A Classification Method for Failure Modes of Underwater Thrusters

Abstract

This invention discloses a method for classifying fault modes of underwater thrusters, comprising: S1, acquiring underwater thruster motor parameters; S2, establishing mechanical motion equations using the underwater thruster motor; S3, constructing a sliding mode observer based on the mechanical motion equations to obtain approximate motor torque load estimates; S4, constructing a least squares matrix based on the online estimated load torque values ​​to obtain torque parameter estimates; and S5, inputting the obtained torque parameter estimates into a support vector machine model to obtain the fault mode category. According to this invention, the problem of fault detection and localization during real-time operation of the thruster is solved, thereby obtaining accurate and reliable fault information and providing necessary decision-making basis for intelligent decision-making in underwater robots.

Description

Technical Field

[0001] This invention relates to the technical field of underwater thrusters, and in particular to a method for classifying failure modes of underwater thrusters. Background Technology

[0002] Autonomous underwater vehicles (AUVs) navigate unmanned and untethered in deep-sea environments. As the primary power component of an AUV, the underwater thruster is crucial; a malfunction can lead to thrust derating or even complete power loss, and in extreme cases, the AUV may be lost altogether. Fault diagnosis can monitor the thruster's operational status in a timely manner, allowing for necessary remedial measures to prevent the malfunction from escalating and causing greater losses. However, current underwater thruster fault diagnosis methods typically start with the AUV system's status signals, mostly only able to determine whether a thruster malfunction has occurred, without identifying the specific type of malfunction. Therefore, they fail to provide sufficiently effective information and basis for intelligent decision-making in the event of a thruster failure. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a fault mode classification method for underwater thrusters, solving the problem of fault detection and localization during real-time operation of thrusters, thereby obtaining accurate and reliable fault information and providing necessary decision-making basis for intelligent decision-making in underwater robots. To achieve the above-mentioned objectives and other advantages of the present invention, a fault mode classification method for underwater thrusters is provided, comprising:

[0004] S1. Obtain the parameters of the underwater thruster motor;

[0005] S2. Establish the mechanical motion equations using the underwater thruster motor: Among them, J and w m T L T f These represent the motor's moment of inertia, mechanical angular velocity, load torque, and static torque, respectively; F is the damping coefficient of the motor drive; and p... n Let i be the number of pole pairs of the motor. q For stator current, ψ f These are permanent magnet rotor flux linkages;

[0006] S3. Construct a sliding mode observer based on the mechanical motion equations: in, For w m The estimated value is given by h, which is the sliding mode gain coefficient, and thus an approximate estimate of the motor torque load is obtained.

[0007] S4. Based on the online estimated load torque value, construct the least squares matrix to obtain the torque parameter estimate;

[0008] S5. Input the obtained torque parameter estimates into the support vector machine model to obtain the fault mode categories.

[0009] Preferably, the motor torque load estimate of the sliding mode observer includes the following steps:

[0010] S31. Establish the mechanical motion equations of the DC brushless motor used in the thruster: Among them, J and w m T L T f These represent the motor's moment of inertia, mechanical angular velocity, load torque, and static torque, respectively; F is the damping coefficient of the motor drive; and p... n Let i be the number of pole pairs of the motor. q For stator current, ψ f These are permanent magnet rotor flux linkages;

[0011] S32. Construct a sliding mode observer based on the aforementioned mechanical motion equations: in, For w m The estimated value, h is the sliding mode gain coefficient;

[0012] S33. The sliding mode gain coefficient h, if greater than the rated load torque of the motor, can ensure that the angular velocity estimation error converges to zero within a finite time. At this point, there exists... By replacing the sign function sgn(.) with the saturation function sat(.), an approximately equivalent load estimate can be obtained.

[0013] Preferably, the online estimation of torque parameters using the least squares method includes the following steps:

[0014] S41. Based on the working characteristics of the propeller, establish an approximate load torque model T. L =kw m |w m |+Bw m , where k and B are the torque model parameters to be identified;

[0015] S42. Construct a least-squares matrix based on the online estimated load torque value: Among them, A and B are both extracted using the time window rolling method, which includes the current moment and the n sets of data used and estimated in the previous period;

[0016] S43. Based on the aforementioned matrices A and B, estimate the torque model parameters using the least squares method: Where pinv(A) is the pseudo-inverse of matrix A.

[0017] Preferably, the fault types include five types: normal, jammed, entangled, blade broken, and blade detached. The torque model parameters are obtained by sliding mode observer and least squares method, and then a pairwise fault multi-classification model is established by support vector machine.

[0018] Compared with the prior art, the beneficial effects of this invention are:

[0019] (1) The underwater propulsion fault mode classification method of the present invention only uses its own control, current, speed and other signals to classify five operating states: normal, jammed, entangled, blade broken and blade lost.

[0020] (2) The sliding mode observer, least squares method and support vector machine used in the underwater thruster fault mode classification method of the present invention are all mature algorithms, and the computational reliability and real-time performance can be guaranteed. Attached Figure Description

[0021] Figure 1 A flowchart illustrating the operation of the underwater thruster failure mode classification method according to the present invention;

[0022] Figure 2 A diagram showing the torque estimation results of the underwater thruster fault mode classification method according to the present invention.

[0023] Figure 3 A graph showing the torque parameter estimation results of the underwater thruster fault mode classification method according to the present invention;

[0024] Figure 4 A graph showing the torque parameter estimation results of the underwater thruster fault mode classification method according to the present invention;

[0025] Figure 5 This is a diagram showing the failure mode classification results of the underwater thruster failure mode classification method according to the present invention. Detailed Implementation

[0026] 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, and 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.

[0027] Reference Figure 1-5 A method for classifying underwater thruster failure modes, comprising: S1, acquiring underwater thruster motor parameters;

[0028] S2. Establish the mechanical motion equations using the underwater thruster motor: Among them, H and w mT L T f These represent the motor's moment of inertia, mechanical angular velocity, load torque, and static torque, respectively; F is the damping coefficient of the motor drive; and p... n Let i be the number of pole pairs of the motor. q For stator current, ψ f These are permanent magnet rotor flux linkages;

[0029] S3. Construct a sliding mode observer based on the mechanical motion equations: in, For w m The estimated value is given by h, which is the sliding mode gain coefficient, and thus an approximate estimate of the motor torque load is obtained.

[0030] S4. Based on the online estimated load torque value, construct the least squares matrix to obtain the torque parameter estimate;

[0031] S5. Input the obtained torque parameter estimates into the support vector machine model to obtain the fault mode categories.

[0032] like Figure 1 As shown, the motor torque load estimate of the sliding mode observer includes the following steps:

[0033] S31. Establish the mechanical motion equations of the DC brushless motor used in the thruster: Among them, H and w m T L T f These represent the motor's moment of inertia, mechanical angular velocity, load torque, and static torque, respectively; F is the damping coefficient of the motor drive; and p... n Let i be the number of pole pairs of the motor. q For stator current, ψ f These are permanent magnet rotor flux linkages;

[0034] S32. Construct a sliding mode observer based on the aforementioned mechanical motion equations: in, For w m The estimated value, h is the sliding mode gain coefficient;

[0035] S33. The sliding mode gain coefficient h, if greater than the rated load torque of the motor, can ensure that the angular velocity estimation error converges to zero within a finite time. At this point, there exists... By replacing the sign function sgn(.) with the saturation function sat(.), an approximately equivalent load estimate can be obtained.

[0036] like Figure 2As shown, the thruster operates for a total of 100 seconds, with the first 50 seconds in normal condition and the last 50 seconds in fault condition. At the moment the thruster transitions from normal to fault condition, there is a brief abrupt change in the load torque estimation error. The thruster's maximum output torque is approximately 2.1 Nm. During this state, the thruster torque estimation error is slightly larger, while at other times the thruster torque error is smaller.

[0037] like Figure 1 As shown, the online estimation of torque parameters using the least squares method includes the following steps:

[0038] S41. Based on the working characteristics of the propeller, establish an approximate load torque model T. L =kw m |w m |+Bw m , where k and B are the torque model parameters to be identified;

[0039] S42. Construct a least-squares matrix based on the online estimated load torque value: Among them, A and B are both extracted using the time window rolling method, which includes the current moment and the n sets of data used and estimated in the previous period;

[0040] S43. Based on the aforementioned matrices A and B, estimate the torque model parameters using the least squares method: Where pinv(A) is the pseudo-inverse of matrix A.

[0041] like Figure 3 , Figure 4 As shown, the overall error between the identified torque parameters and the actual parameters is small. The error fluctuations mainly occur at the fault switching moment at 50s and at 12s and 62s when the data just fills matrix A. In addition, the thruster will also have a large error fluctuation when the thruster speed fluctuates greatly at 73s.

[0042] Furthermore, the fault types include five types: normal, jammed, entangled, blade broken, and blade detached. The torque model parameters are obtained through sliding mode observer and least squares method, and then a pairwise fault multi-classification model is established using support vector machine.

[0043] like Figure 4 As shown, the support vector machine can correctly predict the thruster's state 0.25 seconds after startup under normal thruster conditions. After a thruster malfunction, the support vector machine requires 1-5 seconds to correctly predict the fault. The prediction accuracy for thruster propeller entanglement faults is 98% (392 / 400); the prediction accuracy for thruster propeller jamming faults is 99% (396 / 400); the prediction accuracy for thruster propeller breakage faults is 95% (380 / 400); and the prediction accuracy for thruster propeller detachment faults is 97% (388 / 400).

[0044] The number of devices and processing scale described herein are for the purpose of simplifying the description of the invention, and applications, modifications and variations thereof will be apparent to those skilled in the art.

[0045] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A method for classifying failure modes of underwater thrusters, characterized in that, Includes the following steps: S1. Obtain the parameters of the underwater thruster motor; S2. Establish the mechanical motion equations using the underwater thruster motor: Among them, J and w m T L T f These represent the motor's moment of inertia, mechanical angular velocity, load torque, and static torque, respectively; F is the damping coefficient of the motor drive; and p... n Let i be the number of pole pairs of the motor. q For stator current, ψ f These are permanent magnet rotor flux linkages; S3. Construct a sliding mode observer based on the mechanical motion equations: in, For w m The estimated value is given by h, which is the sliding mode gain coefficient, and thus an approximate estimate of the motor torque load is obtained. S4. Based on the online estimated load torque value, construct the least squares matrix to obtain the torque parameter estimate; S5. Input the obtained torque parameter estimates into the support vector machine model to obtain the fault mode categories.

2. The underwater thruster failure mode classification method as described in claim 1, characterized in that, The motor torque load estimate of the sliding mode observer includes the following steps: S31. Establish the mechanical motion equations of the DC brushless motor used in the thruster: Among them, J and w m T L T f These represent the motor's moment of inertia, mechanical angular velocity, load torque, and static torque, respectively; F is the damping coefficient of the motor drive; and p... n Let i be the number of pole pairs of the motor. q For stator current, ψ f These are permanent magnet rotor flux linkages; S32. Construct a sliding mode observer based on the aforementioned mechanical motion equations: in, For w m The estimated value, h is the sliding mode gain coefficient; S33. The sliding mode gain coefficient h, if greater than the rated load torque of the motor, can ensure that the angular velocity estimation error converges to zero within a finite time. At this point, there exists... By replacing the sign function sgn(.) with the saturation function sat(.), an approximately equivalent load estimate can be obtained.

3. The underwater thruster failure mode classification method as described in claim 2, characterized in that, Online estimation of torque parameters using the least squares method includes the following steps: S41. Based on the working characteristics of the propeller, establish an approximate load torque model T. L =kw m |w m |+Bw m , where k and B are the torque model parameters to be identified; S42. Construct a least-squares matrix based on the online estimated load torque value: Among them, A and B are both extracted using the time window rolling method, which includes the current moment and the n sets of data used and estimated in the previous period; S43. Based on the aforementioned matrices A and B, estimate the torque model parameters using the least squares method: Where pinv(A) is the pseudo-inverse of matrix A.

4. The underwater thruster failure mode classification method as described in claim 3, characterized in that, The fault types include five types: normal, jammed, entangled, blade broken, and blade detached. The torque model parameters are obtained by sliding mode observer and least squares method, and then a pairwise fault multi-classification model is established by support vector machine.

Images