An AUV seabed oil pipeline inspection risk identification system based on deep learning

Abstract

The application discloses a kind of AUV seabed oil pipeline inspection risk identification systems based on deep learning, it is related to marine engineering technical field, including inspection risk management center, the inspection risk management center communication connection has following module, in which: AUV autonomous inspection module is used to deploy AUV to implement seabed oil pipeline inspection task, and according to seabed environment and pipeline state dynamically adjusts the control instruction of AUV, optimizes inspection path and speed.The application can efficiently classify pipeline image data by combining deep learning algorithm to construct abnormal pipeline identification model, identify seabed oil pipeline area in abnormal state, and then further analyze the identified abnormal state area, calculate pipeline risk trend index, evaluate the potential risk level of pipeline, which helps to timely discover and handle potential safety hazards, prevent accidents, and ensure the safe operation of seabed oil pipeline.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of ocean engineering, and particularly relates to an AUV seabed oil pipeline inspection risk identification system based on deep learning. BACKGROUND

[0002] With the gradual decrease of land oil resources, marine oil has become an important energy supplement source, and the expansion of marine oil development scale has continuously increased the number and length of seabed oil pipelines. These pipelines not only undertake the transportation tasks of oil and natural gas, but also involve auxiliary production processes such as water injection and gas injection. The expansion of the scale of the pipelines increases the probability of pipeline failure, and therefore, more efficient and comprehensive inspection means are needed to ensure the safe operation of the pipelines.

[0003] Now, the traditional seabed oil pipeline detection often adopts an ROV (remotely operated vehicle). The ROV is linked with a surface mother ship through a cable, and the operation depends on personnel operation and is limited by the voyage distance, which has limitations and can cause the pipeline risks to be unable to be found in time and the possibility of pipeline failure to be increased. Therefore, how to use an AUV to carry a high-definition camera to perform seabed pipeline inspection, combine image classification and pattern recognition technology, classify the images of the seabed pipeline, identify the areas with large differences from the normal pipeline state, further mine the potential information in the images, and confirm the risk points is a problem to be solved by the application. For this purpose, the application provides an AUV seabed oil pipeline inspection risk identification system based on deep learning. SUMMARY

[0004] The application aims to provide an AUV seabed oil pipeline inspection risk identification system based on deep learning to solve the problems in the background.

[0005] To solve the above technical problems, the technical scheme adopted by the application is as follows:

[0006] An AUV seabed oil pipeline inspection risk identification system based on deep learning comprises an inspection risk management center, and the inspection risk management center is communicatively connected with the following modules, wherein:

[0007] An AUV autonomous inspection module is used to deploy the AUV to perform a seabed oil pipeline inspection task, dynamically adjust the control instructions of the AUV according to the seabed environment and the pipeline state, optimize the inspection path and speed, and automatically return to the water surface after the inspection task is completed, so as to realize the integrated operation of "arming-inspection-recovery".

[0008] An image acquisition and processing module is used to collect and pre-process the image data of the seabed oil pipeline through the inspection process of the AUV, perform feature analysis, extract pipeline defect features, and form a defect feature sequence table.

[0009] An image recognition classification module is configured to construct an abnormal pipeline recognition model in combination with a deep learning algorithm, classify the preprocessed pipeline image data, and identify the seabed oil pipeline region in an abnormal state.

[0010] A risk assessment module is configured to further analyze the identified abnormal state region, calculate a pipeline risk trend index, and evaluate the potential risk level of the pipeline.

[0011] The AUV autonomous inspection module comprises an AUV launching and diving unit and a pipeline identification and tracking unit.

[0012] The AUV launching and diving unit is configured to launch the AUV to a designated position by a water surface launching device and control it to dive to a seabed region autonomously, thereby realizing automatic launching and diving of the AUV, reducing manual intervention, and improving operation efficiency.

[0013] The pipeline identification and tracking unit is configured to identify the seabed oil pipeline by using a multi-angle high-definition camera carried by the AUV, perform inspection along the pipeline, and dynamically adjust the control instructions of the AUV according to the seabed environment and pipeline state by using a reinforcement learning algorithm, thereby optimizing the inspection path and speed.

[0014] The AUV launching and diving unit specifically comprises:

[0015] Before launching, an operator comprehensively checks the AUV to ensure that each system function is normal and the power is sufficient, and presets a diving path and a target region in combination with a task requirement and a high-precision positioning system, and then installs the AUV on the water surface launching device to stably launch the AUV to a designated water surface position.

[0016] After receiving the diving instructions, the autonomous control system of the AUV is started to gradually discharge the gas in the buoyancy tank by controlling the buoyancy adjusting system, increase the weight of the AUV, and make it start to slowly dive, at the same time, the propulsion system of the AUV adjusts the propulsion direction and speed according to the preset diving path and the information fed back by the attitude sensor, when the AUV approaches the seabed target region, the diving depth and position are further accurately controlled by the height sensor and the terrain matching system of the AUV, so that the AUV can accurately reach the predetermined seabed region.

[0017] After reaching the seabed, the positioning system of the AUV communicates with the seabed positioning base station to determine its position on the seabed by using acoustic positioning technology, and compares it with the preset target position, if there is a deviation, the AUV automatically adjusts the position until it completely reaches the target region, at this time, the AUV enters a standby state, the high-definition camera and sensor carried by the AUV start to warm up and initialize, and prepare to implement the pipeline inspection task.

[0018] The further improvement of the technical scheme of the present application is that the pipeline identification and tracking unit specifically comprises:

[0019] After the AUV dives to the seabed target area, the multi-angle high-definition camera carried by the AUV starts to work, captures seabed environment images through 360-degree omnidirectional scanning, and compares and analyzes the characteristics of the seabed oil pipeline through image recognition technology to identify the position and direction of the seabed oil pipeline. At this time, the positioning system of the AUV determines the relative position relationship between itself and the pipeline in combination with the initial position information of the pipeline, and then plans an initial inspection path according to the preset inspection task requirements and the preliminary distribution of the pipeline, and the inspection path covers the key parts of the pipeline.

[0020] During the inspection along the initial path, the AUV continuously perceives seabed environment information, including water flow speed, direction, seabed topography and real-time state of the pipeline, and then analyzes by using a reinforcement learning algorithm, evaluates the influence of different control instructions on the inspection effect according to the current environment and pipeline state, dynamically adjusts the control instructions of the AUV, and optimizes the inspection path and speed.

[0021] The process of dynamically adjusting the control instructions of the AUV is:

[0022] During the inspection of the AUV along the initial inspection path, the AUV continuously collects seabed environment and pipeline state information through various sensors carried by the AUV, including water flow speed and direction obtained through a flow rate sensor, seabed topography scanned through a sonar system, and real-time state of the pipeline photographed by a high-definition camera and identified through image recognition, and the information is transmitted to the central processing unit of the AUV.

[0023] After receiving the seabed environment and pipeline state information collected from the sensors, a reinforcement learning algorithm is run to evaluate the potential influence of different control instructions on the inspection effect according to the current water flow speed, direction and seabed topography, simulate a plurality of control strategies, analyze the safety index, efficiency index and stability index under each control strategy, and then calculate the expected return under each strategy, learn the control instructions that can achieve the optimal inspection effect in a complex environment while avoiding collision and deviation from the pipeline, and adapt to the changing seabed environment;

[0024] According to the analysis result of the reinforcement learning algorithm, the control instructions of the AUV are dynamically adjusted, including adjusting the speed and direction of the thruster to optimize the inspection path and speed, so that efficient inspection is realized in a complex seabed environment while ensuring the safety and stability of the inspection process.

[0025] The further improvement of the technical scheme of the present application is that the image acquisition and processing module comprises an image acquisition unit and an image feature extraction unit.

[0026] The image acquisition unit is configured to acquire image data of the seabed oil pipeline by a high-definition camera carried by the AUV, and to pre-process the image data, including denoising, contrast enhancement and distortion correction.

[0027] The image feature extraction unit is configured to analyze the pre-processed image data, and to extract defect features of the seabed oil pipeline related to the health state of the seabed oil pipeline.

[0028] The image acquisition unit specifically includes:

[0029] After the AUV reaches the seabed oil pipeline inspection area, the carried high-definition camera is self-checked to ensure that all parameters are normal, the lens is not blocked and has no faults, and the brightness of a strong light source provided with the camera is automatically adjusted according to the light condition of the seabed environment. After the strong light source is turned on, sufficient illumination is provided for the camera to cope with the weak light environment of the seabed, and then the high-definition camera is used to acquire images of the seabed oil pipeline in all directions and at multiple angles according to a preset shooting frequency and angle, and at the same time, the shooting position and time information are recorded in real time.

[0030] The acquired image data is pre-processed, including denoising, contrast enhancement and distortion correction, and then the pre-processed image data is stored in the storage system of the AUV.

[0031] The image feature extraction unit specifically includes:

[0032] The pre-processed image data is extracted, and an edge detection algorithm is used to segment the image data to separate the relevant area of the seabed oil pipeline, and the boundary of the seabed oil pipeline is further located by combining the built-in basic data of the seabed oil pipeline in the system.

[0033] Based on the relevant area of the seabed oil pipeline segmented by the image, the geometric features and shape features of the seabed oil pipeline are analyzed, and defect features related to the health state of the seabed oil pipeline are extracted, including pipe diameter, pipe bending degree, crack length, crack width, hole area, hole diameter, corrosion depth and corrosion area ratio.

[0034] The extracted defect features are integrated, and the reference values of the defect features are determined according to the inspection requirements of the seabed oil pipeline, i.e. the maximum allowable values of the defect features, to form a defect feature list. For each defect area, the extracted geometric features and shape features are summarized, and the position information of the defect area on the pipeline is labeled.

[0035] The further improvement of the technical scheme of the present application is that the image recognition classification module specifically comprises:

[0036] A large number of annotated seabed oil pipeline image data are collected, including images of normal state and various abnormal states such as cracks, corrosion and deformation, and pipeline defect features related to the health state of the seabed oil pipeline are extracted from the images to form a comprehensive data set, and the comprehensive data set is divided into a training set and a validation set;

[0037] A convolutional neural network (CNN) is selected as the basic architecture of the deep learning algorithm to construct an abnormal pipeline recognition model, the convolutional neural network model is adjusted according to the characteristics of the seabed oil pipeline image, the images of the training set are input into the model, and the back propagation algorithm and the optimizer are used to continuously adjust the model parameters, so that the error between the prediction result and the true label is reduced, in the training process, the performance of the model is monitored in real time by using the validation set, and after multiple iterations, the performance of the model on the validation set is stable and meets the requirements, and finally the constructed abnormal pipeline recognition model is obtained;

[0038] The trained abnormal pipeline recognition model is deployed, the pipeline image data to be detected is input into the abnormal pipeline recognition model, the abnormal pipeline defect features are identified in combination with the reference values of the pipeline defect features, and then the state category to which the image belongs, i.e., the normal state or the abnormal state, is judged according to the abnormal pipeline defect features and the probability prediction of the model on the image belonging to different state categories, for the image judged as the abnormal state, the specific position and range of the abnormal area on the pipeline are further determined, and finally the abnormal state seabed oil pipeline area is identified.

[0039] The further improvement of the technical scheme of the present application is that the risk assessment module specifically comprises:

[0040] The abnormal state area information output by the image recognition classification module is received, the pipeline defect features of the abnormal state area are quantitatively analyzed, and each pipeline defect feature is compared and analyzed with the pre-set reference value of each pipeline defect feature, if the actual value of a pipeline defect feature exceeds or reaches the reference value, the pipeline defect feature is determined as a risk factor;

[0041] After determining the risk factors, the deviation degree of each risk factor from the corresponding reference value is analyzed, different weights are given to each risk factor in combination with the design specification of the pipeline, the operation history data and the safety requirements, so as to reflect the contribution degree of the risk factor to the overall risk of the pipeline, and then the quantitative value of each risk factor is combined with the corresponding weight to calculate a pipeline risk trend index, reflecting the change trend of the potential risk of the pipeline;

[0042] According to the calculated pipeline risk trend index, according to the pre-set risk level division standard, the potential risk of the pipeline is divided into different risk levels, which are low risk level, medium risk level and high risk level, wherein the risk level division standard is pre-set according to the design specification, operation history and safety requirement of the pipeline, and then the risk threshold of each risk level is determined, after the evaluation is completed, the risk assessment report is generated, including the detailed information of the abnormal area, risk factor analysis, calculation result of the pipeline risk trend index and risk level determination, at the same time, the corresponding suggestion measures are put forward for different risk levels, and the risk assessment report is transmitted to the relevant personnel after the AUV is recovered.

[0043] Further improvement of the technical scheme of the present application is that the calculation process of the pipeline risk trend index is:

[0044] The number of pipeline defect features participating in risk assessment, i.e. the total number of risk factors, is determined, for each risk factor, the quantitative value, the corresponding reference value and the pre-set weight are obtained;

[0045] For each risk factor, the quantitative value is subtracted from the reference value, and the result is divided by the reference value, and the absolute value is taken, to calculate the relative deviation value of each risk factor, to analyze the relative deviation degree of the actual value of the risk factor and the reference value;

[0046] For each risk factor, the ratio of the quantitative value to the reference value is calculated, and the result is added to 1 and the natural logarithm is taken, to calculate the logarithmic term of each risk factor;

[0047] The weight, relative deviation value and logarithmic term of each risk factor are multiplied to obtain the contribution value of each risk factor, and then the contribution values of all risk factors are added to obtain the total contribution value, and the square root of the total contribution value is taken, i.e. the pipeline risk trend index is obtained, the greater the value of the pipeline risk trend index, the higher the potential risk of the pipeline.

[0048] Due to the adoption of the above technical scheme, the present application has the following technical progress compared with the prior art:

[0049] 1、The present application provides an AUV seabed oil pipeline inspection risk identification system based on deep learning, which can efficiently classify pipeline image data by constructing an abnormal pipeline identification model combining deep learning algorithm, identify the seabed oil pipeline area in abnormal state, and further analyze the identified abnormal state area, calculate the pipeline risk trend index, and evaluate the potential risk level of the pipeline, which helps to timely discover and handle potential safety hazards, prevent accidents and ensure the safe operation of the seabed oil pipeline.

[0050] 2. This invention provides a deep learning-based AUV subsea oil pipeline inspection risk identification system. Through the AUV autonomous inspection module, it realizes automated and intelligent inspection of subsea oil pipelines. Compared with the traditional ROV inspection method, it greatly reduces manual intervention and improves inspection efficiency. At the same time, the multi-angle high-definition camera on the AUV can collect pipeline images from all directions and angles, ensuring the comprehensiveness and completeness of the inspection, and effectively improving the inspection quality. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0052] Figure 1 This is a schematic diagram of the workflow of the present invention;

[0053] Figure 2 This is a schematic diagram illustrating the workflow of the image recognition and classification module of the present invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Example 1, as Figure 1 , Figure 2 As shown, this invention provides a deep learning-based AUV subsea oil pipeline inspection risk identification system, including an inspection risk management center. The inspection risk management center is connected to the following modules for communication, wherein:

[0056] The inspection risk management center, as the system's control center, is responsible for coordinating the operation of various modules, receiving and processing data from each module, and issuing risk warnings.

[0057] The AUV autonomous inspection module is used to deploy AUVs to carry out inspection tasks on subsea oil pipelines. It dynamically adjusts the control commands of the AUVs according to the seabed environment and pipeline status, optimizes the inspection path and speed, and automatically returns to the surface after the inspection task is completed, realizing the integrated operation of "deployment-inspection-recovery". The AUV autonomous inspection module includes an AUV deployment and submersion unit and a pipeline identification and tracking unit.

[0058] AUV is installed on the water surface deployment device, the angle and position of the device are adjusted to make it in the best deployment posture, wherein the water surface deployment device is equipped with a positioning system and a communication module, the operator inputs the deployment instruction through the console, the water surface deployment device starts to work, and the AUV is stably put into the specified water surface position, at this time, the buoyancy adjusting system of the AUV starts to work, so that it is kept in the water surface standby state, and is ready to start the diving task, after receiving the diving instruction, the autonomous control system of the AUV is started, the buoyancy adjusting system is controlled to gradually discharge the gas in the buoyancy tank, the weight of the AUV is increased, so that it starts to slowly dive, at the same time, the propulsion system of the AUV adjusts the propulsion direction and speed according to the preset diving path and the information fed back by the attitude sensor, so as to ensure the stability and accuracy of the diving process, in the diving process, the sonar system of the AUV is used to monitor the seabed topography and the surrounding environment in real time, when the AUV approaches the seabed target area, the diving depth and position are further accurately controlled through the height sensor and the terrain matching system of the AUV, so that it can accurately reach the predetermined seabed area, after reaching the seabed, the positioning system of the AUV communicates with the seabed positioning base station, the position of the AUV on the seabed is determined through the acoustic positioning technology, and is compared with the preset target position, if there is deviation, the AUV automatically adjusts the position until it completely reaches the target area, at this time, the AUV enters the standby state, the high-definition camera and sensor carried by the AUV start to warm up and initialize, and are ready to implement the pipeline inspection task, at the same time, the communication system of the AUV transmits the state information and position data of the AUV back to the water surface mother ship in real time, so that the operator can master the working state of the AUV at any time;

[0059] The pipeline identification and tracking unit is used to identify the submarine oil pipeline by using the multi-angle high-definition camera carried by the AUV, patrol along the pipeline, and dynamically adjust the control instructions of the AUV according to the submarine environment and pipeline state by using the reinforcement learning algorithm, so as to optimize the patrol path and speed, improve the adaptability and patrol efficiency of the AUV in the complex submarine environment, and ensure the stability and safety of the patrol process. After the AUV dives to the target area of the seabed, the multi-angle high-definition camera carried by the AUV starts to work, captures the seabed environment image through 360-degree omnidirectional scanning, and compares and analyzes the characteristics of the submarine oil pipeline by using image recognition technology to identify the position and direction of the submarine oil pipeline. At this time, the positioning system of the AUV determines the relative position relationship between itself and the pipeline in combination with the initial position information of the pipeline, and then plans an initial patrol path according to the preset patrol task requirements and the preliminary distribution of the pipeline. The initial patrol path covers the key parts of the pipeline. During the patrol process along the initial path, the AUV continuously senses the seabed environment information, including the water flow speed, direction, seabed topography and real-time state of the pipeline, and then analyzes by using the reinforcement learning algorithm. According to the current environment and pipeline state, the influence of different control instructions on the patrol effect is evaluated, and the control instructions of the AUV are dynamically adjusted to optimize the patrol path and speed.

[0060] In addition, the process of dynamically adjusting the control instructions of the AUV is as follows:

[0061] During the patrol process of the AUV along the initial patrol path, the AUV continuously collects seabed environment and pipeline state information through various sensors carried by the AUV, including water flow speed and direction, seabed topography, and real-time state of the pipeline. The information is transmitted to the central processing unit of the AUV. After receiving the seabed environment and pipeline state information collected from the sensors, the reinforcement learning algorithm is run to evaluate the potential influence of different control instructions on the patrol effect according to the current water flow speed, direction and seabed topography. By simulating various control strategies, the safety index, efficiency index and stability index under each control strategy are analyzed, and the expected return under each strategy is calculated. By using historical data and a preset reward mechanism, the control instructions that can achieve the optimal patrol effect in complex environments are learned while avoiding collision and deviation from the pipeline to adapt to the changing submarine environment. According to the analysis results of the reinforcement learning algorithm, the control instructions of the AUV are dynamically adjusted, including adjusting the speed and direction of the thruster to optimize the patrol path and speed. Through dynamic adjustment, efficient patrol is realized in complex submarine environments while ensuring the safety and stability of the patrol process.

[0062] The calculation expression of the safety expected return under each strategy is as follows:

[0063] R(s, a) = a Safety(s, a) + b Efficiency(s, a) + g Stability(s, a) ;

[0064]

[0065] where R(s, a) is the expected return of performing action a in state s, a, b, g are weight coefficients for balancing the influences of safety, efficiency and stability, taking values in the range [0, 1] and a + b + g = 1, Safety(s, a) is the safety index, Efficiency(s, a) is the efficiency index, Stability(s, a) is the stability index, d(s, a) is the distance to obstacles when performing action a in state s, the smaller the distance, the lower the safety, s d is the standard deviation of distance, for adjusting the sensitivity of the safety index, v(s, a) is the inspection speed when performing action a in state s, the higher the speed, the higher the efficiency, v max is the maximum inspection speed, and Av(s, a) is the speed change rate when performing action a in state s, the smaller the speed change rate, the higher the stability, s v is the standard deviation of speed change rate, for adjusting the sensitivity of the stability index, Safety(s, a) tends to 1 when d(s, a) increases, indicating high safety, Safety(s, a) tends to 0 when d(s, a) decreases, indicating low safety, Efficiency(s, a) tends to 1 when v(s, a) increases, indicating high efficiency, Efficiency(s, a) tends to 0 when v(s, a) decreases, indicating low efficiency, Stability(s, a) tends to 1 when Av(s, a) decreases, indicating high stability, Stability(s, a) tends to 0 when Av(s, a) increases, indicating low stability, R(s, a) is close to 1 when Safety(s, a), Efficiency(s, a) and Stability(s, a) are all high, indicating that the comprehensive effect of performing action a in state s is good, and R(s, a) will decrease when any one of the indexes is low, indicating that the comprehensive effect is poor;

[0066] an image acquisition and processing module, configured to acquire and pre-process image data of the seabed oil pipeline through the inspection process of the AUV, and perform feature analysis to extract pipeline defect features and form a defect feature sequence table, the image acquisition and processing module comprising an image acquisition unit and an image feature extraction unit;

[0067] The image acquisition unit is used to acquire image data of the seabed oil pipeline through the high-definition camera carried by the AUV, and to pre-process the image data, including denoising, contrast enhancement and distortion correction operations, to improve the clarity and recognizability of the image and improve the image quality. The camera is equipped with a strong light source to cope with the weak light environment on the seabed. After the AUV reaches the seabed oil pipeline inspection area, the high-definition camera carried by the AUV is self-checked to ensure that all parameters are normal, the lens is not blocked and there is no fault. According to the light conditions of the seabed environment, the brightness of the strong light source equipped with the camera is automatically adjusted. After the strong light source is turned on, it provides sufficient illumination for the camera to cope with the weak light environment on the seabed. Then, the high-definition camera acquires images of the seabed oil pipeline in all directions and at multiple angles according to the preset shooting frequency and angle. At the same time, the shooting position and time information are recorded in real time. The acquired image data is pre-processed, including denoising, contrast enhancement and distortion correction operations. The image is denoised to effectively remove noise interference and make the image cleaner and clearer. The contrast of the denoised image is enhanced through histogram equalization and adaptive contrast enhancement techniques to improve the contrast between different regions of the image and make the features such as the outline and details of the pipeline more obvious. After the contrast enhancement is completed, the image is corrected for distortion. The lens distortion parameters are used to transform the image to eliminate barrel distortion and pillow distortion and restore the image to its normal geometric shape. Then, the pre-processed image data is stored in the storage system of the AUV.

[0068] The image feature extraction unit is used to analyze the features of the pre-processed image data and extract pipeline defect features related to the health status of the seabed oil pipeline. The defect feature sequence list is obtained by integrating the extracted image data and using the edge detection algorithm (Canny edge detection) to segment the image and separate the relevant areas of the seabed oil pipeline. The system has built-in basic data of the seabed oil pipeline (shape, size and position information of the pipeline) to further locate the boundaries of the seabed oil pipeline. Based on the relevant areas of the seabed oil pipeline segmented by the image, the geometric and shape features of the seabed oil pipeline are analyzed to extract defect features related to the health status of the seabed oil pipeline, including pipeline diameter, pipeline curvature, crack length, crack width, hole area, hole diameter, corrosion depth and corrosion area ratio. The extracted defect features are integrated, and the reference values of each defect feature are determined according to the inspection requirements of the seabed oil pipeline, i.e. the maximum allowable values of each defect feature, forming a defect feature sequence list. For each defect area, the extracted geometric and shape features are summarized and labeled with their position information on the pipeline.

[0069] An image recognition classification module is configured to construct an abnormal pipeline recognition model in combination with a deep learning algorithm, classify the preprocessed pipeline image data, and identify the abnormal state of the subsea oil pipeline region.

[0070] A risk assessment module is configured to further analyze the identified abnormal state region, calculate a pipeline risk trend index, and evaluate the potential risk level of the pipeline.

[0071] In the embodiment 2 as shown in the embodiment 1, the present application provides a technical solution: preferably, the image recognition classification module specifically comprises: Figure 1 、 Figure 2 The image recognition classification module specifically comprises:

[0072] A large number of subsea oil pipeline image data in normal state and various abnormal states such as cracks, corrosion, and deformation are collected, pipeline defect features related to the health state of the subsea oil pipeline are extracted from the image data, a comprehensive data set is formed, the comprehensive data set is divided into a training set and a validation set, the training set is used for model learning, the validation set is used for evaluating the performance of the model, the data distribution is ensured to be reasonable, a convolutional neural network (CNN) is selected as the basic architecture of the deep learning algorithm to construct the abnormal pipeline recognition model, the convolutional neural network model is adjusted according to the characteristics of the subsea oil pipeline image, the images of the training set are input into the model, the images are sequentially subjected to a convolution layer and a pooling layer, the convolution layer automatically extracts pipeline defect features in the images through a convolution kernel, the pooling layer reduces the dimension of the feature map to reduce the amount of calculation, the extracted features are classified and predicted through a full connection layer, and the probability that the image belongs to a normal state or different abnormal states is output. The model parameters are continuously adjusted using the back propagation algorithm and the optimizer to reduce the error between the prediction result and the true label. During the training process, the performance of the model is monitored in real time using the validation set, the generalization ability is evaluated according to the accuracy and recall rate, if overfitting occurs, the model is optimized by adding L1 and L2 regularization terms or using the Dropout method to randomly discard part of the neurons, the performance of the model on the training set and the validation set is ensured to be good, and after multiple iterations, the performance of the model on the validation set is stable and meets the requirements. Finally, the constructed abnormal pipeline recognition model is obtained. The pipeline image data to be detected is input into the trained abnormal pipeline recognition model, the abnormal pipeline defect features are identified in combination with the reference values of the pipeline defect features, the state category to which the image belongs is determined according to the abnormal pipeline defect features and the probability prediction of the model on the image belonging to different state categories, i.e., the normal state or the abnormal state, the specific position and range of the abnormal region on the pipeline are further determined for the image determined as the abnormal state, and finally the abnormal state of the subsea oil pipeline region is identified.

[0073] The risk assessment module specifically comprises:

[0074] The abnormal state region information output by the image recognition classification module is received, and the pipeline defect features of the abnormal state region are quantitatively analyzed. The pipeline defect features are compared with the preset pipeline defect feature reference values. If the actual value of a pipeline defect feature exceeds or reaches the reference value, the pipeline defect feature is determined as a risk factor. After determining the risk factors, the deviation degree of each risk factor from the corresponding reference value is analyzed. In combination with the design specification of the pipeline, the operation history data and the safety requirement, different weights are assigned to each risk factor to reflect the contribution degree of the risk factor to the overall risk of the pipeline. Then, the quantitative value of each risk factor is combined with the corresponding weight to calculate a pipeline risk trend index, which reflects the change trend of the potential risk of the pipeline. According to the calculated pipeline risk trend index, the potential risk of the pipeline is divided into different risk levels according to the preset risk level division standard, which includes a low risk level, a medium risk level and a high risk level. The risk level division standard is preset according to the design specification of the pipeline, the operation history and the safety requirement, and the risk threshold of each risk level is determined. After the evaluation is completed, a risk assessment report is generated, including the detailed information of the abnormal region, the risk factor analysis, the calculation result of the pipeline risk trend index and the determination of the risk level. At the same time, corresponding measures are proposed for different risk levels. For the low risk region, regular monitoring is recommended. For the high risk and extremely high risk regions, immediate repair or replacement measures are recommended. After the AUV is recovered, the risk assessment report is transmitted to the relevant personnel.

[0075] In addition, the calculation process of the pipeline risk trend index is as follows:

[0076] The number of pipeline defect features participating in the risk assessment, i.e. the total number of risk factors, is determined. For each risk factor, its quantitative value, corresponding reference value and preset weight are obtained. For each risk factor, its quantitative value is subtracted from the reference value, and the result is divided by the reference value and taken as an absolute value to calculate the relative deviation value of each risk factor. The relative deviation degree of the actual value of the risk factor from the reference value is analyzed. For each risk factor, the ratio of its quantitative value to the reference value is calculated, and the result is added by 1 and taken as a natural logarithm to calculate the logarithmic term of each risk factor. The relative size relationship between the actual value of the risk factor and the reference value is considered, and the value of the logarithmic function also increases with the increase of the ratio. The weight, relative deviation value and logarithmic term of each risk factor are multiplied to obtain the contribution value of each risk factor. Then, the contribution values of all risk factors are added to obtain the total contribution value. The square root of the total contribution value is taken to obtain the pipeline risk trend index. The larger the value of the pipeline risk trend index, the higher the potential risk of the pipeline.

[0077] The calculation formula of the pipeline risk trend index is as follows:

[0078]

[0079] In the formula, RTI is a pipeline risk trend index, used to intuitively reflect the change trend of the potential risk of the pipeline, n is the total number of risk factors, i.e., the number of pipeline defect characteristics participating in risk assessment, i is the serial number of the risk factor, used to traverse each risk factor, w i is the weight of the i th risk factor, reflecting the contribution of the risk factor to the overall risk of the pipeline, F i is the quantitative value of the i th risk factor, i.e., the specific value of the pipeline defect characteristic obtained through analysis, B i is the reference value corresponding to the i th risk factor, when the quantitative values of all risk factors are equal to their reference values, RTI = 0, indicating that the pipeline is in a risk-free state, and as the deviation of the quantitative values of the risk factors from the reference values increases, the value of RTI also increases, and the greater the deviation of the quantitative value of a risk factor from the reference value, the greater the value of RTI, and at the same time also increases with the increase of , thereby causing the value of this term to increase, ultimately causing the value of RTI to increase, indicating that the potential risk of the pipeline is higher;

[0080] A plurality of risk levels correspond to a plurality of risk thresholds one by one, and the corresponding relationship is as follows:

[0081] The risk threshold of the low risk level is: 0 < RTI ≤ R L ;

[0082] The risk threshold of the medium risk level is: R L < RTI ≤ R M ;

[0083] The risk threshold of the high risk level is: RTI > R M ;

[0084] wherein RTI is a pipeline risk trend index, R L is the upper threshold of the low risk level and the lower threshold of the medium risk level, R M is the upper threshold of the medium risk level and the lower threshold of the high risk level.

[0085] The above merely describes specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be encompassed within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A deep learning-based AUV seabed oil pipeline inspection risk identification system comprising an inspection risk management center, characterized in that, The inspection risk management center is communicatively connected with the following modules, wherein: An AUV autonomous inspection module is configured to deploy the AUV to perform an inspection task of the submarine oil pipeline and dynamically adjust a control instruction of the AUV according to a submarine environment and a pipeline state to optimize an inspection path and a speed; An image acquisition and processing module is configured to acquire and pre-process image data of the submarine oil pipeline through an inspection process of the AUV, perform feature analysis, extract pipeline defect features, and form a defect feature sequence list, the image acquisition and processing module comprising an image acquisition unit and an image feature extraction unit; The image acquisition unit is configured to acquire the image data of the submarine oil pipeline through a high-definition camera carried by the AUV and pre-process the image data; The image feature extraction unit is configured to perform feature analysis on the pre-processed image data, extract pipeline defect features related to a health state of the submarine oil pipeline, and integrate the defect feature sequence list, specifically comprising: extracting the pre-processed image data, performing image segmentation on the pre-processed image data by using an edge detection algorithm, separating out a related area of the submarine oil pipeline, and further positioning a boundary of the submarine oil pipeline in combination with basic data of the submarine oil pipeline built in the system; based on the related area of the submarine oil pipeline segmented out by the image, analyzing geometric features and shape features of the submarine oil pipeline, and extracting defect features related to the health state of the submarine oil pipeline, including a pipeline diameter, a pipeline bending degree, a crack length, a crack width, a hole area, a hole diameter, a corrosion depth, and a corrosion area proportion; integrating the extracted defect features, determining a benchmark value of each defect feature according to an inspection requirement of the submarine oil pipeline, i.e., a maximum allowable value of each defect feature, and forming the defect feature sequence list, and for each defect area, summarizing the extracted geometric features and shape features and labeling position information of the defect area on the pipeline; an image recognition and classification module is configured to construct an abnormal pipeline recognition model in combination with a deep learning algorithm, classify pre-processed pipeline image data, and identify an abnormal state area of the submarine oil pipeline; a risk assessment module is configured to further analyze the identified abnormal state area, calculate a pipeline risk trend index, and evaluate a potential risk level of the pipeline.

2. The AUV seabed oil pipeline inspection risk identification system based on deep learning according to claim 1, characterized in that: The AUV autonomous inspection module comprises an AUV deployment and diving unit and a pipeline recognition and tracking unit; The AUV deployment and diving unit is configured to deploy the AUV to a specified position through a water surface deployment device and control the AUV to autonomously dive to a submarine area; The pipeline recognition and tracking unit is configured to identify the submarine oil pipeline by using a multi-angle high-definition camera carried by the AUV, perform inspection along the pipeline, and dynamically adjust a control instruction of the AUV according to a submarine environment and a pipeline state by using a reinforcement learning algorithm to optimize an inspection path and a speed.

3. The AUV seabed oil pipeline inspection risk identification system based on deep learning according to claim 2, characterized in that: The AUV deployment and diving unit specifically comprises: Before deployment, an operator checks the AUV, presets a diving path and a target area in combination with a task requirement and a high-precision positioning system, and then installs the AUV on the water surface deployment device to stably deploy the AUV to a specified water surface position; After receiving the diving instruction, the AUV's autonomous control system starts, controlling the buoyancy adjustment system to gradually discharge the gas in the buoyancy tank, increasing its own weight, so that it starts to slowly dive, at the same time, the AUV's propulsion system adjusts the propulsion direction and speed according to the preset diving path and the information feedback by the attitude sensor, when the AUV approaches the seabed target area, through its height sensor and terrain matching system, further accurate control of the diving depth and position is realized; After reaching the seabed, the AUV's positioning system communicates with the seabed positioning base station, determines its position on the seabed through acoustic positioning technology, and compares it with the preset target position, if there is deviation, the AUV automatically adjusts the position until it completely reaches the target area, at this time, the AUV enters the standby state, its carried high-definition camera and sensor start preheating and initialization, preparing to implement the pipeline inspection task.

4. The AUV seabed oil pipeline inspection risk identification system based on deep learning according to claim 2, characterized in that: The pipeline identification and tracking unit specifically includes: After the AUV dives to the seabed target area, its carried multi-angle high-definition camera starts working, through 360-degree omnidirectional scanning, it captures seabed environment images, and through image recognition technology, it compares and analyzes the characteristics of the seabed oil pipeline, identifies the position and direction of the seabed oil pipeline, the AUV's positioning system determines its relative position with the pipeline, and then according to the preset inspection task requirements and the preliminary distribution of the pipeline, an initial inspection path is planned; During the inspection along the initial path, the AUV continuously perceives the seabed environment information, including water flow speed, direction, seabed terrain undulation and real-time state of the pipeline, and then uses reinforcement learning algorithm for analysis, according to the current environment and pipeline state, it evaluates the influence of different control instructions on the inspection effect, dynamically adjusts the control instructions of the AUV, optimizes the inspection path and speed.

5. The AUV seabed oil pipeline inspection risk identification system based on deep learning according to claim 1, characterized in that: The image acquisition unit specifically includes: After the AUV reaches the seabed oil pipeline inspection area, it performs self-checking on the carried high-definition camera, and automatically adjusts the brightness of the strong light source equipped with the camera according to the seabed environment light condition, and then through the high-definition camera, it acquires images of the seabed oil pipeline according to the preset shooting frequency and angle, at the same time, it records the shooting position and time information in real time; The collected image data is preprocessed, including denoising, contrast enhancement and distortion correction, and then the preprocessed image data is stored in the AUV's storage system.

6. The AUV seabed oil pipeline inspection risk identification system based on deep learning according to claim 5, characterized in that: The image recognition and classification module specifically includes: Collecting labeled seabed oil pipeline image data, including normal state and abnormal state images, extracting pipeline defect features related to the health status of seabed oil pipeline from them to form a comprehensive data set, and dividing the comprehensive data set into training set and validation set; The convolutional neural network is selected as a deep learning algorithm basic framework to construct an abnormal pipeline identification model, the convolutional neural network model is adjusted according to the characteristics of the seabed oil pipeline image, the images of the training set are input into the model, the back propagation algorithm and the optimizer are used to continuously adjust the model parameters, in the training process, the performance of the model is monitored in real time by using the verification set, after multiple iterations, until the performance of the model on the verification set is stable and meets the requirements, finally, the abnormal pipeline identification model is constructed; The trained abnormal pipeline identification model is deployed, the pipeline image data to be detected is input into the abnormal pipeline identification model, the abnormal pipeline defect features are identified by combining the reference values of the pipeline defect features, and then the state category to which the image belongs, i.e., the normal state or the abnormal state, is judged according to the abnormal pipeline defect features and the probability prediction of the model on the image belonging to different state categories, for the image judged as the abnormal state, the specific position and range of the abnormal area on the pipeline are further determined, and finally the abnormal state seabed oil pipeline area is identified.

7. The AUV seabed oil pipeline inspection risk identification system based on deep learning according to claim 6, characterized in that: The risk assessment module specifically comprises: The abnormal state area information output by the image recognition and classification module is received, the pipeline defect features of the abnormal state area are quantitatively analyzed, the pipeline defect features are compared and analyzed with the pre-set reference values of the pipeline defect features, if the actual value of a pipeline defect feature exceeds or reaches the reference value, the pipeline defect feature is determined as a risk factor; After the risk factor is determined, the deviation degree of each risk factor from the corresponding reference value is analyzed, different weights are given to each risk factor according to the design specification of the pipeline, the operation history data and the safety requirements, and then the quantitative values of the risk factors are combined with the corresponding weights to calculate the pipeline risk trend index, reflecting the change trend of the potential risk of the pipeline; According to the calculated pipeline risk trend index, the potential risk of the pipeline is divided into different risk levels according to the pre-set risk level division standard, which are low risk level, medium risk level and high risk level, wherein the risk level division standard is pre-set according to the design specification of the pipeline, the operation history and the safety requirements, and then the risk threshold values of each risk level are determined, after the evaluation is completed, a risk assessment report is generated, including the detailed information of the abnormal area, the risk factor analysis, the calculation result of the pipeline risk trend index and the determination of the risk level, at the same time, corresponding measures are proposed for different risk levels, and the risk assessment report is transmitted to the relevant personnel after the AUV is recovered.

8. The AUV seabed oil pipeline inspection risk identification system based on deep learning according to claim 7, characterized in that: The calculation process of the pipeline risk trend index is as follows: The number of pipeline defect features participating in the risk assessment is determined, i.e., the total number of risk factors, for each risk factor, the quantitative value, the corresponding reference value and the pre-set weight are obtained; For each risk factor, the quantitative value is subtracted from the reference value, and then the result is divided by the reference value, and the absolute value is taken, to calculate the relative deviation value of each risk factor, and analyze the relative deviation degree of the actual value of the risk factor from the reference value; For each risk factor, the ratio of the quantitative value to the reference value is calculated, and then the result is added by 1 and the natural logarithm is taken, to calculate the logarithmic term of each risk factor. The weight of each risk factor, the relative deviation value and the logarithmic term are multiplied to obtain the contribution value of each risk factor, and then the contribution values of all the risk factors are added to obtain the contribution value sum, and the square root of the contribution value sum is obtained, that is, the pipeline risk trend index.

Images