at time t;

[0144] L d (t1)——The crack length trend chart shows the final length at this stage;

[0145] L d (t0) - the crack length trend diagram initially shows the length at this stage;

[0146] Δ(L)——length difference;

[0147] Δ s t——total time interval of this stage;

[0148] S(t)∈{1,2,3...}——current stage;

[0149] T k ——time of the kth turning point;

[0150] I(t) is the set of times that satisfy the following conditions:

[0151] Condition 1: Belongs to the current stage S(τ) = S(t);

[0152] Condition 2: The time is no later than the current point τ<t;

[0153] Condition 3: The data is accepted as non-outliers;

[0154] v t ——crack growth rate at each stage;

[0155] (2) Calculation of average value for each stage: in:

[0156] I (t-1) ={τ|S(τ)=S(t),τ<t,L c (τ) is accepted as a non-outlier point}, when the data does not participate in the latest average calculation, because the current point may be an outlier;

[0157] (3) Calculation of length difference: Δ(L) = L c (t)-Ave c (t-1);

[0158] (4) Determined as the starting point condition: The starting point of the stage is the first data point in the monitoring stage, which is manually confirmed and verified on site and serves as the data benchmark.

[0159] (5) Condition for determining a normal point: |Δ(L)|≤10∩Lc (t)≥L d (t-1);

[0160] This formula contains two conditions that must be met simultaneously: ① Deviation condition: |Δ(L)|≤10, the absolute deviation between the current recognition length and the average value of the current stage does not exceed 10 units, which is used to filter out significant outliers in the algorithm recognition caused by industrial site noise interference; ② Monotonicity condition: L c (t)≥L d (t-1), the current identified length should be greater than or equal to the last displayed length to ensure that the crack length trend chart always shows a non-declining trend (the crack will not "shorten"); when both conditions are met, the data is displayed in red in the length trend chart and participates in the trend chart connection. When only the first condition is met, the data is displayed in gray in the trend chart and does not participate in the trend chart connection;

[0161] (6) Conditions for determining an unreasonable point: |Δ(L)|≤10∩L c (t)<L d (t-1), unreasonable points participate in subsequent average value calculations.

[0162] This formula contains two conditions that must be met simultaneously: ① Deviation condition: |Δ(L)|≤10, the absolute deviation between the current recognition length and the average value of the current stage does not exceed 10 units, which is used to filter out significant outliers in the algorithm recognition caused by industrial site noise interference; ② Monotonicity condition: L c (t)<L d (t-1), the current identified length should be less than the previously displayed length, and the crack length trend chart shows a downward trend (the crack is "shortening"); this data is displayed in gray in the trend chart and does not participate in the trend chart connection.

[0163] (7) Outlier determination condition: Δ(L)<-10∩L c (t)<L d (t-1), outliers do not participate in subsequent average calculations.

[0164] This formula contains two conditions that must be met simultaneously: ① Negative deviation condition: Δ(L) < -10, the current identified length is 10 units lower than the stage average, which means that the crack length suddenly decreases abnormally. The data may be caused by noise interference; ② Monotonicity condition: L c (t)<L d (t-1), the current identified length is less than the previous valid display value, which means that the crack is "shortened", which violates the laws of physics.

[0165] (8) Condition for determining a turning point: Δ(L)>10∩L c (t)>L d (t-1);

[0166] This formula contains two conditions that must be met simultaneously: ① Positive deviation condition: Δ(L)>10, the current identification length exceeds the average value of the current stage by 10 units, which means that the crack length has increased dramatically and the crack may develop rapidly; ② Monotonicity condition: L c (t)≥L d (t-1), the current identification length should be greater than or equal to the last displayed length to ensure that the crack length trend chart always shows a non-declining trend (the crack will not "shorten"); when both conditions are met, the data point is determined to be a turning point, which means that the previous stage is terminated, and the turning point data does not participate in the calculation of the average value of the previous stage and the development rate of the previous stage.

[0167] (9) Display length of crack length trend chart at time t:

[0168]

[0169] (10) Phase transfer function

[0170] S(t)=S(t-1)+1——when t meets the turning point condition

[0171] (11) Calculate the crack growth rate at each stage

[0172]

[0173] S5. When the risk level is high, generate warning information. The crack length curve and crack growth rate curve of high risk area are as follows: Figure 8 Specifically, when the average crack growth rate in each stage changes 5 times cumulatively or the average crack growth rate in the current stage enters a high-risk level, a warning message is generated.

[0174] The risk levels of fatigue crack risk areas of the steel crane beam in the embodiment of the present invention are specifically divided as shown in Table 3 below.

[0175] Table 3 Risk level classification of fatigue crack risk areas of steel crane beams

[0176]

[0177]

[0178] Example 2

[0179] like Figure 7 As shown, an embodiment of the present invention further provides a fatigue risk monitoring system for in-service steel crane beams, specifically comprising:

[0180] The data acquisition module is used to obtain crack images through the crack image acquisition terminal and send the obtained crack images to the data processing module; the data acquisition module and the crack image acquisition terminal in the embodiment of the present invention use the HTTP protocol to collect, transmit and store image data.

[0181] A data processing module is used to obtain an estimated crack length at the current moment based on the acquired crack image, and to summarize the length estimation data to obtain a crack length trend chart. In the embodiment of the present invention, the data processing module has an embedded fatigue crack recognition algorithm (e.g., the crack recognition method disclosed in the invention patent publication number CN117523470A). The crack image passes through the data processing module to obtain an estimated crack length at the current moment.

[0182] The data post-processing module corrects the crack length on the crack length trend chart, uses the corrected crack length trend chart to obtain the average crack growth rate in each stage, and uses the preset crack rate-risk classification deep neural network to classify the average crack growth rate in each stage, and obtain the risk level of the fatigue crack risk area of ​​the steel crane beam. When the risk level is high, an early warning message is generated.

[0183] The fatigue risk monitoring system for in-service steel crane beams provided by an embodiment of the present invention further includes:

[0184] Data display module, used to display monitoring point distribution information, fatigue crack images, crack length identification results and early warning information;

[0185] The equipment control module is used to automatically adjust the shooting interval of the crack image acquisition terminal in real time based on the current monitoring point reminder information and the number of early warning information, realizing dynamic risk tracking. It is also used to automatically adjust the fill light intensity in real time according to the current monitoring point light and steel structure paint film conditions to avoid overexposure that affects crack identification effect.

[0186] The system management module is used to manage users, roles, organizations and resource menus within the system.

[0187] In the embodiment of the present invention, the preset crack rate-risk classification deep neural network is constructed by the following method:

[0188] Step 1: Establish a fatigue crack rate-risk database;

[0189] Step 2: Construct a second deep neural network and use the data in the fatigue crack rate-risk database for training to obtain a preset crack rate-risk classification deep neural network.

[0190] The performance of the correction method proposed by the data post-processing module in Example 2 of the present invention is studied below.

[0191] The crack images taken in chronological order are identified by the algorithm of the data processing module, and the following identification lengths are obtained. The data is corrected using the correction method, and the crack development rate at each stage is calculated. The results are shown in Table 4 below.

[0192] Table 4 Results of crack development rate at each stage

[0193]

[0194]

[0195]

[0196] Crack growth trend chart Figure 9 As shown, through Figure 9 As can be seen from the results in Table 4, the correction method proposed in the embodiment of the present invention can effectively eliminate the algorithm recognition error caused by industrial environment noise and truly reflect the crack growth trend.

[0197] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A fatigue risk monitoring method for in-service steel crane beams, characterized in that: The following steps are involved: S1. Obtain data of the steel crane beam to be monitored, and classify the obtained crane beam data using a pre-trained fatigue crack location deep neural network to obtain a classification result of high-risk areas of the steel crane beam; S2. Deploy crack image acquisition terminals in high-risk areas based on the high-risk area classification results to acquire crack images; S3. Obtain an estimated crack length at the current moment based on the acquired crack image, summarize the estimated length data, and obtain a crack length trend chart; S4. Correcting the crack length on the crack length trend chart, using the corrected crack length trend chart to obtain the average crack growth rate at each stage, and using a preset crack rate-risk classification deep neural network to perform risk classification on the average crack growth rate at each stage to obtain the risk level of the fatigue crack risk location of the steel crane beam; S5. When the risk level is high, generate warning information.

2. The fatigue risk monitoring method for in-service steel crane beams according to claim 1, characterized in that: In step S1, the pre-trained fatigue crack location deep neural network is constructed by the following method: S11. Establish a fatigue cracking database for steel crane beams; S12. Construct a first deep neural network, train it with data from a steel crane beam fatigue cracking database, obtain a fatigue crack location deep neural network model, and then modify the model to obtain a pre-trained fatigue crack location deep neural network.

3. The fatigue risk monitoring method for in-service steel crane beams according to claim 2, characterized in that: In step S12, the model correction method is as follows: S121, Acquisition of prior knowledge in the field; S122. Modify the model based on prior knowledge in the field to obtain a pre-trained fatigue crack location deep neural network.

4. The fatigue risk monitoring method for in-service steel crane beams according to claim 1, characterized in that: In step S2, the crack image acquisition terminal includes: The main processor is responsible for scheduling and executing the terminal's core business logic, configuring image sensor parameters, performing crack image capture tasks, performing preliminary processing on raw images, and communicating with the microcontroller unit. Communication module, used to upload the collected high-definition crack images and equipment status information to the remote server; Image sensor, used to capture high-definition crack images in conjunction with a fill light; Fill light, used to provide controllable light source for lighting compensation in dark or night environments; FLASH memory, used to temporarily store high-definition crack images taken by the image sensor; The microcontroller unit is used to control the power on and off of each module and to perform power-on self-test and task termination protocols.

5. The fatigue risk monitoring method for in-service steel crane beams according to claim 1, characterized in that: In step S4, the calculation formula for the average crack growth rate in each stage is: Among them, L d (t1) is the final displayed length of the crack at a certain stage in the crack length trend diagram, L d (t0) is the initial display length of a certain stage in the crack length trend diagram, Δ s t is the total time interval of this stage.

6. The fatigue risk monitoring method for in-service steel crane beams according to claim 2, characterized in that: In step S4, the preset crack rate-risk classification deep neural network is constructed by the following method: S41. Establish a fatigue crack rate-risk database; S42. Construct a second deep neural network and use the data in the fatigue crack rate-risk database for training to obtain a preset crack rate-risk classification deep neural network.

7. The fatigue risk monitoring method for in-service steel crane beams according to claim 6, characterized in that: The first deep neural network and the second deep neural network have the same structure, both including an input layer, multiple hidden layers, corresponding multiple Sigmoid activation layers and an output layer connected in sequence.

8. The fatigue risk monitoring method for in-service steel crane beams according to claim 1, characterized in that: In step S4, the displayed length of the crack length trend chart after correction at time t is: Among them, L d (t) is the length displayed on the crack length trend chart at time t, L c (t) is the recognition length at time t.

9. A fatigue risk monitoring system for in-service steel crane beams, characterized in that: include: Data acquisition module; Used to obtain crack images through the crack image acquisition terminal and send the obtained crack images to the data processing module; A data processing module is used to obtain an estimated value of the crack length at the current moment based on the acquired crack image, and to summarize the length estimation data to obtain a crack length trend chart; The data post-processing module corrects the crack length on the crack length trend chart, uses the corrected crack length trend chart to obtain the average crack growth rate in each stage, and uses the preset crack rate-risk classification deep neural network to classify the average crack growth rate in each stage, and obtain the risk level of the fatigue crack risk area of ​​the steel crane beam. When the risk level is high, an early warning message is generated.

10. The fatigue risk monitoring system for in-service steel crane beams according to claim 9, characterized in that: Also includes: Data display module, used to display monitoring point distribution information, fatigue crack images, crack length identification results and early warning information; The equipment control module is used to adjust the shooting interval and fill light intensity of the crack image acquisition terminal in real time; The system management module is used to manage users, roles, organizations, and resource menus.

Images