KR20240028008A - Method and apparatus for diagnosing failure of underwater vehicle thruster based on deep learning - Google Patents

Abstract

In these embodiments, a failure diagnosis model for an underwater propulsion device is applied to an underwater vehicle to diagnose a state in which sudden failures such as propulsion blade breakage and floating matter entanglement occur based on measurement data, and the results of segmenting the blade damage grade are used to diagnose the condition. In order to dynamically determine mission performance and prevent data imbalance due to a lack of failure data required for the failure diagnosis model for underwater propulsion, failure diagnosis is made by utilizing expanded failure data regarding thruster blade breakage and entanglement of floating objects through a data expansion model. Provides a failure diagnosis method and device for underwater propulsion that improves the prediction accuracy of the model.

Description

Deep learning-based underwater vehicle thruster failure diagnosis method and device {METHOD AND APPARATUS FOR DIAGNOSING FAILURE OF UNDERWATER VEHICLE THRUSTER BASED ON DEEP LEARNING}

The technical field to which the present invention belongs relates to a method and device for diagnosing a failure of an underwater vehicle propulsion machine based on deep learning.

The content described in this section simply provides background information for this embodiment and does not constitute prior art.

Due to the operational nature of underwater vehicles, satellite navigation such as GPS (Global Positioning System) cannot be used during mission performance, and real-time communication is limited. As a result, if a malfunction occurs in the underwater propulsion unit, which is the propulsion unit of the underwater vehicle, not only is the mission performance of the underwater vehicle interrupted, but there is a problem in which the status of the underwater vehicle cannot be confirmed.

Methods for diagnosing failures in underwater propulsion systems can be broadly divided into rule-based methods, model-based methods, and data-driven methods.

The rule-based method has a simple structure, but it is difficult to respond flexibly. Model-based methods require knowledge of specialized dynamic process models. The data-based method extracts and utilizes the features contained in the actual measured data, so it does not require a complex dynamic process model and can actively respond to unexpected errors.

Deep learning, one of the data-based methods, basically requires a large amount of learning data and requires high-quality data to extract embedded features.

Because underwater vehicles are operated for long periods of time in normal conditions, the amount of normal data is very large, but when a failure occurs, there is relatively insufficient failure data that can be acquired, resulting in data imbalance. If data with data imbalance is used as learning data, it affects the prediction accuracy of the deep learning model. However, if a propeller in a faulty state is operated for a long time to acquire failure data, there is a problem that affects the lifespan of the underwater thruster.

Korea Patent Publication No. 10-2017-0109770 (2017.10.10)

Embodiments of the present invention apply a failure diagnosis model related to an underwater propeller to an underwater vehicle to diagnose a state in which hard type failures such as damage to the propeller blade and entanglement of floating objects occur based on measurement data. In order to prevent the data imbalance phenomenon where failure data for learning is lacking in the failure diagnosis model for underwater propulsion, the accuracy of the failure diagnosis model is improved by utilizing expanded failure data regarding thruster blade breakage and floating object entanglement through a data expansion model, and improved The main purpose is to ensure sustainable performance of underwater vehicle missions by providing a process for recognizing failure situations through a failure diagnosis model and then responding to them.

Embodiments of the present invention distinguish and diagnose floating matter entanglement and blade damage through a failure diagnosis model. If floating objects become entangled in the propeller of the propulsion machine, significant current consumption and load are applied, which can cause the underwater vehicle to float and take action after recovery. Blade breakage determines whether or not the mission will be performed depending on the degree of damage. Therefore, by analyzing the state data, the damage threshold is determined according to the degree of damage, and a failure diagnosis model is developed to enable status diagnosis based on the damage grade compared to the damage standard. The purpose is to apply it.

Other unspecified objects of the present invention can be additionally considered within the scope that can be easily inferred from the following detailed description and its effects.

According to one aspect of the present embodiment, a method for diagnosing a failure of an underwater propulsion device includes: moving an underwater vehicle according to an underwater navigation algorithm; Collecting measurement data of the underwater propulsion device mounted on the underwater vehicle using a measurement sensor; Diagnosing the state of the underwater thruster by inputting the measurement data based on a pre-learned failure diagnosis model and outputting a normal state or a failure state; and controlling the operation of the underwater moving object according to the state of the underwater thruster.

In the step of outputting the failure state, the failure state may be output by dividing the failure state into a tangled floating state or a blade damage state.

In the step of controlling the operation of the underwater propeller, if the state of the underwater propeller is in the state of entanglement of the floating object, the movement of the underwater moving object is stopped and rises to the sea surface, and if the state of the underwater propeller is in the blade damage state, Performance can be subdivided.

The failure diagnosis model receives and learns floating object entanglement failure data and outputs the floating object entanglement state, and the failure diagnosis model receives blade breakage failure data and learns to output the blade damage state, and compares it with a set damage standard value. Damage grades related to blade damage can be classified and output.

The failure diagnosis model may output the floating matter entanglement state or the blade damage state by applying a Recurrent Neural Network (RNN) or Long Short Term Memory (LSTM) that analyzes time series data.

The failure diagnosis model sets the number of data between normal data and failure data regarding the underwater propulsion machine as a first ratio, and sets the number of data between the floating object entanglement failure data and the blade breakage failure data as a second ratio, The ratio of data input to the fault diagnosis model can be adjusted by applying a first weight to the first ratio and a second weight to the second ratio.

The failure diagnosis model expands the floating object entanglement failure data using a data expansion model based on a GAN (Generative Adversarial Network) and is learned by receiving the expanded floating object entanglement failure data, and the failure diagnosis model uses the data expansion model. Thus, the blade breakage failure data can be expanded and the expanded blade breakage failure data can be input and learned.

The data expansion model is learned by connecting a generator and a discriminator, the generator includes a first generator and a second generator, and the discriminator may include a first discriminator and a second discriminator. You can.

The first generator generates floating object entanglement failure data, the second generator generates blade breakage failure data, the first discriminator determines similarity of the generated floating object entanglement failure data, and the second discriminator determines the similarity of the generated floating object entanglement failure data. The similarity of the generated blade breakage failure data can be determined.

The data expansion model is learned such that (i) a first loss function for the first generator and the first discriminator and (ii) a second loss function for the second generator and the second discriminator are minimized. You can.

According to another aspect of the present embodiment, a fault diagnosis device includes: an underwater thruster implemented to move an underwater mobile object; a measurement sensor attached to the underwater thruster to collect measurement data; And a fault diagnosis model that inputs the measurement data to diagnose the state of the underwater thruster and outputs a normal state or a failure state, and includes a control unit that controls the operation of the underwater mobile device according to the state of the underwater thruster. Provides a device.

The measurement sensor may measure current consumption through a Hall sensor attached to the power supply line of the underwater propulsion machine, and transmit the measured power consumption to the fault diagnosis model.

The measurement sensor may measure thruster vibration data through an acceleration sensor or vibration sensor attached to the body or power transmission shaft of the underwater thruster, and transmit the measured thruster vibration data to the failure diagnosis model.

The measurement sensor may measure motor rotation data through an encoder sensor attached to the motor of the underwater propulsion device, and transmit the measured motor rotation data to the failure diagnosis model.

As described above, according to embodiments of the present invention, a failure diagnosis model for an underwater propulsion device can be applied to an underwater vehicle to diagnose a state in which sudden failures such as damage to the propulsion blade and entanglement of floating objects occur based on measurement data. There is an effect.

According to embodiments of the present invention, in order to prevent a data imbalance phenomenon in which the failure data required for the failure diagnosis model for the underwater propulsion is insufficient, the extended failure data regarding the damage of the propeller blade and entanglement of floating objects through the data expansion model is used to detect the failure. It has the effect of improving the accuracy of the diagnostic model.

According to embodiments of the present invention, there is an effect of making underwater vehicle missions sustainable by providing a process for recognizing a situation in which a failure has occurred through an improved failure diagnosis model and then responding to it.

Even if the effects are not explicitly mentioned here, the effects described in the following specification and their potential effects expected by the technical features of the present invention are treated as if described in the specification of the present invention.

Figures 1A and 1B are diagrams illustrating a situation in which a failure occurs in an existing underwater propulsion device.
Figures 2a and 2b are diagrams illustrating a method for diagnosing a failure of an underwater propulsion device according to an embodiment of the present invention.
Figure 3 is a diagram illustrating the operation of the failure diagnosis method of an underwater propulsion device according to an embodiment of the present invention to determine the failure type through a failure diagnosis model.
Figure 4 is a diagram illustrating the operation of the failure diagnosis method of an underwater propulsion device according to an embodiment of the present invention to generate failure data through a data expansion model.
Figure 5 is a diagram illustrating a failure diagnosis device for an underwater propulsion engine according to another embodiment of the present invention.

Hereinafter, in describing the present invention, if it is determined that related known functions may unnecessarily obscure the gist of the present invention as they are obvious to those skilled in the art, the detailed description will be omitted, and some embodiments of the present invention will be described. It will be described in detail through exemplary drawings.

Failures that occur in underwater propulsion engines are largely divided into two types: hard type failure and soft type failure. A hard type failure is defined as a situation where a sudden change in state occurs in a short period of time during normal operation, while a soft type failure refers to a situation that falls outside the scope of the normal state due to reasons such as lifespan during long-term operation.

Examples of hard-type failures include situations such as broken thruster blades and entanglement in ocean floats, while examples of soft-type failures include phenomena such as aging.

This embodiment provides a method and apparatus for diagnosing thruster blade failure and marine flotation entanglement as hard type failures.

Figures 1A and 1B are diagrams illustrating a situation in which a failure occurs in an existing underwater propulsion device.

The underwater vehicle starts operating, and the underwater vehicle makes a movement plan and moves according to the underwater navigation algorithm (S110). A malfunction may occur in the propeller of an underwater vehicle during mission performance (S120). Previously, even if a propeller malfunction occurred during the performance of an underwater vehicle's mission, the status of the underwater vehicle could not be confirmed and recovery of the underwater vehicle was not easy (S130).

The data-based fault diagnosis method applied to solve the problem of difficulty in checking the status of underwater propulsion requires a large amount of data, so it requires long-term platform operation and experience and a data accumulation system. Since the underwater propulsion is used as a platform propulsion, the risk of platform loss must be accepted in order to obtain failure data.

In this embodiment, in order to solve the problem of lack of failure data, data acquired in a short period of time can be expanded, and the expanded data can be used as learning data to derive failure diagnosis results through a deep learning model with high prediction accuracy.

Figures 2a and 2b are diagrams illustrating a method for diagnosing a failure of an underwater propulsion device according to an embodiment of the present invention.

In step S210, the method of diagnosing a failure of an underwater propulsion device may include moving an underwater vehicle according to an underwater navigation algorithm. The underwater navigation algorithm uses an external environment detection sensor attached to an underwater vehicle to measure depth, speed, and direction and estimate the route based on the measured information.

In step S220, measurement data of the underwater propulsion device mounted on the underwater vehicle may be collected using a measurement sensor. Measurement sensors measure the status of the underwater propulsion machine.

In step S230, the state of the underwater thruster can be diagnosed by inputting measurement data based on a pre-learned failure diagnosis model, and a normal state or failure state can be output.

In steps S240, S250, and S260, steps of controlling the operation of the underwater mobile device can be performed depending on the state of the underwater propulsion device. In the step of controlling the operation of the underwater vehicle, the movement of the underwater vehicle is performed when the state of the underwater propulsion device is normal. If the underwater propulsion device is in a malfunction state, the underwater vehicle stops moving and rises to the sea surface (S240). The underwater moving body is recovered by recovering the underwater moving body (S250). The thrusters can be repaired using repair equipment in an underwater vehicle recovery vessel. After completing repairs, the underwater vehicle performs its designated mission again (S260).

The surfaced underwater vehicle transmits a signal to the recovery ship through a satellite navigation device such as GPS, and the recovery ship that receives the signal recovers the underwater vehicle floating in the relevant sea area, corrects the malfunction, and then completes the mission again. .

The status is diagnosed based on the status data generated from the underwater propulsion by the propeller failure diagnosis model, and when a failure occurs, the failure diagnosis model recognizes it and enables follow-up action.

Figure 3 is a diagram illustrating the operation of the failure diagnosis method of an underwater propulsion device according to an embodiment of the present invention to determine the failure type through a failure diagnosis model.

The learning data of the fault diagnosis model utilizes state data sets such as vibration, current, and voltage acquired while using existing thrusters. The failure data set includes not only measured data but also an expanded data set generated using a data expansion model based on the GAN model (S310).

Using a deep learning-based failure diagnosis model, the measured failure data and extended failure data are applied to a time series learning model, RNN (Recurrent Neural Network) or LSTM (Long Short Term Memory) (S320).

Real-time status data for determining the status of the underwater propulsion device is arranged based on the Nyquist-Shannon Sampling Theorem (S330). An array is created by combining data sets in certain units. The arranged data set is used as prediction data for the learned fault diagnosis model.

The failure diagnosis model determines whether the underwater thruster is in a normal state or a failure state through the prediction result (S350). If a failure is determined to be a failure, the failure diagnosis model diagnoses whether it corresponds to tangled floating matter or blade damage. Depending on the type and state of the failure, the underwater vehicle's ability to perform its mission is determined, and if it is unable to perform its mission, it temporarily stops and rises to sea level.

The failure state is classified into a floating object entanglement state or a blade damage state. The failure diagnosis model receives and learns the floating object entanglement failure data and outputs the floating object entanglement state. The failure diagnosis model receives the blade breakage failure data and learns the blade damage state. can be output.

The failure diagnosis model can output the state of floating matter entanglement or blade damage by applying RNN (Recurrent Neural Network) or LSTM (Long Short Term Memory) that analyzes time series data.

The failure diagnosis model may set the number of data between normal data and failure data regarding the underwater propulsion machine as a first ratio. For example, the first ratio can be set to 2:1 to 10:1, etc. and can be changed according to required design details.

The failure diagnosis model may set the number of data between the floating object entanglement failure data and the blade breakage failure data as a second ratio. For example, the second ratio can be set to 1:0.3 ~ 1:0.7, etc. and can be changed according to required design details.

The failure diagnosis model may adjust the ratio of data input to the failure diagnosis model by applying a first weight to the first ratio and a second weight to the second ratio. The sum of the first weight and the second weight may be set to 1.

In this embodiment, failure data measured over a short period of time is expanded using a GAN (Generative Adversarial Network) model, one of the unsupervised deep learning techniques.

The failure diagnosis model may set the number of data between the measured failure data and the extended generated failure data as a third ratio. For example, the third ratio can be set to 1:0.3 ~ 1:0.7, etc. and can be changed according to required design details.

The failure diagnosis model may adjust the ratio of data input to the failure diagnosis model by applying a first weight to the first ratio, a second weight to the second ratio, and a third weight to the third ratio. The sum of the first weight, second weight, and third weight may be set to 1.

The expanded data is input as training data for time series learning models such as RNN (Recurrent Neural Network) or LSTM (Long Short Term Memory).

Figure 4 is a diagram illustrating the operation of the failure diagnosis method of an underwater propulsion device according to an embodiment of the present invention to generate failure data through a data expansion model.

When the underwater thruster is driven, the thruster status data is acquired through the measurement sensor (S410). When an abnormal signal occurs from the propeller (S420), the control unit that receives the abnormal signal collects propeller failure data (S440). If no abnormal signal occurs from the thruster, the mission continues (S430).

There are limitations in learning a fault diagnosis model using only fault data collected through the operation of an underwater propulsion machine.

In order to solve the data imbalance problem, which is a limitation of the fault diagnosis model, the GAN (Generative Adversarial Network) model, one of the unsupervised learning models, is used to expand the fault data and use it as learning data. Extended data verification can use Fast-Fourier Transform (FFT), which is one of the discrete Fourier transform algorithms.

The failure diagnosis model can be learned by expanding floating object entanglement failure data using a data expansion model based on GAN (Generative Adversarial Network) and receiving the expanded floating object entanglement failure data as input.

The failure diagnosis model can be learned by expanding blade breakage failure data using a data expansion model based on GAN (Generative Adversarial Network) and receiving the expanded blade breakage failure data as input.

The data expansion model is learned by connecting a generator and a discriminator. The generator may include a first generator and a second generator, and the discriminator may include a first discriminator and a second discriminator.

The first generator generates floating object entanglement failure data (S451), and the first discriminator may determine the similarity of the generated floating object entanglement failure data (S461).

The second generator generates blade breakage failure data (S452), and the second discriminator may determine the similarity of the generated blade breakage failure data (S462).

The data expansion model may be learned such that (i) a first loss function for the first generator and the first discriminator and (ii) a second loss function for the second generator and the second discriminator are minimized. The data expansion model optimizes the model based on the loss function to complete the generation of expanded failure data (S470).

The expanded failure data is used as training data for RNN (Recurrent Neural Network) or LSTM (Long Short Term Memory), deep learning models suitable for processing real-time data. The learned model is used as an internal failure diagnosis algorithm of the underwater vehicle to predict the state of the underwater vehicle through status data measured in real time, and when a failure occurs, whether floating objects are entangled in the propeller or the blade is damaged by external factors. Diagnose.

Depending on the type and state of the failure, the underwater vehicle's ability to perform its mission is determined, and if it is unable to perform its mission, it temporarily stops and rises to sea level. The floating underwater vehicle transmits a signal to the recovery ship through GPS, etc., and the recovery ship that receives the signal recovers the floating underwater vehicle in the relevant sea area, corrects the malfunction, and then completes the mission again.

Figure 5 is a diagram illustrating a failure diagnosis device for an underwater propulsion engine according to another embodiment of the present invention.

The fault diagnosis device includes an underwater thruster, a measurement sensor, and a control unit.

The underwater thruster is implemented to move the underwater vehicle. The underwater propulsion device includes a body 510 installed at the rear end of the underwater vehicle, a power transmission shaft 520 that is inserted into the body and rotates, a blade 530 connected to the power transmission shaft to generate propulsion in an underwater environment, and a power transmission shaft. It includes a motor 540 that transmits power, a power source 550 that supplies power to the motor, and a power supply line 560 that transmits power from the power source to each component.

The measurement sensor is attached to the underwater thruster and collects measurement data. The measurement sensor includes an acceleration sensor or vibration sensor 515, 525, an encoder sensor 545, a Hall sensor 565, etc.

The control unit 500 includes a failure diagnosis model that inputs measurement data to diagnose the state of the underwater thruster and outputs a normal state or failure state, and transmits a control command to control the operation of the underwater vehicle according to the state of the underwater thruster.

The measurement sensor can measure current consumption through a Hall sensor 565 attached to the power supply line 560 of the underwater propulsion machine, and transmit the measured power consumption to the fault diagnosis model.

The measurement sensor measures thruster vibration data through an acceleration sensor or vibration sensor (515, 525) attached to the body 510 or power transmission shaft 520 of the underwater thruster, and transmits the measured thruster vibration data to a fault diagnosis model. You can.

The measurement sensor can measure motor rotation data through the encoder sensor 545 attached to the motor 540 of the underwater propulsion machine and transmit the measured motor rotation data to the fault diagnosis model.

Normal and faulty data sets measured by multiple underwater vehicle operations are arranged at least twice the sampling frequency of the measurement sensor based on the Nyquist-Shannon Sampling Theorem. The number of fault data is very small compared to normal data, and to solve the data imbalance problem, the arrayed fault data is expanded and used.

The fault diagnosis device may include at least one processor, a computer-readable storage medium, and a communication bus.

The processor can control the fault diagnosis device to operate. For example, the processor may execute one or more programs stored on a computer-readable storage medium. One or more programs may include one or more computer-executable instructions, which, when executed by a processor, may be configured to cause a fault diagnosis device to perform operations according to example embodiments.

A computer-readable storage medium is configured to store computer-executable instructions or program code, program data, and/or other suitable form of information. Computer-executable instructions, program code, program data, and/or other suitable forms of information may also be provided through an input/output interface or communication interface. A program stored on a computer-readable storage medium includes a set of instructions executable by a processor. In one embodiment, the computer-readable storage medium includes memory (volatile memory, such as random access memory, non-volatile memory, or an appropriate combination thereof), one or more magnetic disk storage devices, optical disk storage devices, flash memory devices. , other types of storage media that can be accessed by the integrated test system and store desired information, or a suitable combination thereof.

A communication bus interconnects various other components of an information processing device, including a processor and computer-readable storage media.

The fault diagnosis device may also include one or more input/output interfaces and one or more communication interfaces that provide interfaces for one or more input/output devices. The input/output interface and communication interface are connected to a communication bus. The input/output device may be connected to other components of the fault diagnosis device through an input/output interface.

Each device may be implemented within a logic circuit using hardware, firmware, software, or a combination thereof, and may also be implemented using a general-purpose or special-purpose computer. The device may be implemented using hardwired devices, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), etc. Additionally, the device may be implemented as a System on Chip (SoC) including one or more processors and a controller.

Each device may be mounted on a computing device or server equipped with hardware elements in the form of software, hardware, or a combination thereof. A computing device or server includes all or part of a communication device such as a communication modem for communicating with various devices or wired and wireless communication networks, a memory for storing data to execute a program, and a microprocessor for executing a program to perform calculations and commands. It can refer to a variety of devices, including:

In FIGS. 2B, 3, and 4, each process is described as being sequentially executed, but this is merely an illustrative explanation, and those skilled in the art will be able to perform the procedures in FIGS. 2B and 4 without departing from the essential characteristics of the embodiments of the present invention. Various modifications and modifications may be made by executing by changing the order shown in FIGS. 3 and 4, executing one or more processes in parallel, or adding other processes.

Operations according to the present embodiments may be implemented in the form of program instructions that can be performed through various computer means and recorded on a computer-readable medium. Computer-readable media refers to any media that participates in providing instructions to a processor for execution. Computer-readable media may include program instructions, data files, data structures, or combinations thereof. For example, there may be magnetic media, optical recording media, memory, etc. A computer program may be distributed over networked computer systems so that computer-readable code can be stored and executed in a distributed manner. Functional programs, codes, and code segments for implementing this embodiment can be easily deduced by programmers in the technical field to which this embodiment belongs.

These embodiments are intended to explain the technical idea of the present embodiment, and the scope of the technical idea of the present embodiment is not limited by these examples. The scope of protection of this embodiment should be interpreted in accordance with the claims below, and all technical ideas within the equivalent scope should be interpreted as being included in the scope of rights of this embodiment.

500: Control unit
510: body
515: Acceleration sensor
520: Power transmission shaft
525: Vibration sensor

Claims (12)

In the method of diagnosing a failure of an underwater propulsion machine,
Moving an underwater vehicle according to an underwater navigation algorithm;
Collecting measurement data of the underwater propulsion device mounted on the underwater vehicle using a measurement sensor;
Diagnosing the state of the underwater thruster by inputting the measurement data based on a pre-learned failure diagnosis model and outputting a normal state or a failure state; and
A method of diagnosing a failure of an underwater thruster, comprising the step of controlling the operation of the underwater moving object according to the state of the underwater thruster. According to paragraph 1,
The step of outputting the failure state divides the failure state into a state of entangled floating matter or a broken blade state,
The step of controlling the operation of the underwater vehicle is,
If the state of the underwater thruster is in a normal state, the movement of the underwater mobile body proceeds,
If the state of the underwater propeller is in the state of entanglement of the floating object, the underwater mobile body stops moving and rises to the sea surface,
A failure diagnosis method for an underwater thruster, characterized in that, when the state of the underwater thruster is a damaged state of the blade, the operation is subdivided according to the degree of damage. According to paragraph 2,
The failure diagnosis model is learned by receiving floating object entanglement failure data and outputs the floating object entanglement state,
The failure diagnosis model is learned by receiving blade damage failure data, outputs the blade damage state, and compares it to a set damage standard value to classify and output damage grades related to the blade damage state. . According to paragraph 3,
The failure diagnosis model is characterized in that the floating object entanglement state or the blade damage state is output by applying RNN (Recurrent Neural Network) or LSTM (Long Short Term Memory) that analyzes time series data. A method for diagnosing a failure of an underwater propulsion machine. According to paragraph 3,
The fault diagnosis model is,
Setting the number of data between normal data and failure data regarding the underwater propulsion machine as a first ratio,
Set the number of data between the floating object entanglement failure data and the blade breakage failure data as a second ratio,
A failure diagnosis method for an underwater propulsion machine, characterized in that the ratio of data input to the failure diagnosis model is adjusted by applying a first weight to the first ratio and a second weight to the second ratio. According to paragraph 3,
The failure diagnosis model is learned by expanding the floating object entanglement failure data using a GAN (Generative Adversarial Network)-based data expansion model and receiving the expanded floating object entanglement failure data as input,
The failure diagnosis model is a failure diagnosis method of an underwater propulsion machine, characterized in that the blade breakage failure data is expanded using the data expansion model and the expanded blade breakage failure data is input and learned. According to clause 6,
The data expansion model is learned by connecting a generator and a discriminator,
The constructor includes a first constructor and a second constructor,
The discriminator includes a first discriminator and a second discriminator,
The first generator generates floating matter entanglement failure data,
The second generator generates blade breakage failure data,
The first discriminator determines the similarity of the generated floating object entanglement failure data,
The second discriminator determines the similarity of the generated blade damage failure data,
In water, characterized in that (i) a first loss function for the first generator and the first discriminator and (ii) a second loss function for the second generator and the second discriminator are learned to be minimum. How to diagnose propulsion failure. In the fault diagnosis device,
An underwater propulsion device implemented to move an underwater vehicle;
a measurement sensor attached to the underwater propulsion device to collect measurement data; and
A fault diagnosis device including a fault diagnosis model that inputs the measurement data to diagnose the state of the underwater thruster and outputs a normal state or a failure state, and includes a control unit that controls the operation of the underwater moving object according to the state of the underwater thruster. . According to clause 8,
The failure state is divided into a floating matter entanglement state or a blade damage state,
The failure diagnosis model is learned by receiving floating object entanglement failure data and outputs the floating object entanglement state,
The failure diagnosis model is a failure diagnosis device characterized in that it is learned by receiving blade damage failure data and outputs the blade damage state. According to clause 8,
The measurement sensor measures current consumption through a Hall sensor attached to the power supply line of the underwater propulsion machine, and transmits the measured power consumption to the fault diagnosis model. According to clause 8,
The measurement sensor measures thruster vibration data through an acceleration sensor or vibration sensor attached to the body or power transmission shaft of the underwater thruster, and transmits the measured thruster vibration data to the fault diagnosis model. Device. According to clause 8,
The measurement sensor measures motor rotation data through an encoder sensor attached to the motor of the underwater propulsion device, and transmits the measured motor rotation data to the failure diagnosis model.

Images