Unmanned underwater vehicle biological attachment fault diagnosis method and system based on parameter identification

Abstract

The invention discloses an unmanned underwater vehicle biological attachment fault diagnosis method and system based on parameter identification, and belongs to the technical field of intelligent operation and maintenance of unmanned underwater vehicles, and the method comprises the steps: S1, collecting time sequence data generated in the operation process of an unmanned underwater vehicle in real time; s2, constructing a behavior agent model: constructing a buoyancy-speed model reflecting the resistance performance by using the buoyancy regulating variable and the longitudinal speed; constructing a pitch angle dynamic model reflecting the longitudinal control performance by using the pitch angle and the elevator angle; constructing a roll angle dynamic model reflecting the transverse control performance by using the roll angle and the rudder angle; s3, identifying parameters of the behavior agent model in real time; s4, constructing a feature vector and establishing a reference; and S5, carrying out fault detection, and comprehensively judging whether the three features deviate from a normal range at the same time or not. According to the method, attachment types such as marine organism attachment and foreign matter winding can be efficiently identified, and high-level, medium-level and low-level early warning is triggered according to the severity degree.

Description

Technical Field

[0001] This invention belongs to the field of intelligent operation and maintenance technology for underwater unmanned vehicles, specifically relating to a method and system for diagnosing bio-attachment faults in unmanned underwater vehicles based on parameter identification. Background Technology

[0002] When unmanned underwater vehicles (UUVs) operate in the marine environment for extended periods, their hull surfaces are prone to biofouling caused by the adhesion of marine organisms such as barnacles and algae. This biofouling accumulates gradually, leading to multiple negative impacts: (1) significantly increased fluid resistance, reducing gliding efficiency and thus shortening endurance and increasing energy consumption; (2) altering the mass distribution of the UUV, causing attitude imbalances such as pitch or roll deviations, requiring continuous control surface corrections to maintain the intended trajectory, which not only increases energy consumption but also burdens the control system. If not diagnosed and addressed in a timely manner, biofouling may further lead to mission failure or even equipment damage, such as propeller overload or structural damage due to excessive drag.

[0003] Existing technologies mostly employ fixed threshold methods for anomaly detection, such as classifying anomalies as occurring when a depth deviation exceeds ±5 meters or a rudder angle consistently greater than 10°. However, such methods have three inherent drawbacks: First, they lack adaptability, failing to consider the differences in normal operating conditions across different sea areas (e.g., temperature and salinity variations) or different mission phases (e.g., acceleration and cruising), leading to false alarms or missed alarms; second, they cannot distinguish between attachment types, failing to determine whether it is uniform attachment (increased overall drag) or localized non-uniform attachment (e.g., torque deviation caused by unilateral attachment), thus affecting the formulation of handling strategies; finally, they lack positioning capabilities, unable to indicate the specific area where attachment may occur (e.g., head, tail, left, or right), limiting targeted maintenance.

[0004] In recent years, some studies have attempted to introduce machine learning methods, such as identifying attachment patterns through supervised learning. However, these methods rely on a large number of labeled fault samples for training, while in actual marine missions, fault data is scarce and costly to acquire, making it difficult to meet the needs of the models. Therefore, there is an urgent need for an online diagnostic method that does not require prior fault data and only utilizes conventional navigation logs (such as sensor readings and control commands). This method needs to possess mathematical rigor (e.g., based on statistical models or state estimation) and engineering feasibility (e.g., low computational complexity and real-time operation) to improve the autonomy and reliability of UUVs. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for diagnosing bio-attachment faults of unmanned underwater vehicles based on parameter identification, overcoming the shortcomings of existing technologies such as fixed thresholds, inability to identify attachment types, and reliance on fault samples.

[0006] To achieve the above-mentioned objectives, the first objective of this invention is a method for diagnosing bio-attachment faults in unmanned underwater vehicles based on parameter identification, comprising:

[0007] S1. Real-time acquisition of the following time-series data generated by the unmanned underwater vehicle during operation: buoyancy adjustment, longitudinal speed, pitch angle, roll angle, elevator angle, and rudder angle; S2. Construct a behavioral agent model, including: A buoyancy-velocity model reflecting drag performance is constructed using buoyancy adjustment and longitudinal velocity. A dynamic pitch angle model reflecting longitudinal control performance is constructed using pitch angle and elevator angle. A dynamic roll angle model reflecting lateral control performance is constructed using roll angle and rudder angle. S3. Real-time identification of behavioral proxy model parameters to obtain drag coefficient proxy, pitch control surface response coefficient and yaw control surface response coefficient. S4. Construct feature vectors and establish a baseline, including: Acquire data from multiple sampling points before the initial stage of the task; Normalize the drag coefficient proxy, pitch surface response coefficient, and yaw surface response coefficient to obtain normalized eigenvectors. The mean vector and covariance matrix of statistically normalized eigenvectors; S5. Fault detection: The Hotelling statistic is used to comprehensively determine whether the three characteristics deviate from the normal range at the same time. If the three characteristics deviate from the normal range at the same time, there is a fault; otherwise, there is no fault.

[0008] Preferably, S1 includes: filtering the data.

[0009] Preferably, in S2: The mathematical expression for the buoyancy-velocity model is:

[0010] in: This is a constant term, reflecting constant environmental disturbances; This is the gain coefficient of buoyancy on velocity, characterizing propulsion efficiency; is the velocity autocorrelation coefficient, with a value range of (0,1); The model residuals include unmodeled dynamics and random perturbations. The mathematical expression for the pitch angle dynamic model is:

[0011] in: This is the first constant bias term; The pitch angle autocorrelation coefficient reflects attitude inertia; The elevator response coefficient represents the pitch angle change caused by a unit rudder deflection; For pitch channel model residuals; The mathematical expression for the roll angle dynamic model is:

[0012] in: This is the second constant bias term; This is the autocorrelation coefficient of the roll angle; This refers to the rudder response coefficient; This represents the residual of the roll channel model.

[0013] Preferably, S3 includes: using a recursive least squares method with a forgetting factor to identify the parameters of the behavioral agent model in real time; Recursive least squares method with forgetting factor introduces forgetting factor More recent data is given higher weight, thus enabling online parameter updates; the following key parameters are output: Resistance coefficient agent quantity: ; Pitch control surface response coefficient: ; Yaw control surface response coefficient: .

[0014] Preferably, S4 includes: S4. Construct feature vectors and establish a baseline, including: Obtain the first N sampling points in the initial stage of the task. The initial running time of the task shall not be less than 10 minutes; Definition of normalized eigenvectors:

[0015] in: This represents the sample mean of the resistance surcharge during the normal phase. The pitch control effect is the sample mean. This represents the sample mean of the roll rudder effect. Statistical baseline calculation: The formula for calculating the mean vector is: , is a 3-dimensional column vector; The formula for calculating the covariance matrix is: , is a 3×3 positive definite symmetric matrix.

[0016] Preferably, it also includes S6, fault type identification, which first uses a contribution map to quantify each feature pair. The contribution level exceeding the limit is then used to determine the fault type using a mapping rule table; the mapping rule table includes the maximum contribution item, its corresponding physical meaning, and the attachment type determination.

[0017] Preferably, it also includes S7, which triggers a graded early warning mechanism based on the severity of the fault.

[0018] A second objective of this invention is to provide a bio-attachment fault diagnosis system for unmanned underwater vehicles based on parameter identification, comprising: The data acquisition module collects the following time-series data generated by the unmanned underwater vehicle in real time during operation: buoyancy adjustment, longitudinal speed, pitch angle, roll angle, elevator angle, and rudder angle; The behavioral agent model building module includes: A buoyancy-velocity model reflecting drag performance is constructed using buoyancy adjustment and longitudinal velocity. A dynamic pitch angle model reflecting longitudinal control performance is constructed using pitch angle and elevator angle. A dynamic roll angle model reflecting lateral control performance is constructed using roll angle and rudder angle. The estimation module uses recursive least squares with a forgetting factor to identify the parameters of the behavioral surrogate model in real time, and obtains the drag coefficient surrogate, pitch surface response coefficient and yaw surface response coefficient. Constructing feature vectors and establishing a baseline module includes: Acquire multiple sampling points before the initial stage of the task; Normalize the drag coefficient proxy, pitch surface response coefficient, and yaw surface response coefficient to obtain normalized eigenvectors. The mean vector and covariance matrix of statistically normalized eigenvectors; The fault detection module uses the Hotelling statistic to comprehensively determine whether three features deviate from the normal range simultaneously. If all three features deviate from the normal range simultaneously, there is a fault; otherwise, there is no fault.

[0019] A third objective of this invention is to provide a computer program product, comprising a computer program that is executed by a processor of the above-described method for diagnosing bio-attachment faults in unmanned underwater vehicles based on parameter identification.

[0020] A fourth objective of this invention is to provide a computer-readable storage medium including instructions that, when executed on a computer, cause the computer to perform the aforementioned parameter identification-based bio-attachment fault diagnosis method for unmanned underwater vehicles.

[0021] Compared with the prior art, the present invention has the following technical effects: This invention utilizes real-time acquisition of conventional sensor data such as buoyancy, speed, and angle from the underwater vehicle during navigation to construct three ARX (autoregressive exogenous input) models indexed by a unified timestamp: buoyancy-speed, elevator-pitch angle, and rudder-roll angle, accurately describing the system's dynamic characteristics. It employs recursive least squares (FFRLS) with a forgetting factor to estimate drag coefficients and control surface response coefficients online in real time, normalizing them into feature vectors to eliminate dimensional influences and enhance feature comparability. A multivariate statistical control model is established based on data from the initial normal navigation phase of the mission, using Hotelling's... The test performs hypothesis testing on real-time feature vectors to achieve early fault detection. Once an anomaly is detected, the contribution map analysis method is further used to locate the key variables that cause the anomaly, thereby identifying attachment types such as marine organism attachment and foreign object entanglement, and triggering high, medium and low level warnings based on the severity.

[0022] This invention eliminates the need for dynamic modeling: it relies solely on sensor data collected during navigation and the system's input-output relationship to complete the modeling, thus avoiding complex and uncertain first-principles analysis of fluid dynamics. This invention is highly adaptable: it automatically establishes statistical benchmarks using normal data collected at the beginning of the mission, and can adapt to different sea conditions, sea areas and mission profiles, and has online self-learning capabilities. This invention is highly interpretable: it visually identifies anomalous variables and their contribution through a contribution graph, clearly indicating possible attachment types and locations; This invention is computationally lightweight: FFRLS and Hotelling's T² verification calculations are highly efficient and can run in real time in embedded systems, meeting the long-term on-site monitoring requirements of underwater vehicles; This invention requires no fault samples: the statistical model is trained entirely based on normal system operation data, without relying on historical fault data, which is more in line with the reality of scarce fault samples in actual marine missions; This invention has strong anti-interference capabilities: it effectively suppresses uncertain interferences such as water flow and waves in the marine environment through signal filtering and multivariate statistical methods, thereby reducing the false alarm rate; This invention is applicable to various long-term deployed unmanned underwater vehicles and can significantly improve their mission success rate and equipment operational reliability in complex marine environments. Attached Figure Description

[0023] Figure 1 The flowchart provided for the first preferred embodiment of the present invention; Figure 2 The flowchart provided for the second preferred embodiment of the present invention; Figure 3 The flowchart provided is for the third preferred embodiment of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. The described embodiments are only a part of the embodiments of the present invention, and not all of them. Generally, the embodiments of the present invention described and shown in the accompanying drawings are characteristic technologies and solutions. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0025] For the first embodiment, please refer to... Figure 1 A method for diagnosing bio-attachment faults in unmanned underwater vehicles based on parameter identification, comprising: S1. Data Acquisition: Real-time acquisition of the following time-series data generated by the unmanned underwater vehicle during operation: buoyancy adjustment. Longitudinal velocity Pitch angle Roll angle Elevator angle and rudder angle ; In a preferred embodiment, the sampling frequency in this step is not less than 1Hz; For discrete sampling time indexes, Corresponding physical time The sampling period Second; This is the buoyancy adjustment amount, representing the working intensity of the buoyancy pump or the net buoyancy control command, with units of milliliters (mL) or newtons (N). The longitudinal velocity represents the forward speed of the submersible along its longitudinal axis, measured in meters per second (m / s), and can be obtained through differential depth sensor readings or DVL velocity measurement. The pitch angle represents the angle between the aircraft's longitudinal axis and the local horizontal plane, with pitch being positive. The unit is radians (rad), and it is output by the inertial measurement unit (IMU). The roll angle represents the angle of rotation of the aircraft around its longitudinal axis, with rightward tilt being positive. The unit is radians (rad), and it is output by the IMU. Elevation is the angle of deflection of the control surface that controls pitch motion, measured in radians (rad). The rudder angle is the angle of deflection of the rudder surface that controls yaw or induces roll, and is measured in radians (rad).

[0026] As a preferred embodiment, based on this step, the above-mentioned sampled data can be subjected to preliminary filtering processing (such as Kalman filtering) to reduce interference from sensor noise and ocean current disturbances, and improve the accuracy of subsequent modeling.

[0027] S2. Construct a behavioral agent model, including: A buoyancy-velocity model reflecting drag performance is constructed using buoyancy adjustment and longitudinal velocity. A dynamic pitch angle model reflecting longitudinal control performance is constructed using pitch angle and elevator angle. A dynamic roll angle model reflecting lateral control performance is constructed using roll angle and rudder angle. To avoid complex fluid dynamics modeling processes and improve model practicality, this embodiment uses a data-driven first-order autoregressive exogenous model (ARX) to describe the key dynamic behaviors of the submersible. The ARX model has a simple structure and high computational efficiency, effectively capturing the dynamic relationship between system input and output, and does not rely on complex physical mechanism modeling, making it particularly suitable for the design of control systems with obvious input-output response characteristics.

[0028] Based on this, this embodiment establishes three independent ARX models, each expressed in discrete-time form where "the current output depends on the input and state at the current or previous moment." This modeling strategy fully considers the temporal coupling relationship between control variables and state variables in the actual system, and can effectively characterize the dynamic response characteristics of the submersible in motion control.

[0029] Buoyancy-velocity model (reflecting changes in fluid resistance)

[0030] Symbol definition: : Constant term, reflecting constant environmental disturbances (such as background water flow), dimensionless; : The gain coefficient of buoyancy on velocity, which characterizes the propulsion efficiency, and is expressed in (m / s) / unit buoyancy; : Velocity autocorrelation coefficient, dimensionless, typically ranging from (0,1); Model residuals include unmodeled dynamics and random disturbances.

[0031] Model Interpretation: This model describes the longitudinal velocity at the current moment. How is it affected by the current buoyancy input? Speed ​​at the previous moment Impact. As bioattachment increases, fluid resistance increases, making it more difficult to maintain velocity, which manifests as... Significantly reduced. Therefore, It can be used as a proxy indicator for changes in resistance.

[0032] Pitch angle dynamic model (reflecting elevator efficiency)

[0033] Parameter definition: The first constant bias term is dimensionless. Pitch angle autocorrelation coefficient, reflecting attitude inertia, dimensionless; Elevator response coefficient, in rad / rad, represents the pitch angle change caused by a unit rudder deflection; : Pitch channel model residuals.

[0034] Model Interpretation: This model describes the current pitch angle. From the posture of the previous moment elevator command from the previous moment Joint decision. Introduction The timing of the rudder angle is to match the physical delay in the rudder surface response (it takes time for the rudder action and hydrodynamic setup to occur). If non-uniform adhesion occurs at the nose or tail, the rudder effectiveness decreases. It will decrease.

[0035] Roll angle dynamic model (reflecting rudder efficiency)

[0036] Parameter definition: : Second constant bias term; : Roll angle autocorrelation coefficient; : Rudder response coefficient, in rad / rad; : Residual of the roll channel model.

[0037] Model Interpretation: This model describes the current roll angle. The relationship between the historical roll angle and the rudder angle at the previous moment. If adhesion occurs on the left or right side, lateral control capability decreases. Decrease.

[0038] The three models mentioned above are independent of each other, and respectively capture the changes in three key indicators: drag performance, longitudinal control performance, and lateral control performance, forming the basis for fault diagnosis.

[0039] S3. Real-time identification of behavioral proxy model parameters to obtain drag coefficient proxy, pitch control surface response coefficient and yaw control surface response coefficient. The core objective of this step is to perform online parameter estimation, which mainly involves using Forgetting Factor Recursive Least Squares (FFRLS) to identify the model parameters in real time.

[0040] FFRLS Principle Explanation: Traditional least squares methods weight all historical data equally, making it difficult to track slow time-varying systems. FFRLS introduces a forgetting factor. This invention assigns higher weight to recent data, thereby enabling online parameter updates. This indicates that the estimation results are dominated by approximately the most recent 100 samples, which are suitable for slow degradation processes such as bioattachment.

[0041] Output key parameters: Resistance coefficient agent quantity: ; Pitch control surface response coefficient: ; Yaw control surface response coefficient: ; These three time-varying parameters constitute the core features for subsequent fault diagnosis. Their physical meaning is clear, and they can be estimated online using conventional sensor data.

[0042] S4. Construct feature vectors and establish a baseline, including: Acquire multiple sampling points before the initial stage of the task; Normalize the drag coefficient proxy, pitch surface response coefficient, and yaw surface response coefficient to obtain normalized eigenvectors. The mean vector and covariance matrix of statistically normalized eigenvectors; Based on this preferred embodiment, the following examples are provided for explanation: In the early stages of the task (before...) One sampling point, (corresponding to an operating time of no less than 10 minutes), the submersible hull is clean and considered to be in normal operating condition. Statistical benchmarks are established using data from this phase.

[0043] Definition of normalized eigenvectors: ; in: : Sample mean of resistance surcharge during normal phase; : Sample mean of pitch control effect; : Sample mean of roll rudder effect; The purpose of normalization is to eliminate differences in the dimensions and amplitudes of different parameters, make the features comparable, and facilitate multivariate fusion analysis.

[0044] Statistical baseline calculation: Mean vector: , is a 3-dimensional column vector; Covariance matrix: , is a 3×3 positive definite symmetric matrix.

[0045] It not only reflects the fluctuation range of each feature, but also characterizes the correlation between them (for example, increased drag may be accompanied by decreased rudder effectiveness), providing a statistical basis for subsequent multivariate detection.

[0046] S5. Fault detection: The Hotelling statistic is used to comprehensively determine whether the three characteristics deviate from the normal range at the same time. If the three characteristics deviate from the normal range at the same time, there is a fault; otherwise, there is no fault.

[0047] To comprehensively determine whether all three characteristics deviate from the normal range simultaneously, this embodiment uses Hotelling's index, a classic indicator in multivariate statistical process control (MSPC). This statistic, based on the assumption of a multivariate normal distribution, effectively detects coordinated anomalous changes among multiple variables by calculating the Mahalanobis distance between sample points and the center of the normal range, thus providing a comprehensive and statistically rigorous method.

[0048] definition:

[0049] Interpretation of statistical significance: Essentially, it's the square of the Mahalanobis distance, measuring the current feature vector. Relative to the normal distribution center The standardized deviation. Unlike Euclidean distance, Mahalanobis distance uses the covariance matrix to measure the degree of deviation. By scaling each dimension and considering the correlation between variables, multivariate joint anomalies can be detected more accurately.

[0050] Control Limit (UCL) setting:

[0051] in, It has degrees of freedom. The 99th percentile (i.e., significance level) of the F-distribution This formula originates from the multivariate normal distribution. The exact distribution theory of statistics.

[0052] like If the system is deemed to be abnormal with a 99% confidence level, then a bioattachment failure has occurred.

[0053] For the second embodiment, please refer to... Figure 2 A method for diagnosing bio-attachment faults in unmanned underwater vehicles based on parameter identification, building upon the first embodiment described above, further includes S6, fault type identification, which first utilizes a contribution map to quantify each feature pair. The contribution level exceeding the limit is then used to determine the fault type using a mapping rule table; the mapping rule table includes the maximum contribution item, its corresponding physical meaning, and the attachment type determination.

[0054] Once a fault is detected, it is necessary to further determine the attachment type, that is, to identify the specific category and attributes of the fault in order to carry out targeted treatment. This invention employs the Contribution Plot method, which uses statistical analysis techniques to quantify the impact of each feature variable on... The contribution of out-of-limit phenomena can be used to identify key influencing factors and assist in fault diagnosis and decision-making.

[0055] Contribution calculation formula:

[0056] in Represents the vector of the first One element, yes The Each component.

[0057] Table 1 shows the mapping rules.

[0058] This method provides interpretable fault location capabilities, enabling maintenance personnel to quickly determine the attachment location.

[0059] Third embodiment, please refer to Figure 3 A method for diagnosing bio-attachment faults of unmanned underwater vehicles based on parameter identification, which, in addition to the second embodiment described above, also includes S7, triggering a graded early warning mechanism according to the severity of the fault.

[0060] For example, a three-level early warning output triggers a tiered early warning mechanism based on the severity of the fault: This includes the warning level, triggering conditions, engineering implications, and recommendations; Level 1, Triggering Conditions: Only Contribution1 is abnormal. Engineering meaning and suggestions: Slight and uniform adhesion, battery life begins to decrease. It is recommended to strengthen monitoring. The task can continue.

[0061] Level 2, triggering condition: Non-uniform characteristics may appear (Contribution2 or Contribution3 is the largest). Engineering implications and recommendations: moderate adhesion, difficult track control, significantly increased energy consumption. It is recommended to adjust the mission plan or prepare for recovery.

[0062] Level 3, triggering condition: Or any rudder efficiency coefficient decreases by more than 30% (e.g. (Engineering meaning and recommendations: Severe adhesion, equipment is on the verge of going out of control, and it is necessary to return to base immediately or activate emergency procedures.)

[0063] The early warning information is transmitted back to the mother ship via underwater acoustic communication for use in mission scheduling and decision-making.

[0064] Fourth embodiment: A bio-attachment fault diagnosis system for unmanned underwater vehicles based on parameter identification, used to implement the method of the first embodiment described above, comprising: The data acquisition module collects the following time-series data generated by the unmanned underwater vehicle in real time during operation: buoyancy adjustment, longitudinal speed, pitch angle, roll angle, elevator angle, and rudder angle; The behavioral agent model building module includes: A buoyancy-velocity model reflecting drag performance is constructed using buoyancy adjustment and longitudinal velocity. A dynamic pitch angle model reflecting longitudinal control performance is constructed using pitch angle and elevator angle. A dynamic roll angle model reflecting lateral control performance is constructed using roll angle and rudder angle. The estimation module uses recursive least squares with a forgetting factor to identify the parameters of the behavioral surrogate model in real time, and obtains the drag coefficient surrogate, pitch surface response coefficient and yaw surface response coefficient. Constructing feature vectors and establishing a baseline module includes: Acquire multiple sampling points before the initial stage of the task; Normalize the drag coefficient proxy, pitch surface response coefficient, and yaw surface response coefficient to obtain normalized eigenvectors. The mean vector and covariance matrix of statistically normalized eigenvectors; The fault detection module uses the Hotelling statistic to comprehensively determine whether three features deviate from the normal range simultaneously. If all three features deviate from the normal range simultaneously, there is a fault; otherwise, there is no fault.

[0065] The fifth embodiment is a bio-attachment fault diagnosis system for unmanned underwater vehicles based on parameter identification, used to implement the method of the second embodiment described above. Building upon the fourth embodiment, it further includes a fault type identification module, which first quantifies each feature pair using a contribution map. The contribution level exceeding the limit is then used to determine the fault type using a mapping rule table; the mapping rule table includes the maximum contribution item, its corresponding physical meaning, and the attachment type determination.

[0066] The sixth embodiment is a bio-attachment fault diagnosis system for unmanned underwater vehicles based on parameter identification, which implements the method of the third embodiment above. Based on the fifth embodiment above, it also includes a graded early warning module, which triggers a graded early warning mechanism according to the severity of the fault.

[0067] The following example of an unmanned underwater vehicle illustrates the implementation process of this invention: Step 1: After the underwater glider is launched, record the buoyancy pump's operating status every second (converted to...). ), GPS / pressure gauge fusion depth (derived by differentiation) IMU output Servo commands Kalman filtering is applied to the collected data to reduce noise interference.

[0068] Step 2: The first 2 hours ( Points (based on a 1Hz sampling frequency) are used to calculate and It has been verified that... (10 minutes) is the minimum reliable window. It can provide a more stable statistical benchmark.

[0069] Step 3: Starting from the 3rd hour, perform an FFRLS update every 10 seconds (sliding window length ≈ 300 points, approximately 5 minutes), and output... .

[0070] Step 4: Real-time Calculation If the limit is exceeded three times consecutively, a Level 1 warning will be triggered; if simultaneously... If so, it will be upgraded to Level 2 and reported as "suspected right-side attachment".

[0071] Step 5: The early warning information is transmitted back to the mother ship via underwater acoustic communication for mission scheduling and decision-making.

[0072] The seventh embodiment is a computer program product, including a computer program that is executed by a processor as described above in the parameter identification-based bio-attachment fault diagnosis method for unmanned underwater vehicles.

[0073] Eighth embodiment, a computer-readable storage medium including instructions, which, when executed on a computer, cause the computer to perform the above-described parameter identification-based bio-attachment fault diagnosis method for unmanned underwater vehicles.

[0074] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented, in whole or in part, as a computer program product, the computer program product includes one or more computer instructions. When the computer program instructions are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0075] The above description is only a preferred embodiment of the present invention. It should be noted that any improvements, modifications, substitutions or variations made by those skilled in the art without departing from the principle of the present invention should be considered as being included within the protection scope of the present invention.

Claims

1. A method for diagnosing bio-attachment faults in unmanned underwater vehicles based on parameter identification, characterized in that, include: S1. Real-time acquisition of the following time-series data generated by the unmanned underwater vehicle during operation: buoyancy adjustment, longitudinal speed, pitch angle, roll angle, elevator angle, and rudder angle; S2. Construct a behavioral agent model, including: A buoyancy-velocity model reflecting drag performance is constructed using buoyancy adjustment and longitudinal velocity. A dynamic pitch angle model reflecting longitudinal control performance is constructed using pitch angle and elevator angle. A dynamic roll angle model reflecting lateral control performance is constructed using roll angle and rudder angle. S3. Real-time identification of behavioral proxy model parameters to obtain drag coefficient proxy, pitch control surface response coefficient and yaw control surface response coefficient. S4. Construct feature vectors and establish a baseline, including: Acquire data from multiple sampling points before the initial stage of the task; Normalize the drag coefficient proxy, pitch surface response coefficient, and yaw surface response coefficient to obtain normalized eigenvectors. The mean vector and covariance matrix of statistically normalized eigenvectors; S5. Fault detection: The Hotelling statistic is used to comprehensively determine whether the three characteristics deviate from the normal range at the same time. If the three characteristics deviate from the normal range at the same time, there is a fault; otherwise, there is no fault.

2. The method for diagnosing bio-attachment faults of unmanned underwater vehicles based on parameter identification according to claim 1, characterized in that, S1 includes: The data is filtered.

3. The method for diagnosing bio-attachment faults of unmanned underwater vehicles based on parameter identification according to claim 1 or 2, characterized in that, In S2: The mathematical expression for the buoyancy-velocity model is: in: This is a constant term, reflecting constant environmental disturbances; This is the gain coefficient of buoyancy on velocity, characterizing propulsion efficiency; is the velocity autocorrelation coefficient, with a value range of (0,1); The model residuals include unmodeled dynamics and random perturbations. The mathematical expression for the pitch angle dynamic model is: in: This is the first constant bias term; The pitch angle autocorrelation coefficient reflects attitude inertia; The elevator response coefficient represents the pitch angle change caused by a unit rudder deflection; For pitch channel model residuals; The mathematical expression for the roll angle dynamic model is: in: This is the second constant bias term; This is the autocorrelation coefficient of the roll angle; This refers to the rudder response coefficient; This represents the residual of the roll channel model.

4. The method for diagnosing bio-attachment faults of unmanned underwater vehicles based on parameter identification according to claim 3, characterized in that, S3 includes: using recursive least squares with a forgetting factor to identify behavioral agent model parameters in real time; Recursive least squares method with forgetting factor introduces forgetting factor More recent data is given higher weight, thus enabling online parameter updates; the following key parameters are output: Resistance coefficient agent quantity: ; Pitch control surface response coefficient: ; Yaw control surface response coefficient: .

5. The method for diagnosing bio-attachment faults of unmanned underwater vehicles based on parameter identification according to claim 4, characterized in that, S4 includes: S4. Construct feature vectors and establish a baseline, including: Obtain the first N sampling points in the initial stage of the task. The initial running time of the task shall not be less than 10 minutes; Definition of normalized eigenvectors: in: This represents the sample mean of the resistance surcharge during the normal phase. The pitch control effect is the sample mean. This represents the sample mean of the roll rudder effect. Statistical baseline calculation: The formula for calculating the mean vector is: , is a 3-dimensional column vector; The formula for calculating the covariance matrix is: , is a 3×3 positive definite symmetric matrix.

6. The method for diagnosing bio-attachment faults of unmanned underwater vehicles based on parameter identification according to claim 5, characterized in that, It also includes S6, fault type identification, which first uses a contribution map to quantify each feature pair. The contribution level exceeding the limit is then used to determine the fault type using a mapping rule table; the mapping rule table includes the maximum contribution item, its corresponding physical meaning, and the attachment type determination.

7. The method for diagnosing bio-attachment faults of unmanned underwater vehicles based on parameter identification according to claim 6, characterized in that, It also includes S7, which triggers a graded early warning mechanism based on the severity of the fault.

8. A bio-attachment fault diagnosis system for unmanned underwater vehicles based on parameter identification, characterized in that, include: The data acquisition module collects the following time-series data generated by the unmanned underwater vehicle in real time during operation: buoyancy adjustment, longitudinal speed, pitch angle, roll angle, elevator angle, and rudder angle; The behavioral agent model building module includes: A buoyancy-velocity model reflecting drag performance is constructed using buoyancy adjustment and longitudinal velocity. A dynamic pitch angle model reflecting longitudinal control performance is constructed using pitch angle and elevator angle. A dynamic roll angle model reflecting lateral control performance is constructed using roll angle and rudder angle. The estimation module uses recursive least squares with a forgetting factor to identify the parameters of the behavioral surrogate model in real time, and obtains the drag coefficient surrogate, pitch surface response coefficient and yaw surface response coefficient. Constructing feature vectors and establishing a baseline module includes: Acquire data from multiple sampling points before the initial stage of the task; Normalize the drag coefficient proxy, pitch surface response coefficient, and yaw surface response coefficient to obtain normalized eigenvectors. The mean vector and covariance matrix of statistically normalized eigenvectors; The fault detection module uses the Hotelling statistic to comprehensively determine whether three features deviate from the normal range simultaneously. If all three features deviate from the normal range simultaneously, there is a fault; otherwise, there is no fault.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it is the bio-attachment fault diagnosis method for unmanned underwater vehicles based on parameter identification as described in any one of claims 1-7.

10. A computer-readable storage medium comprising instructions, characterized in that, When it is run on a computer, it causes the computer to perform the parameter identification-based bio-attachment fault diagnosis method for unmanned underwater vehicles as described in any one of claims 1-7.

Images