AI-based pipeline dredging robot regulation system

Abstract

The AI-based pipeline dredging robot regulation system belongs to the field of artificial intelligence, comprises a data acquisition unit, which acquires preset pipeline static data and dynamic data acquired by a sensor in real time; an impact force solving unit, which solves jet impact force generated by a jet on a pipeline wall based on the dynamic data; a fragility assessment unit, which assesses a pipeline wall fragility index based on the dynamic data and the static data; a damage risk judgment unit, which calculates a fatigue damage index based on the jet impact force and the pipeline wall fragility index, judges the index, generates a risk judgment result, and sends a safety or warning signal and a fatigue damage index signal to better match the logic of a flowchart; and an adaptive regulation unit, which generates a safety damping factor based on the safety signal or the risk warning signal, and generates a final regulation instruction based on the safety damping factor and preset initial operation parameters, so that the fuzziness and uncertainty of existing operation intensity assessment are changed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and particularly to an AI-based pipeline dredging robot control system. BACKGROUND

[0002] With the continuous development of modern pipeline networks, the complexity of dredging operations has significantly increased. This complexity presents many challenges, particularly in terms of manual control and operation management.

[0003] Current general dredging operations only focus on cleaning efficiency, while ignoring the potential long-term health impact on the pipeline structure. Traditional dredging methods rely on the experience of operators for subjective judgment, lacking means for quantifying and assessing the physical effects and the health status of the pipeline during the operation process. These methods, although capable of completing the dredging task, are usually inefficient and may fail to discover potential structural damage in real time, or even cause structural damage to the pipeline due to excessive scouring, thereby shortening the service life of the pipeline asset.

[0004] Therefore, how to solve the technical problem that existing dredging operations only focus on cleaning efficiency while ignoring the long-term health impact on the pipeline structure has become a problem that needs to be solved in the field.

[0005] The above information disclosed in the above background section is only used to strengthen the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0006] The present application aims to provide an AI-based pipeline dredging robot control system to solve the problems raised in the above background.

[0007] The technical solution of the present application is as follows:

[0008] A data acquisition unit is configured to acquire pre-set pipeline static data and dynamic data obtained in real time by a sensor;

[0009] An impact force solving unit is configured to solve jet impact force generated by the jet flow on the pipeline wall based on the dynamic data;

[0010] A fragility assessment unit is configured to assess a pipeline wall fragility index based on the dynamic data and the static data;

[0011] A damage risk determination unit is configured to calculate a fatigue damage index based on the jet impact force and the pipeline wall fragility index, and to perform risk judgment on the index, and to generate a signal containing the risk judgment result, safety or warning, and the fatigue damage index to better match the logic of the flowchart;

[0012] An adaptive regulation unit is configured to generate a safety damping factor based on the safety operation signal or the risk warning signal, and generate a final regulation instruction based on the safety damping factor and a preset initial operation parameter.

[0013] Preferably, the impact force calculation unit is specifically configured to:

[0014] determine a distance attenuation factor based on the distance between the nozzle and the pipe wall in the dynamic data;

[0015] and based on the jet pressure and the impact angle in the dynamic data, the jet impact force is calculated by combining the distance attenuation factor and the preset nozzle flow coefficient and nozzle cross-sectional area.

[0016] Preferably, the vulnerability evaluation unit is specifically configured to:

[0017] The acoustic signal time-frequency spectrum in the dynamic data and the pipe wall micro-topography image are taken as inputs, and a pre-trained convolutional neural network model is used for feature extraction to obtain normalized sensor feature values.

[0018] Preferably, the vulnerability evaluation unit is further configured to:

[0019] Based on the normalized sensor feature values and the pipeline service information in the static data, the pipe wall vulnerability index is calculated by a multi-factor weighting model.

[0020] Preferably, the damage risk determination unit is specifically configured to:

[0021] The product of the jet impact force and the pipe wall vulnerability index is divided by the preset material reference resistance to obtain the fatigue damage index.

[0022] Preferably, the damage risk determination unit is further configured to:

[0023] The fatigue damage index is compared and analyzed with the preset risk threshold value;

[0024] When the fatigue damage index is greater than the risk threshold value, a risk warning signal is generated;

[0025] When the fatigue damage index is less than or equal to the risk threshold value, a safety operation signal is generated.

[0026] Preferably, the adaptive regulation unit is specifically configured to:

[0027] When the risk warning signal is received, the safety damping factor is calculated by a preset exponential decay function;

[0028] When the safety operation signal is received, the safety damping factor is set to an initial value.

[0029] Preferably, the adaptive regulation unit is further configured to:

[0030] The initial command pressure in the initial operation parameter is multiplied by the safety damping factor to obtain a final command pressure;

[0031] The initial command speed in the initial operation parameter is multiplied by the safety damping factor to obtain a final command speed;

[0032] The final control command is composed of the final command pressure and the final command speed.

[0033] The AI-based pipeline dredging robot control system is provided by improving the prior art, and has the following improvements and advantages compared with the prior art.

[0034] 1. The system can accurately quantify the physical action strength of the operation; Through the impact force calculation unit, the system no longer simply depends on indirect parameters such as jet pressure, but comprehensively considers real-time dynamic data such as nozzle-to-pipe wall distance, impact angle, and jet pressure to calculate the jet impact force directly acting on the pipe wall; This design converts abstract control commands into specific physical effect measurements, providing a solid physical foundation for subsequent scientific and reliable damage risk assessment, and changing the fuzziness and uncertainty of existing operation strength evaluation;

[0035] 2. The system can deeply understand and comprehensively evaluate the internal health state of the pipeline; The vulnerability evaluation unit innovatively combines machine learning and a multi-factor weighting model; This unit not only uses a convolutional neural network model to extract deep features reflecting material micro changes from dynamic data such as acoustic signal time-frequency spectrograms and pipe wall micro-topography images, but also combines long-term aging trends represented by static data such as pipeline service information; By combining micro insights into instantaneous state with macro considerations of long-term trends, the system can generate a pipe wall vulnerability index that comprehensively reflects the current carrying capacity of the pipeline, greatly improving the ability to identify early and hidden structural damage;

[0036] 3. The system establishes an objective and standardized risk determination mechanism; The damage risk determination unit integrates the quantified jet impact force and the pipe wall vulnerability index to calculate a unified fatigue damage index, and generates a clear safe operation signal or risk warning signal according to the comparison result of the index and the preset risk threshold; This converts the fuzzy judgment that relies on the subjective experience of the operator into a repeatable and standardized quantitative risk assessment, providing a clear and unambiguous decision basis for subsequent adaptive control;

[0037] 4. The system realizes closed-loop adaptive regulation of work intensity; the adaptive regulation unit can dynamically generate a safety damping factor according to the received safety or warning signal, and use it to real-time correct the initial work parameters to generate the final regulation instruction; this mechanism ensures that the work intensity of the dredging robot can be real-time constrained within the range safe for the pipeline structure, and can quickly and smoothly reduce the work intensity when facing damage risk; seamlessly connects perception evaluation and control execution, realizes the essential transformation from blind execution to intelligent guardianship, effectively avoids structural damage of the pipeline caused by excessive flushing, significantly prolongs the service life of the pipeline asset, and upgrades the pipeline maintenance concept from passive remedial maintenance to proactive predictive health management. BRIEF DESCRIPTION OF DRAWINGS

[0038] The application will be further explained below in conjunction with the accompanying drawings and embodiments:

[0039] Figure 1 is a flow chart of the system of the application. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical scheme and advantages of the application more clear and explicit, the application will be further described in detail below in conjunction with specific embodiments.

[0041] Embodiment 1

[0042] Please refer to Figure 1 The application provides an AI-based pipeline dredging robot regulation system, which comprises:

[0043] A data acquisition unit is configured to acquire preset pipeline static data and dynamic data acquired by a sensor in real time;

[0044] An impact force solving unit is configured to solve the jet impact force generated by the jet flow on the pipeline wall based on the dynamic data;

[0045] A fragility evaluation unit is configured to evaluate the pipeline wall fragility index based on the dynamic data and the static data;

[0046] A damage risk determination unit is configured to calculate a fatigue damage index based on the jet impact force and the pipeline wall fragility index, and perform risk determination on the index to generate a signal containing the risk determination result, safety or warning and the fatigue damage index to better match the logic of the flow chart;

[0047] An adaptive regulation unit is configured to generate a safety damping factor based on a safety operation signal or a risk warning signal, and generate a final regulation instruction based on the safety damping factor and preset initial work parameters;

[0048] The embodiment of the application provides a pipeline dredging robot regulation system based on AI, aiming at solving the technical problem that existing dredging operation only pays attention to cleaning efficiency and ignores the long-term health influence on the pipeline structure; the system realizes the core target change from guaranteeing cleanliness to guaranteeing health through the construction of an intelligent closed loop of perception-modeling-evaluation-decision-execution; the system comprises a data acquisition unit, an impact force solving unit, a vulnerability evaluation unit, a damage risk judgment unit and a self-adaptive regulation unit;

[0049] The data acquisition unit aims to provide comprehensive and multi-dimensional data input for the whole regulation system, and is the basis for realizing accurate modeling and intelligent decision; in the embodiment, the unit is responsible for acquiring two types of core data:

[0050] The preset pipeline static data: before the operation starts, the system inquires the city pipe network information system through an interface to obtain inherent attributes associated with the target pipeline ID; the pipeline static data refers to parameters that do not change after the pipeline is built, for example, the construction year of the pipeline, the design maximum life , the material label, such as C30 concrete, and the design pipe diameter; these data provide a basis for evaluating the long-term aging trend of the pipeline and setting the material mechanics benchmark;

[0051] The dynamic data acquired by the sensor in real time: in the operation process, the sensor array carried by the robot continuously acquires real-time information reflecting the working state and environmental interaction; the dynamic data refers to parameters that change with time and the position of the robot, including: the three-dimensional geometric morphology of the inner wall of the pipeline acquired by the high-frequency laser scanner, which is used to build the pipe wall model in real time; the distribution of the dirt in the pipe sensed by the forward-looking sonar system; the real-time position, travel speed and jet angle of the spray head provided by the robot pose sensor ; the jet flow pressure monitored by the pressure sensor ; and the jet impact acoustic signal collected by the operation click acoustic sensor;

[0052] The impact force solving unit aims to convert the controllable operation parameters of the robot into a quantifiable index of the direct physical effect on the pipe wall, that is, the jet impact force ; the unit receives dynamic data such as jet pressure , jet angle and the like from the data acquisition unit, and calculates the instantaneous impact force of the high-pressure water jet on the pipe wall in real time through the physical model described in detail below ; the technical value lies in completing the quantitative conversion from the control instruction to the physical effect, which is the cornerstone of subsequent physical damage evaluation;

[0053] The vulnerability assessment unit aims to accurately assess the health condition of the pipeline itself, i.e., its vulnerability to external impact. Traditional assessment methods often rely on offline, periodic manual detection, which cannot reflect dynamic changes during operation. In this embodiment, the unit combines static data provided by the data acquisition unit, such as service life and dynamic data such as acoustic signals and micro-topography, through machine learning models and multi-factor weighting models described later, to output a quantitative pipe wall vulnerability index . The higher the index, the more likely the pipe wall is to be damaged under the same impact.

[0054] The damage risk judgment unit aims to comprehensively consider external actions, impact force, and internal state, vulnerability, and real-time, dynamic judgment of the risk level of the current dredging operation. This unit takes the jet impact force output by the impact force calculation unit and the pipe wall vulnerability index output by the vulnerability assessment unit as input, combines the pre-set material reference resistance , and calculates a comprehensive fatigue damage index . This unit compares the value of with the pre-set risk threshold to generate one of two signals: a safe operation signal when the risk is within an acceptable range, and a risk warning signal when the risk exceeds the safety threshold.

[0055] The adaptive control unit aims to dynamically adjust the operation parameters of the robot based on the risk judgment result, forming a closed-loop control to actively avoid potential structural damage. This unit receives signals generated by the damage risk judgment unit, including risk judgment results and fatigue damage index values. Based on these signals, the unit calculates a safety damping factor . This unit uses this factor to dynamically correct the initial operation parameters preset by the operator or recommended by the system, such as the initial command pressure and the initial command speed , to generate the final control instructions issued to the robot's actuator, i.e., the final command pressure and the final command speed .

[0056] The embodiment builds a complete intelligent operation closed loop through the cooperation of the above units; instead of blindly executing preset dredging parameters, the health feedback of the pipeline can be perceived in real time, the operation risk can be dynamically evaluated, and the operation intensity can be autonomously adjusted to the best balance point considering the cleaning efficiency and the safety of the pipeline structure; this not only avoids structural damage of the pipeline caused by excessive flushing, significantly prolongs the service life of the pipeline asset, but also improves the concept of pipeline maintenance from passive and remedial maintenance to active and predictive health management, thereby improving the operation and maintenance efficiency and safety as a whole.

[0057] The impact force solving unit is specifically used for:

[0058] determining a distance attenuation factor based on the nozzle-pipe wall distance in the dynamic data;

[0059] and based on the jet pressure and the impact angle in the dynamic data, the jet impact force is solved in combination with the distance attenuation factor and the preset nozzle flow coefficient and nozzle cross-sectional area;

[0060] In the embodiment, the implementation mode of the impact force solving unit is refined; the purpose is to build a semi-empirical physical model with more physical meaning, which can accurately quantify the impact of high-pressure water jet on the pipe wall, thereby providing more accurate and reliable data input for subsequent risk assessment;

[0061] In the embodiment, the impact force solving unit solves the jet impact force by the following steps :

[0062] To quantify the attenuation effect of jet energy in the propagation process, the unit introduces a distance attenuation factor ;

[0063] The distance attenuation factor is a dimensionless parameter between (0, 1], which represents the attenuation of jet energy in the propagation process due to momentum exchange with the surrounding fluid medium; based on the nozzle-pipe wall distance calculated by fusing the laser scanner and the robot pose sensor in the dynamic data , the current value of the distance attenuation factor is determined through a preset function relationship ; the function relationship can be obtained in advance through computational fluid dynamics simulation or experimental platform calibration;

[0064] The function form can be or , wherein and are coefficients obtained by experimental calibration;

[0065] To calibrate, a group of calibration distances and the corresponding impact force measured at this distance The experimental data set is composed of the distance and the corresponding impact force measured at this distance The function form is determined based on the data set by numerical fitting method; when the distance , , represents no attenuation;

[0066] Based on the above parameters, the impact force calculation unit uses the following semi-empirical formula to calculate the instantaneous impact force according to the impact jet theory :

[0067] ;

[0068] The construction of this formula aims to convert multiple indirect operation control parameters into physical quantities that directly cause material fatigue;

[0069] The derivation of this formula is based on the momentum theorem, that is, the impact force generated by the jet on the pipe wall is equal to the change in momentum of the jet per unit time; in the ideal case, when the high-pressure water jet vertically impacts the plane, the impact force is ; wherein, is the density of water, unit: , Q is the volume flow rate, unit: , v is the jet velocity, unit: ; Through Bernoulli equation, the relationship between velocity and pressure can be obtained, that is:

[0070] ;

[0071] The flow rate can be represented as:

[0072] ;

[0073] Therefore, the ideal impact force can be represented as:

[0074] ;

[0075] Considering the influence of the nozzle flow coefficient and the impact angle , as well as the distance attenuation factor , the formula is modified as:

[0076] ;

[0077] Thus, the actual impact force acting on the pipe wall is more accurately quantified;

[0078] wherein, : Jet impact force, the core physical input for evaluating subsequent fatigue damage, unit in Newton, N, is the final output of this unit;

[0079] : Nozzle flow coefficient, a dimensionless parameter, reflects the inherent performance of the nozzle; its source is to consult the nozzle design manual or to calibrate through flow test, in this embodiment, according to the nozzle model used, its value is preset between 0.6 and 0.95;

[0080] : Jet pressure, one of the key parameters that can be controlled by the robot, is monitored in real time by the pressure sensor at the outlet of the robot water pump, unit in Pascals, Pa, derived from dynamic data;

[0081] : Nozzle cross-sectional area, an inherent design parameter of the robot nozzle, unit in square meters, m², stored in the system as a preset parameter;

[0082] : Distance decay factor, as above, determined in real time by the distance between the nozzle and the pipe wall ;

[0083] : Impact angle, refers to the angle between the nozzle axis and the normal direction of the pipe wall surface at this point; it is calculated by the robot pose sensor and the local three-dimensional model of the pipe wall constructed in real time according to the laser scanning data, derived from dynamic data;

[0084] : Function for determining the distance decay factor according to the distance between the nozzle and the pipe wall ;

[0085] This formula is a simplified macroscopic physical model that ignores complex microscopic fluid dynamics effects. Its effectiveness is highly dependent on the accurate calibration of the distance decay factor , which should be done through a large amount of experimental data or high-fidelity CFD simulation;

[0086] The behavior of this formula under different input values conforms to physical common sense: when the nozzle is far away from the pipe wall , the distance decay factor , the impact force reaches the maximum value; when the impact angle , the jet is parallel to the pipe wall, , the impact force tends to zero, which verifies the physical robustness of the model;

[0087] Compared with the prior art of using only a single parameter such as water pump pressure to roughly estimate the work intensity, the embodiment can more truly and accurately reflect the actual physical action of the flow on the pipe wall by introducing a multi-factor physical model that comprehensively considers pressure, distance, angle, and nozzle performance; this accurate quantification eliminates uncertainty in the evaluation process, enabling the subsequent fatigue damage risk determination to be based on a more solid physical foundation, thereby significantly improving the accuracy and reliability of the entire system regulation decision.

[0088] Embodiment 2

[0089] The vulnerability assessment unit is specifically used for:

[0090] The acoustic signal time-frequency spectrogram in the dynamic data and the pipe wall microscopic morphology image are taken as inputs, and a pre-trained convolutional neural network model is used for feature extraction to obtain normalized sensor feature values.

[0091] The vulnerability assessment unit is also used for:

[0092] Based on the normalized sensor feature values and the pipe service information in the static data, a multi-factor weighted model is used to calculate a pipe wall vulnerability index.

[0093] In the embodiment, the implementation of the vulnerability assessment unit is described in detail; the innovation lies in that by fusing a deep learning model and a multi-factor weighted model, microscopic insight into the pipe wall health condition is achieved, and early and hidden damage that is difficult to find by traditional detection methods can be identified.

[0094] In the embodiment, the vulnerability assessment unit uses a pre-trained convolutional neural network model to deeply mine hidden state features in the sensor data.

[0095] The vulnerability assessment unit takes the acoustic signal time-frequency spectrogram in the dynamic data and the pipe wall microscopic morphology image as dual-channel inputs, and uses a pre-trained convolutional neural network model for feature extraction; the CNN model here is trained with a large number of labeled samples containing pipe wall data of different health conditions, so that it can automatically learn and identify deep features related to the degradation of the pipe wall mechanical performance, such as changes in material density or the appearance of micro-cracks; the output of the model is normalized sensor feature values. ;

[0096] The CNN model adopts a multi-layer structure, including but not limited to: an input layer that accepts acoustic signal time-frequency spectrograms and pipe wall microscopic morphology images, multiple convolutional layers for automatically extracting spatial and temporal features in images and time-frequency spectrograms, pooling layers for dimension reduction and feature compression, and fully connected layers for mapping extracted features to the final output. The first layer can use a convolutional kernel with a size of 3x3 and be coupled with a ReLU activation function, followed by a max-pooling layer, and this process is repeated to build a multi-layer structure. The final output layer is a linear activation layer to output continuous normalized sensor feature values.

[0097] The CNN model is trained using a supervised learning approach, which requires a large-scale labeled data set. The data set is composed of pipe samples in different health states, each sample containing the following data pairs:

[0098] Input data: acoustic signal time-frequency spectrograms and microscopic morphology images inside the pipe;

[0099] Output label: a continuous numerical label representing the pipe vulnerability level, labeled by professionals or through physical detection methods such as ultrasonic flaw detection. During the model training process, the internal parameters of the CNN model, including sensor feature weights , are automatically optimized to minimize the error between the predicted value and the true label. Therefore, is the result of automatic learning of the model, reflecting the importance of different sensor features in vulnerability assessment;

[0100] Normalized sensor feature value refers to the neuron activation value before the last hidden layer and the classification output layer of the CNN model. The CNN model is modified here, with the last layer no longer being a softmax layer for classification, but a linear activation layer for regression, to directly output continuous feature values and undergo standardization processing such as Min-Max or Z-score. Selecting this layer output as the feature is because it integrates all the abstract information from the previous layers, providing the most concise and high-dimensional representation of the input data, which is most suitable for representing the complex health state of the pipe wall;

[0101] To obtain a more comprehensive health assessment, the vulnerability assessment unit also combines the instantaneous state monitored by the real-time sensor with the aging trend based on historical data;

[0102] This unit calculates the pipe wall vulnerability index based on the normalized sensor feature value and the pipe service information in static data, which is the current service life and the designed maximum life ; The form of this self-defined comprehensive evaluation model is as follows:

[0103] ;

[0104] This model aims to combine the instantaneous state monitored by real-time sensors with the aging trend based on historical data to get a more comprehensive health assessment;

[0105] wherein, : pipe wall vulnerability index, the final output of this unit, is a dimensionless number of ; the reference value 1 represents a brand-new, perfect pipeline, and the larger the value, the more vulnerable the pipe wall is;

[0106] : sensor feature weight, these weight values are not manually set, but are internal parameters automatically optimized by the CNN model during supervised learning training. The entire system can be trained as an end-to-end deep learning model; the input is the original sensor data and static data, and the output is the vulnerability index ; through back propagation and gradient descent algorithm, the and in the model can be optimized simultaneously to minimize the error between the predicted value and the true labeled value, reflecting the contribution of different sensor features to vulnerability;

[0107] : the kth normalized sensor feature value, which is extracted by the CNN model from the original sensor data in real time;

[0108] the total number of extracted feature values; : summation operation from the first feature value to the Nth feature value;

[0109] : aging influence weight, this parameter is optimized together with in end-to-end training. During the training phase, the model learns how to balance the influence of real-time sensor features and long-term service information on vulnerability; this ensures that the value of β is based on actual data rather than empirical values, thereby improving the accuracy of the model;

[0110] : current service life of the pipeline, derived from static data;

[0111] : maximum design life of the pipeline, derived from static data;

[0112] Although this model simplifies the aging process as a linear relationship with service life, in more complex applications, more influencing factors such as soil corrosion index, geological subsidence rate, etc. can be introduced to construct a more accurate nonlinear aging model;

[0113] The embodiment realizes automatic extraction of deep features related to damage from original and high-dimensional sensor data by introducing a CNN model, overcomes the limitations of traditional methods that rely on manual feature engineering and are difficult to find early hidden damage, and further combines sensor features representing instantaneous state with service aging effects representing long-term trends through a multi-factor weighting model, so that the final vulnerability assessment result can not only sensitively capture local and sudden structural changes, but also macroscopically consider the overall life cycle stage of the pipeline, and the assessment result is more comprehensive and accurate, which provides a key prerequisite for the whole adaptive regulation system to realize accurate operation.

[0114] Embodiment 3

[0115] The damage risk determination unit is specifically used for:

[0116] multiplying the jet impact force and the pipe wall vulnerability index, and dividing the preset material reference resistance to obtain a fatigue damage index;

[0117] The damage risk determination unit is further used for:

[0118] comparing and analyzing the fatigue damage index with a preset risk threshold;

[0119] generating a risk warning signal when the fatigue damage index is greater than the risk threshold;

[0120] generating a safe operation signal when the fatigue damage index is less than or equal to the risk threshold;

[0121] In the embodiment, the working logic of the damage risk determination unit is refined, and the purpose is to establish clear and quantifiable risk assessment standards, and to convert complex physical actions and state assessment results into direct driving, explicit risk level signals for subsequent decision-making;

[0122] In the embodiment, the core task of the damage risk determination unit is to calculate the fatigue damage index and make a determination accordingly;

[0123] The unit is specifically used for multiplying the jet impact force and the pipe wall vulnerability index , and dividing the preset material reference resistance to obtain a fatigue damage index ; the calculation formula is as follows:

[0124] ;

[0125] The internal logic of this formula is that the operation risk is proportional to the product of external impact and internal vulnerability , and the inherent resistance of the material Normalization is performed to obtain a relative risk indicator with clear physical meaning;

[0126] wherein, : jet impact force, the value is calculated in real time by the aforementioned impact force calculation unit and transmitted to this;

[0127] : pipe wall fragility index, the value is evaluated in real time by the aforementioned fragility evaluation unit and transmitted to this;

[0128] : material reference resistance, which is a pre-set reference physical quantity representing the impact force threshold that the specific pipe material can withstand in an ideal state without causing significant fatigue damage accumulation; Its source mainly depends on the pipe material, such as the material mechanics properties of C30 concrete, which can be set by referring to the S-N curve in the relevant design specification or material fatigue experiment database, stress-life curve;

[0129] For common C30 concrete pipes, according to industry standards and experimental data, The value of XX Newton can be set as the impact force threshold that the pipe material can withstand without causing significant fatigue damage. XX needs to be filled according to actual technical data to ensure implementability;

[0130] Material reference resistance The setting of can refer to industry standards and experimental data; For example, for common C30 concrete pipes, the fatigue life under different stress levels can be determined through fatigue experiments, S-N curve; Can be set as the impact force value corresponding to the fatigue limit in the S-N curve, that is, below this force value, the material can withstand infinite cyclic loads without fatigue failure;

[0131] After calculating the fatigue damage index The damage risk determination unit is also used to compare and analyze the fatigue damage index with the pre-set risk threshold; In this embodiment, the core risk threshold is set to 1;

[0132] When the fatigue damage index Is less than or equal to the risk threshold 1, the system determines that the current operation is in the safe interval, and generates a safe operation signal;

[0133] When the fatigue damage index Is greater than the risk threshold 1, the system determines that the operation has entered the risk interval, which may cause irreversible cumulative damage to the pipe structure, and generates a risk warning signal;

[0134] In a more detailed control strategy, multiple thresholds can also be set, for example, Defined as a Level 1 warning, Defined as a Level 2 early warning system to enable more tiered regulation;

[0135] This index reflects instantaneous risk, while a more comprehensive model should incorporate a historical damage accumulation term. This makes the final risk assessment as Historical damage accumulation can be achieved by analyzing each operation. The value is obtained by integrating or weighting the average.

[0136] The accuracy of this model is highly dependent on the quality of the training dataset and the rigor of the annotations. To improve physical fidelity, the labeled data should be cross-validated using multiple non-destructive testing techniques, such as ultrasound and X-ray, to ensure the objectivity and accuracy of the labels.

[0137] This embodiment defines and calculates the fatigue damage index. Successfully combined multiple parameters from different modules, each with different physical meanings, to determine the impact force. Vulnerability Material durability It is integrated into a single, intuitive, and standardized risk metric; this quantitative risk assessment method replaces the vague judgment that relies on the operator's subjective experience, making the risk determination process objective and repeatable; and by comparing with clear risk thresholds, it can generate clear and unambiguous decision signals, whether safe or warning, providing a direct and reliable driving basis for subsequent adaptive control units, and is a key link connecting assessment and control.

[0138] Example 4

[0139] The adaptive control unit is specifically used for:

[0140] When a risk warning signal is received, the safety damping factor is calculated using a preset exponential decay function;

[0141] When a safe operation signal is received, the safety damping factor is set to its initial value;

[0142] The adaptive control unit is also used for:

[0143] Multiply the initial command pressure in the initial operating parameters by the safety damping factor to obtain the final command pressure;

[0144] The initial command speed in the initial operating parameters is multiplied by the safety damping factor to obtain the final command speed;

[0145] The final control command consists of the final command pressure and the final command speed;

[0146] Based on the above embodiments, this embodiment elaborates on the decision-making and execution logic of the adaptive control unit. The core innovation lies in the design of a non-linear and smooth control strategy to ensure that the system can react quickly and stably when facing risks, thereby achieving strong and precise constraints on the intensity of work.

[0147] In this embodiment, the adaptive control unit generates a safety damping factor in different ways based on the received signal. :

[0148] When a risk warning signal is received from the damage risk assessment unit, that is The unit calculates the safety damping factor using a preset exponential decay function. The logic of this function is that once the operation enters the risk zone, a damping mechanism is immediately activated, and the damping rate increases with the risk index. As the value increases, the damping effect increases exponentially; the calculation formula is:

[0149] ;

[0150] When a safe operation signal is received, i.e. This means that the current workload does not need to be suppressed, and the unit will then apply a safety damping factor. Set the initial value to 1;

[0151] in, Safety damping factor is a dimensionless multiplier between (0, 1] that directly affects the preset operating parameters.

[0152] The sensitivity coefficient is a dimensionless positive number; it is a hyperparameter that can be preset by maintenance engineers according to the maintenance strategies of different pipelines, and determines the severity of the control system's response to risks.

[0153] The base of the natural logarithm ;

[0154] Control sensitivity coefficient The settings should be adjusted according to the specific application scenario and security requirements; for example, in high-risk, high-value pipelines, such as dredging operations of water supply networks under urban main roads, the settings should be adjusted accordingly. Set to a larger value to ensure that the system can quickly and significantly reduce operational intensity when a risk signal occurs, prioritizing pipeline safety; while in low-risk scenarios, it can be... The settings are set relatively low to maintain operational efficiency as much as possible while ensuring basic safety.

[0155] For scenarios with high security requirements, The value can be set in the range of 5-10; while for low-risk scenarios, it can be set in the range of 1-3, and the specific optimal value should be fine-tuned through field experiments and historical data;

[0156] A larger value means that the system is more sensitive to reaction, which is suitable for scenarios with extremely high safety requirements;

[0157] Fatigue damage index, calculated in real time by the risk assessment system and passed to this module;

[0158] After generating the safety damping factor , the adaptive control unit is also used to generate the final control instruction based on the factor and the preset initial operation parameters ;

[0159] The index decay function has good robustness: when the fatigue damage index exceeds the risk threshold 1 and continues to increase, the safety damping factor will quickly and continuously approach zero. Compared with simple step control, this exponential decay ensures smooth transition of operation parameters, avoiding additional impact on the pipeline and robot system due to sudden changes, thereby effectively suppressing operation intensity. When , is equal to 1, ensuring efficiency during safe operation;

[0160] This unit is used to:

[0161] Multiply the initial instruction pressure in the initial operation parameters by the safety damping factor to obtain the final instruction pressure :

[0162] ;

[0163] Multiply the initial instruction speed in the initial operation parameters by the safety damping factor to obtain the final instruction speed :

[0164] ;

[0165] The final control instruction is composed of the final instruction pressure and the final instruction speed , and is issued to the hydraulic pump and drive motor of the robot for execution; this strategy ensures that when the risk index rises, , the operation intensity, reflected in pressure and speed, will be smoothly and automatically reduced;

[0166] The embodiment realizes a nonlinear, responsive and smooth adaptive control mechanism by introducing a safety damping factor based on an exponential decay function; compared with simple linear weakening or start-stop control, the method can intervene gently when the risk just appears and suppress strongly when the risk rises sharply, thereby ensuring safety and maximizing work efficiency; by multiplying the safety damping factor with the initial work parameters, the method directly generates executable control instructions, thereby building a seamless closed loop from risk assessment to work execution; this logic not only greatly improves the intelligence and safety of the dredging work, but also accurately avoids damage, thereby practically improving the maintenance target to a new height of structural health and asset life cycle maximization.

[0167] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application rather than limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. An AI-based pipeline dredging robot control system, characterized in that, include: The data acquisition unit is used to acquire preset static pipeline data and dynamic data acquired in real time by sensors; The impact force calculation unit is used to calculate the jet impact force generated by the jet on the pipe wall based on dynamic data; The vulnerability assessment unit is used to evaluate the vulnerability index of the pipe wall based on dynamic and static data. The damage risk assessment unit is used to calculate the fatigue damage index based on the jet impact force and the pipe wall fragility index, and to assess the risk of the index. It generates a signal that includes the risk assessment result, safety or warning and fatigue damage index, so as to better match the logic of the flowchart. The damage risk assessment unit is also used for: The fatigue damage index was compared and analyzed with the preset risk threshold. When the fatigue damage index exceeds the risk threshold, a risk warning signal is generated. When the fatigue damage index is less than or equal to the risk threshold, a safe operation signal is generated; An adaptive control unit is used to generate a safety damping factor based on a safe operation signal or a risk warning signal. Based on the safety damping factor and preset initial operating parameters, the final control command is generated; The vulnerability assessment unit is specifically used for: The acoustic signal time-spectrum diagram and the microscopic morphology image of the pipe wall in the dynamic data are used as inputs, and a pre-trained convolutional neural network model is used to extract features to obtain normalized sensor feature values. The vulnerability assessment unit is also used for: Based on the normalized sensor feature values ​​and pipeline service information in static data, the pipe wall fragility index is calculated through a multi-factor weighted model. The damage risk assessment unit is specifically used for: The fatigue damage index is obtained by dividing the product of the jet impact force and the pipe wall fragility index by the preset material reference tolerance.

2. The AI-based pipeline dredging robot control system according to claim 1, characterized in that, The impact force calculation unit is specifically used for: Determine the distance attenuation factor based on the nozzle-to-pipe wall distance in dynamic data; Based on the jet pressure and impact angle in the dynamic data, combined with the distance attenuation factor and the preset nozzle flow coefficient and nozzle cross-sectional area, the jet impact force is calculated.

3. The AI-based pipeline dredging robot control system according to claim 1, characterized in that, The adaptive control unit is specifically used for: When a risk warning signal is received, the safety damping factor is calculated using a preset exponential decay function; When a safe operation signal is received, the safety damping factor is set to its initial value.

4. The AI-based pipeline dredging robot control system according to claim 3, characterized in that, The adaptive control unit is also used for: Multiply the initial command pressure in the initial operating parameters by the safety damping factor to obtain the final command pressure; The initial command speed in the initial operating parameters is multiplied by the safety damping factor to obtain the final command speed; The final control command consists of the final command pressure and the final command speed.

Images